refactor(project): apply simplify review

- Fix prunable parser to match 'prunable <reason>' lines, not just bare 'prunable'.
- Parallelize Effect.forEach in Worktree.list with concurrency: 'unbounded' to match the rest of the file.
- Simplify the experimental worktree GET handler to use Effect.fn(fn, Effect.map(...)).
- Drop unused isProjectList/isProjectUpdate type guards and the unknown-cast set/setBootStore wrappers in global-sync; expose setGlobalStore directly.
- Centralize project worktree mutations as globalSync.project.addWorktree / removeWorktree, replacing three near-duplicate produce blocks in pages/layout.tsx and prompt-input/submit.ts. Use pathKey for dedup in all sites.
- Cast project.updated event properties as Project (not ProjectInfo) and seed worktrees: [] on insert.
- Extract projectDirectories(project) helper in pages/layout/helpers.ts; use at six call sites that built [project.worktree, ...(project.worktrees ?? [])] inline.

bun.lock

@@ -29,7 +29,7 @@
     },
     "packages/app": {
       "name": "@opencode-ai/app",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@kobalte/core": "catalog:",
         "@opencode-ai/core": "workspace:*",
@@ -85,7 +85,7 @@
     },
     "packages/console/app": {
       "name": "@opencode-ai/console-app",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@cloudflare/vite-plugin": "1.15.2",
         "@ibm/plex": "6.4.1",
@@ -120,7 +120,7 @@
     },
     "packages/console/core": {
       "name": "@opencode-ai/console-core",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@aws-sdk/client-sts": "3.782.0",
         "@jsx-email/render": "1.1.1",
@@ -147,7 +147,7 @@
     },
     "packages/console/function": {
       "name": "@opencode-ai/console-function",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@ai-sdk/anthropic": "3.0.64",
         "@ai-sdk/openai": "3.0.48",
@@ -171,7 +171,7 @@
     },
     "packages/console/mail": {
       "name": "@opencode-ai/console-mail",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@jsx-email/all": "2.2.3",
         "@jsx-email/cli": "1.4.3",
@@ -195,7 +195,7 @@
     },
     "packages/core": {
       "name": "@opencode-ai/core",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "bin": {
         "opencode": "./bin/opencode",
       },
@@ -229,7 +229,7 @@
     },
     "packages/desktop": {
       "name": "@opencode-ai/desktop",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "drizzle-orm": "catalog:",
         "effect": "catalog:",
@@ -283,7 +283,7 @@
     },
     "packages/enterprise": {
       "name": "@opencode-ai/enterprise",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@opencode-ai/core": "workspace:*",
         "@opencode-ai/ui": "workspace:*",
@@ -313,7 +313,7 @@
     },
     "packages/function": {
       "name": "@opencode-ai/function",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@octokit/auth-app": "8.0.1",
         "@octokit/rest": "catalog:",
@@ -329,7 +329,7 @@
     },
     "packages/http-recorder": {
       "name": "@opencode-ai/http-recorder",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@effect/platform-node": "catalog:",
         "effect": "catalog:",
@@ -342,7 +342,7 @@
     },
     "packages/llm": {
       "name": "@opencode-ai/llm",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@smithy/eventstream-codec": "4.2.14",
         "@smithy/util-utf8": "4.2.2",
@@ -360,7 +360,7 @@
     },
     "packages/opencode": {
       "name": "opencode",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "bin": {
         "opencode": "./bin/opencode",
       },
@@ -496,7 +496,7 @@
     },
     "packages/plugin": {
       "name": "@opencode-ai/plugin",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@opencode-ai/sdk": "workspace:*",
         "effect": "catalog:",
@@ -534,7 +534,7 @@
     },
     "packages/sdk/js": {
       "name": "@opencode-ai/sdk",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "cross-spawn": "catalog:",
       },
@@ -549,7 +549,7 @@
     },
     "packages/slack": {
       "name": "@opencode-ai/slack",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@opencode-ai/sdk": "workspace:*",
         "@slack/bolt": "^3.17.1",
@@ -584,7 +584,7 @@
     },
     "packages/ui": {
       "name": "@opencode-ai/ui",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@kobalte/core": "catalog:",
         "@opencode-ai/core": "workspace:*",
@@ -633,7 +633,7 @@
     },
     "packages/web": {
       "name": "@opencode-ai/web",
-      "version": "1.14.46",
+      "version": "1.14.48",
       "dependencies": {
         "@astrojs/cloudflare": "12.6.3",
         "@astrojs/markdown-remark": "6.3.1",

infra/monitoring.ts

@@ -1,6 +1,8 @@
 import { SECRET } from "./secret"
 import { domain } from "./stage"
 
+const description = "Managed by SST (Don't edit in Honeycomb UI)"
+
 const webhookRecipient = new honeycomb.WebhookRecipient("DiscordAlerts", {
   name: $app.stage === "production" ? "Discord Alerts" : `Discord Alerts (${$app.stage})`,
   url: `https://${domain}/honeycomb/webhook`,
@@ -25,6 +27,16 @@ const webhookRecipient = new honeycomb.WebhookRecipient("DiscordAlerts", {
   ],
 })
 
+// Honeycomb can keep stale query-local calculated fields when the name is unchanged,
+// so tie the field name to the expression while avoiding deploy-to-deploy churn.
+// https://github.com/honeycombio/terraform-provider-honeycombio/issues/852
+const calculatedField = (field: { name: string; expression: string }) => ({
+  ...field,
+  name: `${field.name}_${(
+    Array.from(field.expression).reduce((result, char) => Math.imul(31, result) + char.charCodeAt(0), 0) >>> 0
+  ).toString(36)}`,
+})
+
 const modelHttpErrorsQuery = (product: "go" | "zen") => {
   const filters = [
     { column: "model", op: "exists" },
@@ -32,21 +44,26 @@ const modelHttpErrorsQuery = (product: "go" | "zen") => {
     { column: "user_agent", op: "contains", value: "opencode" },
     { column: "isGoTier", op: "=", value: product === "go" ? "true" : "false" },
   ]
+  const failedHttpStatus = calculatedField({
+    name: "is_failed_http_status",
+    expression:
+      product === "go"
+        ? `IF(AND(GTE($status, "400"), NOT(EQUALS($status, "401")), NOT(EQUALS($status, "429"))), 1, 0)`
+        : `IF(AND(EQUALS($status, "429"), $isFreeTier), 0, AND(GTE($status, "400"), NOT(EQUALS($status, "401"))), 1, 0)`,
+  })
 
   return honeycomb.getQuerySpecificationOutput({
     breakdowns: ["model"],
-    calculatedFields: [
-      {
-        name: "is_failed_http_status",
-        expression:
-          product === "go"
-            ? `IF(AND(GTE($status, "400"), NOT(EQUALS($status, "401")), NOT(EQUALS($status, "429"))), 1, 0)`
-            : `IF(AND(GTE($status, "400"), NOT(EQUALS($status, "401"))), 1, 0)`,
-      },
-    ],
+    calculatedFields: [failedHttpStatus],
     calculations: [
       { op: "COUNT", name: "TOTAL", filterCombination: "AND", filters },
-      { op: "SUM", name: "FAILED", column: "is_failed_http_status", filterCombination: "AND", filters },
+      {
+        op: "SUM",
+        name: "FAILED",
+        column: failedHttpStatus.name,
+        filterCombination: "AND",
+        filters,
+      },
     ],
     formulas: [{ name: "ERROR", expression: "IF(GTE($TOTAL, 100), DIV($FAILED, $TOTAL), 0)" }],
     timeRange: 900,
@@ -59,31 +76,30 @@ const providerHttpErrorsQuery = (product: "go" | "zen") => {
     { column: "user_agent", op: "contains", value: "opencode" },
     { column: "isGoTier", op: "=", value: product === "go" ? "true" : "false" },
   ]
+  const successHttpStatus = calculatedField({
+    name: "is_success_http_status",
+    expression: `IF(AND(GTE($status, "200"), LT($status, "400")), 1, 0)`,
+  })
+  const failedProviderHttpStatus = calculatedField({
+    name: "is_failed_provider_http_status",
+    expression: `IF(GT($llm.error.code, "400"), 1, 0)`,
+  })
 
   return honeycomb.getQuerySpecificationOutput({
     breakdowns: ["provider"],
-    calculatedFields: [
-      {
-        name: "is_success_http_status",
-        expression: `IF(AND(GTE($status, "200"), LT($status, "400")), 1, 0)`,
-      },
-      {
-        name: "is_failed_provider_http_status",
-        expression: `IF(GT($llm.error.code, "400"), 1, 0)`,
-      },
-    ],
+    calculatedFields: [successHttpStatus, failedProviderHttpStatus],
     calculations: [
       {
         op: "SUM",
         name: "SUCCESS",
-        column: "is_success_http_status",
+        column: successHttpStatus.name,
         filterCombination: "AND",
         filters: [...filters, { column: "event_type", op: "=", value: "completions" }],
       },
       {
         op: "SUM",
         name: "FAILED",
-        column: "is_failed_provider_http_status",
+        column: failedProviderHttpStatus.name,
         filterCombination: "AND",
         filters: [...filters, { column: "event_type", op: "=", value: "llm.error" }],
       },
@@ -95,8 +111,6 @@ const providerHttpErrorsQuery = (product: "go" | "zen") => {
   }).json
 }
 
-const description = "Managed by SST (Don't edit in Honeycomb UI)"
-
 new honeycomb.Trigger("IncreasedModelHttpErrorsGo", {
   name: "Increased Model HTTP Errors [Go]",
   description,

packages/app/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/app",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "description": "",
   "type": "module",
   "exports": {

packages/app/src/components/dialog-select-directory.tsx

@@ -10,6 +10,7 @@ import { useGlobalSDK } from "@/context/global-sdk"
 import { useGlobalSync } from "@/context/global-sync"
 import { useLayout } from "@/context/layout"
 import { useLanguage } from "@/context/language"
+import { projectDirectories } from "@/pages/layout/helpers"
 
 interface DialogSelectDirectoryProps {
   title?: string
@@ -284,7 +285,7 @@ export function DialogSelectDirectory(props: DialogSelectDirectoryProps) {
 
     for (const project of projects) {
       let at = 0
-      const dirs = [project.worktree, ...(project.sandboxes ?? [])]
+      const dirs = projectDirectories(project)
       for (const directory of dirs) {
         const sessions = sync.child(directory, { bootstrap: false })[0].session
         for (const session of sessions) {

packages/app/src/components/dialog-select-file.tsx

@@ -16,6 +16,7 @@ import { useFile } from "@/context/file"
 import { useLanguage } from "@/context/language"
 import { useSessionLayout } from "@/pages/session/session-layout"
 import { createSessionTabs } from "@/pages/session/helpers"
+import { projectDirectories } from "@/pages/layout/helpers"
 import { decode64 } from "@/utils/base64"
 import { getRelativeTime } from "@/utils/time"
 
@@ -280,14 +281,14 @@ export function DialogSelectFile(props: { mode?: DialogSelectFileMode; onOpenFil
   const project = createMemo(() => {
     const directory = projectDirectory()
     if (!directory) return
-    return layout.projects.list().find((p) => p.worktree === directory || p.sandboxes?.includes(directory))
+    return layout.projects.list().find((p) => p.worktree === directory || p.worktrees?.includes(directory))
   })
   const workspaces = createMemo(() => {
     const directory = projectDirectory()
     const current = project()
     if (!current) return directory ? [directory] : []
 
-    const dirs = [current.worktree, ...(current.sandboxes ?? [])]
+    const dirs = projectDirectories(current)
     if (directory && !dirs.includes(directory)) return [...dirs, directory]
     return dirs
   })

packages/app/src/components/prompt-input/submit.ts

@@ -342,6 +342,7 @@ export function createPromptSubmit(input: PromptSubmitInput) {
           return
         }
         WorktreeState.pending(createdWorktree.directory)
+        globalSync.project.addWorktree(projectDirectory, createdWorktree.directory)
         sessionDirectory = createdWorktree.directory
       }
 

packages/app/src/components/session/session-header.tsx

@@ -144,7 +144,7 @@ export function SessionHeader() {
   const project = createMemo(() => {
     const directory = projectDirectory()
     if (!directory) return
-    return layout.projects.list().find((p) => p.worktree === directory || p.sandboxes?.includes(directory))
+    return layout.projects.list().find((p) => p.worktree === directory || p.worktrees?.includes(directory))
   })
   const name = createMemo(() => {
     const current = project()

packages/app/src/components/session/session-new-view.tsx

@@ -20,8 +20,8 @@ export function NewSessionView(props: NewSessionViewProps) {
   const sdk = useSDK()
   const language = useLanguage()
 
-  const sandboxes = createMemo(() => sync.project?.sandboxes ?? [])
-  const options = createMemo(() => [MAIN_WORKTREE, ...sandboxes(), CREATE_WORKTREE])
+  const worktrees = createMemo(() => sync.project?.worktrees ?? [])
+  const options = createMemo(() => [MAIN_WORKTREE, ...worktrees(), CREATE_WORKTREE])
   const current = createMemo(() => {
     const selection = props.worktree
     if (options().includes(selection)) return selection

packages/app/src/context/global-sync.tsx

@@ -2,7 +2,6 @@ import type {
   Config,
   OpencodeClient,
   Path,
-  Project,
   ProviderAuthResponse,
   ProviderListResponse,
   Todo,
@@ -27,18 +26,19 @@ import { applyDirectoryEvent, applyGlobalEvent, cleanupDroppedSessionCaches } fr
 import { clearSessionPrefetchDirectory } from "./global-sync/session-prefetch"
 import { estimateRootSessionTotal, loadRootSessionsWithFallback } from "./global-sync/session-load"
 import { trimSessions } from "./global-sync/session-trim"
-import type { ProjectMeta } from "./global-sync/types"
+import type { ProjectInfo, ProjectMeta } from "./global-sync/types"
 import { SESSION_RECENT_LIMIT } from "./global-sync/types"
 import { formatServerError } from "@/utils/server-errors"
 import { queryOptions, useMutation, useQueries, useQuery, useQueryClient } from "@tanstack/solid-query"
 import { createRefreshQueue } from "./global-sync/queue"
 import { directoryKey } from "./global-sync/utils"
+import { pathKey } from "@/utils/path-key"
 
 type GlobalStore = {
   ready: boolean
   error?: InitError
   path: Path
-  project: Project[]
+  project: ProjectInfo[]
   session_todo: {
     [sessionID: string]: Todo[]
   }
@@ -122,18 +122,6 @@ function createGlobalSync() {
     if (eventTimer !== undefined) clearTimeout(eventTimer)
   })
 
-  const setProjects = (next: Project[] | ((draft: Project[]) => Project[])) => {
-    setGlobalStore("project", next)
-  }
-
-  const setBootStore = ((...input: unknown[]) => {
-    if (input[0] === "project" && Array.isArray(input[1])) {
-      setProjects(input[1] as Project[])
-      return input[1]
-    }
-    return (setGlobalStore as (...args: unknown[]) => unknown)(...input)
-  }) as typeof setGlobalStore
-
   const bootstrap = useQuery(() => ({
     queryKey: ["bootstrap"],
     queryFn: async () => {
@@ -142,7 +130,7 @@ function createGlobalSync() {
         requestFailedTitle: language.t("common.requestFailed"),
         translate: language.t,
         formatMoreCount: (count) => language.t("common.moreCountSuffix", { count }),
-        setGlobalStore: setBootStore,
+        setGlobalStore,
         queryClient,
       })
       bootedAt = Date.now()
@@ -150,13 +138,31 @@ function createGlobalSync() {
     },
   }))
 
-  const set = ((...input: unknown[]) => {
-    if (input[0] === "project" && (Array.isArray(input[1]) || typeof input[1] === "function")) {
-      setProjects(input[1] as Project[] | ((draft: Project[]) => Project[]))
-      return input[1]
-    }
-    return (setGlobalStore as (...args: unknown[]) => unknown)(...input)
-  }) as typeof setGlobalStore
+  const set = setGlobalStore
+
+  const addProjectWorktree = (projectWorktree: string, directory: string) => {
+    setGlobalStore(
+      "project",
+      produce((draft) => {
+        const project = draft.find((item) => item.worktree === projectWorktree)
+        if (!project) return
+        const key = pathKey(directory)
+        project.worktrees = [directory, ...(project.worktrees ?? []).filter((item) => pathKey(item) !== key)]
+      }),
+    )
+  }
+
+  const removeProjectWorktree = (projectWorktree: string, directory: string) => {
+    setGlobalStore(
+      "project",
+      produce((draft) => {
+        const project = draft.find((item) => item.worktree === projectWorktree)
+        if (!project) return
+        const key = pathKey(directory)
+        project.worktrees = (project.worktrees ?? []).filter((item) => pathKey(item) !== key)
+      }),
+    )
+  }
 
   const setSessionTodo = (sessionID: string, todos: Todo[] | undefined) => {
     if (!sessionID) return
@@ -344,7 +350,7 @@ function createGlobalSync() {
           if (recent) return
           bootstrap.refetch()
         },
-        setGlobalProject: setProjects,
+        setGlobalProject: (next) => setGlobalStore("project", next),
       })
       if (event.type === "server.connected" || event.type === "global.disposed") {
         if (recent) return
@@ -408,6 +414,8 @@ function createGlobalSync() {
     icon(directory: string, value: string | undefined) {
       children.projectIcon(directory, value)
     },
+    addWorktree: addProjectWorktree,
+    removeWorktree: removeProjectWorktree,
   }
 
   const updateConfigMutation = useMutation(() => ({

packages/app/src/context/global-sync/bootstrap.ts

@@ -3,7 +3,6 @@ import type {
   OpencodeClient,
   Path,
   PermissionRequest,
-  Project,
   ProviderAuthResponse,
   ProviderListResponse,
   QuestionRequest,
@@ -15,7 +14,7 @@ import { getFilename } from "@opencode-ai/core/util/path"
 import { retry } from "@opencode-ai/core/util/retry"
 import { batch } from "solid-js"
 import { reconcile, type SetStoreFunction, type Store } from "solid-js/store"
-import type { State, VcsCache } from "./types"
+import type { ProjectInfo, State, VcsCache } from "./types"
 import { cmp, normalizeAgentList, normalizeProviderList } from "./utils"
 import { formatServerError } from "@/utils/server-errors"
 import { QueryClient, queryOptions } from "@tanstack/solid-query"
@@ -24,7 +23,7 @@ import { loadMcpQuery } from "../global-sync"
 type GlobalStore = {
   ready: boolean
   path: Path
-  project: Project[]
+  project: ProjectInfo[]
   session_todo: {
     [sessionID: string]: Todo[]
   }
@@ -94,12 +93,21 @@ export const loadProjectsQuery = (sdk: OpencodeClient) =>
     queryKey: ["project"],
     queryFn: () =>
       retry(() =>
-        sdk.project.list().then((x) => {
-          return (x.data ?? [])
-            .filter((p) => !!p?.id)
-            .filter((p) => !!p.worktree && !p.worktree.includes("opencode-test"))
-            .slice()
-            .sort((a, b) => cmp(a.id, b.id))
+        sdk.project.list().then(async (x) => {
+          return Promise.all(
+            (x.data ?? [])
+              .filter((p) => !!p?.id)
+              .filter((p) => !!p.worktree && !p.worktree.includes("opencode-test"))
+              .slice()
+              .sort((a, b) => cmp(a.id, b.id))
+              .map(async (project) => ({
+                ...project,
+                worktrees: await sdk.worktree
+                  .list({ directory: project.worktree })
+                  .then((x) => x.data ?? [])
+                  .catch(() => []),
+              })),
+          )
         }),
       ),
   })
@@ -140,8 +148,8 @@ function groupBySession<T extends { id: string; sessionID: string }>(input: T[])
   }, {})
 }
 
-function projectID(directory: string, projects: Project[]) {
-  return projects.find((project) => project.worktree === directory || project.sandboxes?.includes(directory))?.id
+function projectID(directory: string, projects: ProjectInfo[]) {
+  return projects.find((project) => project.worktree === directory || project.worktrees?.includes(directory))?.id
 }
 
 function mergeSession(setStore: SetStoreFunction<State>, session: Session) {
@@ -207,7 +215,7 @@ export async function bootstrapDirectory(input: {
   global: {
     config: Config
     path: Path
-    project: Project[]
+    project: ProjectInfo[]
     provider: ProviderListResponse
   }
   queryClient: QueryClient

packages/app/src/context/global-sync/child-store.ts

@@ -231,6 +231,7 @@ export function createChildStoreManager(input: {
             limit: 5,
             message: {},
             part: {},
+            part_text_accum_delta: {},
           })
           children[key] = child
           disposers.set(key, dispose)

packages/app/src/context/global-sync/event-reducer.test.ts

@@ -81,6 +81,7 @@ const baseState = (input: Partial<State> = {}) =>
     limit: 10,
     message: {},
     part: {},
+    part_text_accum_delta: {},
     ...input,
   }) as State
 

packages/app/src/context/global-sync/event-reducer.ts

@@ -11,7 +11,7 @@ import type {
   SnapshotFileDiff,
   Todo,
 } from "@opencode-ai/sdk/v2/client"
-import type { State, VcsCache } from "./types"
+import type { ProjectInfo, State, VcsCache } from "./types"
 import { trimSessions } from "./session-trim"
 import { dropSessionCaches } from "./session-cache"
 import { diffs as list, message as clean } from "@/utils/diffs"
@@ -20,8 +20,8 @@ const SKIP_PARTS = new Set(["patch", "step-start", "step-finish"])
 
 export function applyGlobalEvent(input: {
   event: { type: string; properties?: unknown }
-  project: Project[]
-  setGlobalProject: (next: Project[] | ((draft: Project[]) => Project[])) => void
+  project: ProjectInfo[]
+  setGlobalProject: (next: ProjectInfo[] | ((draft: ProjectInfo[]) => ProjectInfo[])) => void
   refresh: () => void
 }) {
   if (input.event.type === "global.disposed" || input.event.type === "server.connected") {
@@ -42,7 +42,7 @@ export function applyGlobalEvent(input: {
   }
   input.setGlobalProject(
     produce((draft) => {
-      draft.splice(result.index, 0, properties)
+      draft.splice(result.index, 0, { ...properties, worktrees: [] })
     }),
   )
 }
@@ -211,6 +211,12 @@ export function applyDirectoryEvent(input: {
             const result = Binary.search(messages, props.messageID, (m) => m.id)
             if (result.found) messages.splice(result.index, 1)
           }
+          const parts = draft.part[props.messageID]
+          if (parts) {
+            for (const part of parts) {
+              delete draft.part_text_accum_delta[part.id]
+            }
+          }
           delete draft.part[props.messageID]
         }),
       )
@@ -219,6 +225,11 @@ export function applyDirectoryEvent(input: {
     case "message.part.updated": {
       const part = (event.properties as { part: Part }).part
       if (SKIP_PARTS.has(part.type)) break
+      input.setStore(
+        produce((draft) => {
+          delete draft.part_text_accum_delta[part.id]
+        }),
+      )
       const parts = input.store.part[part.messageID]
       if (!parts) {
         input.setStore("part", part.messageID, [part])
@@ -240,6 +251,11 @@ export function applyDirectoryEvent(input: {
     }
     case "message.part.removed": {
       const props = event.properties as { messageID: string; partID: string }
+      input.setStore(
+        produce((draft) => {
+          delete draft.part_text_accum_delta[props.partID]
+        }),
+      )
       const parts = input.store.part[props.messageID]
       if (!parts) break
       const result = Binary.search(parts, props.partID, (p) => p.id)
@@ -263,6 +279,7 @@ export function applyDirectoryEvent(input: {
       if (!parts) break
       const result = Binary.search(parts, props.partID, (p) => p.id)
       if (!result.found) break
+      input.setStore("part_text_accum_delta", props.partID, (existing) => (existing ?? "") + props.delta)
       input.setStore(
         "part",
         props.messageID,

packages/app/src/context/global-sync/session-cache.test.ts

@@ -39,6 +39,7 @@ describe("app session cache", () => {
       part: Record<string, Part[] | undefined>
       permission: Record<string, PermissionRequest[] | undefined>
       question: Record<string, QuestionRequest[] | undefined>
+      part_text_accum_delta: Record<string, string | undefined>
     } = {
       session_status: { ses_1: { type: "busy" } as SessionStatus },
       session_diff: { ses_1: [] },
@@ -47,12 +48,14 @@ describe("app session cache", () => {
       part: { msg_1: [part("prt_1", "ses_1", "msg_1")] },
       permission: { ses_1: [] as PermissionRequest[] },
       question: { ses_1: [] as QuestionRequest[] },
+      part_text_accum_delta: { prt_1: "streamed text" },
     }
 
     dropSessionCaches(store, ["ses_1"])
 
     expect(store.message.ses_1).toBeUndefined()
     expect(store.part.msg_1).toBeUndefined()
+    expect(store.part_text_accum_delta.prt_1).toBeUndefined()
     expect(store.todo.ses_1).toBeUndefined()
     expect(store.session_diff.ses_1).toBeUndefined()
     expect(store.session_status.ses_1).toBeUndefined()
@@ -70,6 +73,7 @@ describe("app session cache", () => {
       part: Record<string, Part[] | undefined>
       permission: Record<string, PermissionRequest[] | undefined>
       question: Record<string, QuestionRequest[] | undefined>
+      part_text_accum_delta: Record<string, string | undefined>
     } = {
       session_status: {},
       session_diff: {},
@@ -78,6 +82,7 @@ describe("app session cache", () => {
       part: { [m.id]: [part("prt_1", "ses_1", m.id)] },
       permission: {},
       question: {},
+      part_text_accum_delta: {},
     }
 
     dropSessionCaches(store, ["ses_1"])

packages/app/src/context/global-sync/session-cache.ts

@@ -18,6 +18,7 @@ type SessionCache = {
   part: Record<string, Part[] | undefined>
   permission: Record<string, PermissionRequest[] | undefined>
   question: Record<string, QuestionRequest[] | undefined>
+  part_text_accum_delta: Record<string, string | undefined>
 }
 
 export function dropSessionCaches(store: SessionCache, sessionIDs: Iterable<string>) {
@@ -27,6 +28,9 @@ export function dropSessionCaches(store: SessionCache, sessionIDs: Iterable<stri
   for (const key of Object.keys(store.part)) {
     const parts = store.part[key]
     if (!parts?.some((part) => stale.has(part?.sessionID ?? ""))) continue
+    for (const part of parts) {
+      delete store.part_text_accum_delta[part.id]
+    }
     delete store.part[key]
   }
 

packages/app/src/context/global-sync/types.ts

@@ -8,6 +8,7 @@ import type {
   Part,
   Path,
   PermissionRequest,
+  Project,
   ProviderListResponse,
   QuestionRequest,
   Session,
@@ -30,6 +31,10 @@ export type ProjectMeta = {
   }
 }
 
+export type ProjectInfo = Project & {
+  worktrees?: string[]
+}
+
 export type State = {
   status: "loading" | "partial" | "complete"
   agent: Agent[]
@@ -72,6 +77,9 @@ export type State = {
   part: {
     [messageID: string]: Part[]
   }
+  part_text_accum_delta: {
+    [partID: string]: string
+  }
 }
 
 export type VcsCache = {

packages/app/src/context/global-sync/utils.ts

@@ -1,4 +1,5 @@
-import type { Agent, Project, ProviderListResponse } from "@opencode-ai/sdk/v2/client"
+import type { Agent, ProviderListResponse } from "@opencode-ai/sdk/v2/client"
+import type { ProjectInfo } from "./types"
 export { pathKey as directoryKey, type PathKey as DirectoryKey } from "@/utils/path-key"
 
 export const cmp = (a: string, b: string) => (a < b ? -1 : a > b ? 1 : 0)
@@ -27,7 +28,7 @@ export function normalizeProviderList(input: ProviderListResponse): ProviderList
   }
 }
 
-export function sanitizeProject(project: Project) {
+export function sanitizeProject(project: ProjectInfo) {
   if (!project.icon?.url && !project.icon?.override) return project
   return {
     ...project,

packages/app/src/context/layout.tsx

@@ -6,8 +6,8 @@ import { useGlobalSync } from "./global-sync"
 import { useGlobalSDK } from "./global-sdk"
 import { useServer } from "./server"
 import { usePlatform } from "./platform"
-import { Project } from "@opencode-ai/sdk/v2"
 import { Persist, persisted, removePersisted } from "@/utils/persist"
+import type { ProjectInfo } from "./global-sync/types"
 import { decode64 } from "@/utils/base64"
 import { same } from "@/utils/same"
 import { createScrollPersistence, type SessionScroll } from "./layout-scroll"
@@ -51,7 +51,7 @@ type TabHandoff = {
   at: number
 }
 
-export type LocalProject = Partial<Project> & { worktree: string; expanded: boolean }
+export type LocalProject = Partial<ProjectInfo> & { worktree: string; expanded: boolean }
 
 export type ReviewDiffStyle = "unified" | "split"
 
@@ -404,9 +404,8 @@ export const { use: useLayout, provider: LayoutProvider } = createSimpleContext(
     const roots = createMemo(() => {
       const map = new Map<string, string>()
       for (const project of globalSync.data.project) {
-        const sandboxes = project.sandboxes ?? []
-        for (const sandbox of sandboxes) {
-          map.set(sandbox, project.worktree)
+        for (const worktree of project.worktrees ?? []) {
+          map.set(worktree, project.worktree)
         }
       }
       return map

packages/app/src/pages/layout.tsx

@@ -70,6 +70,7 @@ import {
   effectiveWorkspaceOrder,
   errorMessage,
   latestRootSession,
+  projectDirectories,
   sortedRootSessions,
 } from "./layout/helpers"
 import {
@@ -560,8 +561,8 @@ export default function Layout(props: ParentProps) {
 
     const projects = layout.projects.list()
 
-    const sandbox = projects.find((p) => p.sandboxes?.some((item) => pathKey(item) === key))
-    if (sandbox) return sandbox
+    const worktree = projects.find((p) => p.worktrees?.some((item) => pathKey(item) === key))
+    if (worktree) return worktree
 
     const direct = projects.find((p) => pathKey(p.worktree) === key)
     if (direct) return direct
@@ -646,7 +647,7 @@ export default function Layout(props: ParentProps) {
       if (!expanded) continue
       const key = pathKey(directory)
       const project = projects.find(
-        (item) => pathKey(item.worktree) === key || item.sandboxes?.some((sandbox) => pathKey(sandbox) === key),
+        (item) => pathKey(item.worktree) === key || item.worktrees?.some((worktree) => pathKey(worktree) === key),
       )
       if (!project) continue
       if (project.vcs === "git" && layout.sidebar.workspaces(project.worktree)()) continue
@@ -1223,7 +1224,7 @@ export default function Layout(props: ParentProps) {
     const key = pathKey(directory)
     const project = layout.projects
       .list()
-      .find((item) => pathKey(item.worktree) === key || item.sandboxes?.some((sandbox) => pathKey(sandbox) === key))
+      .find((item) => pathKey(item.worktree) === key || item.worktrees?.some((worktree) => pathKey(worktree) === key))
     if (project) return project.worktree
 
     const known = Object.entries(store.workspaceOrder).find(
@@ -1274,9 +1275,7 @@ export default function Layout(props: ParentProps) {
     const root = projectRoot(directory)
     server.projects.touch(root)
     const project = layout.projects.list().find((item) => item.worktree === root)
-    let dirs = project
-      ? effectiveWorkspaceOrder(root, [root, ...(project.sandboxes ?? [])], store.workspaceOrder[root])
-      : [root]
+    let dirs = project ? effectiveWorkspaceOrder(root, projectDirectories(project), store.workspaceOrder[root]) : [root]
     const canOpen = (value: string | undefined) => {
       if (!value) return false
       return dirs.some((item) => pathKey(item) === pathKey(value))
@@ -1509,14 +1508,7 @@ export default function Layout(props: ParentProps) {
       clearLastProjectSession(root)
     }
 
-    globalSync.set(
-      "project",
-      produce((draft) => {
-        const project = draft.find((item) => item.worktree === root)
-        if (!project) return
-        project.sandboxes = (project.sandboxes ?? []).filter((sandbox) => sandbox !== directory)
-      }),
-    )
+    globalSync.project.removeWorktree(root, directory)
     setStore("workspaceOrder", root, (order) => (order ?? []).filter((workspace) => workspace !== directory))
 
     layout.projects.close(directory)
@@ -1528,7 +1520,7 @@ export default function Layout(props: ParentProps) {
     const nextKey = pathKey(nextCurrent)
     const project = layout.projects.list().find((item) => item.worktree === root)
     const dirs = project
-      ? effectiveWorkspaceOrder(root, [root, ...(project.sandboxes ?? [])], store.workspaceOrder[root])
+      ? effectiveWorkspaceOrder(root, projectDirectories(project), store.workspaceOrder[root])
       : [root]
     const valid = dirs.some((item) => pathKey(item) === nextKey)
 
@@ -1862,7 +1854,7 @@ export default function Layout(props: ParentProps) {
   function workspaceIds(project: LocalProject | undefined) {
     if (!project) return []
     const local = project.worktree
-    const dirs = [local, ...(project.sandboxes ?? [])]
+    const dirs = projectDirectories(project)
     const active = currentProject()
     const directory = pathKey(active?.worktree ?? "") === pathKey(project.worktree) ? currentDir() : undefined
     const extra =
@@ -1954,6 +1946,7 @@ export default function Layout(props: ParentProps) {
       })
       return [created.directory, ...next]
     })
+    globalSync.project.addWorktree(project.worktree, created.directory)
 
     globalSync.child(created.directory)
     navigateWithSidebarReset(`/${base64Encode(created.directory)}/session`)

packages/app/src/pages/layout/helpers.ts

@@ -55,6 +55,11 @@ export const childSessionOnPath = (sessions: Session[] | undefined, rootID: stri
 export const displayName = (project: { name?: string; worktree: string }) =>
   project.name || getFilename(project.worktree)
 
+export const projectDirectories = (project: { worktree: string; worktrees?: string[] }) => [
+  project.worktree,
+  ...(project.worktrees ?? []),
+]
+
 export const errorMessage = (err: unknown, fallback: string) => {
   if (err && typeof err === "object" && "data" in err) {
     const data = (err as { data?: { message?: string } }).data

packages/app/src/pages/layout/sidebar-items.tsx

@@ -15,7 +15,7 @@ import { usePermission } from "@/context/permission"
 import { messageAgentColor } from "@/utils/agent"
 import { sessionTitle } from "@/utils/session-title"
 import { sessionPermissionRequest } from "../session/composer/session-request-tree"
-import { childSessionOnPath, hasProjectPermissions } from "./helpers"
+import { childSessionOnPath, hasProjectPermissions, projectDirectories } from "./helpers"
 
 const OPENCODE_PROJECT_ID = "4b0ea68d7af9a6031a7ffda7ad66e0cb83315750"
 
@@ -35,7 +35,7 @@ export const ProjectIcon = (props: {
   const globalSync = useGlobalSync()
   const notification = useNotification()
   const permission = usePermission()
-  const dirs = createMemo(() => [props.project.worktree, ...(props.project.sandboxes ?? [])])
+  const dirs = createMemo(() => projectDirectories(props.project))
   const unseenCount = createMemo(() =>
     dirs().reduce((total, directory) => total + notification.project.unseenCount(directory), 0),
   )

packages/console/app/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/console-app",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "type": "module",
   "license": "MIT",
   "scripts": {

packages/console/app/src/routes/workspace/[id]/usage/usage-section.module.css

@@ -67,6 +67,7 @@
         display: inline-flex;
         align-items: center;
         justify-content: center;
+        flex-shrink: 0;
         padding: 0;
         background: transparent;
         border: none;
@@ -79,6 +80,7 @@
         }
 
         svg {
+          flex-shrink: 0;
           width: 16px;
           height: 16px;
         }

packages/console/app/src/routes/workspace/[id]/usage/usage-section.tsx

@@ -53,7 +53,7 @@ export function UsageSection() {
   }
 
   const calculateTotalOutputTokens = (u: Awaited<ReturnType<typeof getUsageInfo>>[0]) => {
-    return u.outputTokens + (u.reasoningTokens ?? 0)
+    return u.outputTokens
   }
 
   const goPrev = async () => {

packages/console/app/src/routes/zen/util/handler.ts

@@ -889,10 +889,6 @@ export async function handler(
 
     const inputCost = modelCost.input * inputTokens * 100
     const outputCost = modelCost.output * outputTokens * 100
-    const reasoningCost = (() => {
-      if (!reasoningTokens) return undefined
-      return modelCost.output * reasoningTokens * 100
-    })()
     const cacheReadCost = (() => {
       if (!cacheReadTokens) return undefined
       if (!modelCost.cacheRead) return undefined
@@ -909,17 +905,11 @@ export async function handler(
       return modelCost.cacheWrite1h * cacheWrite1hTokens * 100
     })()
     const totalCostInCent =
-      inputCost +
-      outputCost +
-      (reasoningCost ?? 0) +
-      (cacheReadCost ?? 0) +
-      (cacheWrite5mCost ?? 0) +
-      (cacheWrite1hCost ?? 0)
+      inputCost + outputCost + (cacheReadCost ?? 0) + (cacheWrite5mCost ?? 0) + (cacheWrite1hCost ?? 0)
     return {
       totalCostInCent,
       inputCost,
       outputCost,
-      reasoningCost,
       cacheReadCost,
       cacheWrite5mCost,
       cacheWrite1hCost,
@@ -941,8 +931,7 @@ export async function handler(
   ) {
     const { inputTokens, outputTokens, reasoningTokens, cacheReadTokens, cacheWrite5mTokens, cacheWrite1hTokens } =
       usageInfo
-    const { totalCostInCent, inputCost, outputCost, reasoningCost, cacheReadCost, cacheWrite5mCost, cacheWrite1hCost } =
-      costInfo
+    const { totalCostInCent, inputCost, outputCost, cacheReadCost, cacheWrite5mCost, cacheWrite1hCost } = costInfo
 
     logger.metric({
       "tokens.input": inputTokens,
@@ -953,14 +942,12 @@ export async function handler(
       "tokens.cache_write_1h": cacheWrite1hTokens,
       "cost.input.microcents": centsToMicroCents(inputCost),
       "cost.output.microcents": centsToMicroCents(outputCost),
-      "cost.reasoning.microcents": reasoningCost ? centsToMicroCents(reasoningCost) : undefined,
       "cost.cache_read.microcents": cacheReadCost ? centsToMicroCents(cacheReadCost) : undefined,
       "cost.cache_write.microcents": cacheWrite5mCost ? centsToMicroCents(cacheWrite5mCost) : undefined,
       "cost.total.microcents": centsToMicroCents(totalCostInCent),
       // deprecated - remove after May 20, 2026
       "cost.input": Math.round(inputCost),
       "cost.output": Math.round(outputCost),
-      "cost.reasoning": reasoningCost ? Math.round(reasoningCost) : undefined,
       "cost.cache_read": cacheReadCost ? Math.round(cacheReadCost) : undefined,
       "cost.cache_write_5m": cacheWrite5mCost ? Math.round(cacheWrite5mCost) : undefined,
       "cost.cache_write_1h": cacheWrite1hCost ? Math.round(cacheWrite1hCost) : undefined,

packages/console/app/src/routes/zen/util/provider/openai.ts

@@ -50,7 +50,7 @@ export const openaiHelper: ProviderHelper = ({ workspaceID }) => ({
     const cacheReadTokens = usage.input_tokens_details?.cached_tokens ?? undefined
     return {
       inputTokens: inputTokens - (cacheReadTokens ?? 0),
-      outputTokens: outputTokens - (reasoningTokens ?? 0),
+      outputTokens,
       reasoningTokens,
       cacheReadTokens,
       cacheWrite5mTokens: undefined,

packages/console/core/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "@opencode-ai/console-core",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "private": true,
   "type": "module",
   "license": "MIT",

packages/console/function/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/console-function",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "$schema": "https://json.schemastore.org/package.json",
   "private": true,
   "type": "module",

packages/console/mail/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/console-mail",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "dependencies": {
     "@jsx-email/all": "2.2.3",
     "@jsx-email/cli": "1.4.3",

packages/core/package.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "name": "@opencode-ai/core",
   "type": "module",
   "license": "MIT",

packages/core/src/effect-zod.ts

@@ -36,7 +36,7 @@ export function zod<S extends Schema.Top>(schema: S): z.ZodType<Schema.Schema.Ty
  * mapped `.omit()` / `.extend()` surface triggers brand-intersection
  * explosions for branded primitives (`string & Brand<"SessionID">` extends
  * `object` via the brand and gets walked into the prototype by `DeepPartial`,
- * `updateSchema`, etc.), and zod's inference through `z.ZodType<T | undefined>`
+ * mapped-schema helpers, and zod's inference through `z.ZodType<T | undefined>`
  * wrappers also can't reconstruct `T` cleanly. Consumers that care about the
  * post-`.omit()` shape should cast `c.req.valid(...)` to the expected type.
  */

packages/desktop/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@opencode-ai/desktop",
   "private": true,
-  "version": "1.14.46",
+  "version": "1.14.48",
   "type": "module",
   "license": "MIT",
   "homepage": "https://opencode.ai",

packages/enterprise/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/enterprise",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "private": true,
   "type": "module",
   "license": "MIT",

packages/extensions/zed/extension.toml

@@ -1,7 +1,7 @@
 id = "opencode"
 name = "OpenCode"
 description = "The open source coding agent."
-version = "1.14.46"
+version = "1.14.48"
 schema_version = 1
 authors = ["Anomaly"]
 repository = "https://github.com/anomalyco/opencode"
@@ -11,26 +11,26 @@ name = "OpenCode"
 icon = "./icons/opencode.svg"
 
 [agent_servers.opencode.targets.darwin-aarch64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.46/opencode-darwin-arm64.zip"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.48/opencode-darwin-arm64.zip"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.darwin-x86_64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.46/opencode-darwin-x64.zip"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.48/opencode-darwin-x64.zip"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.linux-aarch64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.46/opencode-linux-arm64.tar.gz"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.48/opencode-linux-arm64.tar.gz"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.linux-x86_64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.46/opencode-linux-x64.tar.gz"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.48/opencode-linux-x64.tar.gz"
 cmd = "./opencode"
 args = ["acp"]
 
 [agent_servers.opencode.targets.windows-x86_64]
-archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.46/opencode-windows-x64.zip"
+archive = "https://github.com/anomalyco/opencode/releases/download/v1.14.48/opencode-windows-x64.zip"
 cmd = "./opencode.exe"
 args = ["acp"]

packages/function/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opencode-ai/function",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "$schema": "https://json.schemastore.org/package.json",
   "private": true,
   "type": "module",

packages/http-recorder/README.md

@@ -70,19 +70,15 @@ Cassettes are normal source files — review them, diff them, commit them.
 
 ## Request matching
 
-By default, requests match on canonicalized method, URL, headers, and JSON
-body (object keys sorted). Two dispatch strategies are available:
+Replay walks the cassette in record order via an internal cursor: the Nth
+request executed at runtime is served by the Nth recorded interaction, and
+each one is validated as the cursor advances. Request equality is computed
+on canonicalized method, URL, headers, and JSON body (object keys sorted).
 
-- **`match`** (default) — find the first recorded interaction whose request
-  matches the incoming request. Same request twice returns the same response.
-- **`sequential`** — return interactions in the order they were recorded,
-  validating each one matches as the cursor advances. Use for ordered flows
-  where the same URL is hit multiple times with meaningful state changes
-  (pagination, retries, polling).
-
-```ts
-HttpRecorder.cassetteLayer("flow/poll-until-done", { dispatch: "sequential" })
-```
+This is deliberately strict — content-based dispatch was removed because
+it silently returns the first recorded response for repeated identical
+requests, masking state changes that retry/polling/cache-hit tests need to
+observe. If you reorder requests in a test, re-record the cassette.
 
 Supply your own matcher via `match: (incoming, recorded) => boolean` for
 custom equivalence (e.g. ignoring a timestamp field in the body).
@@ -172,9 +168,9 @@ const audit = Effect.gen(function* () {
   const cassettes = yield* HttpRecorder.Cassette.Service
   const entries = yield* cassettes.list()
   const issues = yield* Effect.forEach(entries, (entry) =>
-    cassettes.read(entry.name).pipe(
-      Effect.map((interactions) => ({ name: entry.name, findings: HttpRecorder.secretFindings(interactions) })),
-    ),
+    cassettes
+      .read(entry.name)
+      .pipe(Effect.map((interactions) => ({ name: entry.name, findings: HttpRecorder.secretFindings(interactions) }))),
   )
   return issues.filter((i) => i.findings.length > 0)
 })
@@ -194,7 +190,6 @@ type RecordReplayOptions = {
   directory?: string // default: <cwd>/test/fixtures/recordings
   metadata?: Record<string, unknown> // merged into cassette.metadata
   redactor?: Redactor // default: Redactor.defaults()
-  dispatch?: "match" | "sequential" // default: "match"
   match?: (incoming, recorded) => boolean // custom matcher
 }
 ```
@@ -211,4 +206,4 @@ type RecordReplayOptions = {
 | `redaction.ts` | Lower-level header/URL primitives + secret pattern detection.                    |
 | `schema.ts`    | Effect Schema definitions for the cassette JSON format.                          |
 | `storage.ts`   | Path resolution, JSON encode/decode, sync existence check.                       |
-| `matching.ts`  | Request matcher, canonicalization, dispatch strategies, mismatch diagnostics.    |
+| `matching.ts`  | Request matcher, canonicalization, sequential cursor, mismatch diagnostics.      |

packages/http-recorder/package.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "name": "@opencode-ai/http-recorder",
   "type": "module",
   "license": "MIT",

packages/http-recorder/src/cassette.ts

@@ -6,10 +6,9 @@ import { decodeCassette, encodeCassette, type Cassette, type CassetteMetadata, t
 
 const DEFAULT_RECORDINGS_DIR = path.resolve(process.cwd(), "test", "fixtures", "recordings")
 
-export class CassetteNotFoundError extends Schema.TaggedErrorClass<CassetteNotFoundError>()(
-  "CassetteNotFoundError",
-  { cassetteName: Schema.String },
-) {
+export class CassetteNotFoundError extends Schema.TaggedErrorClass<CassetteNotFoundError>()("CassetteNotFoundError", {
+  cassetteName: Schema.String,
+}) {
   override get message() {
     return `Cassette "${this.cassetteName}" not found`
   }
@@ -21,11 +20,7 @@ export interface AppendResult {
 
 export interface Interface {
   readonly read: (name: string) => Effect.Effect<ReadonlyArray<Interaction>, CassetteNotFoundError>
-  readonly append: (
-    name: string,
-    interaction: Interaction,
-    metadata?: CassetteMetadata,
-  ) => Effect.Effect<AppendResult>
+  readonly append: (name: string, interaction: Interaction, metadata?: CassetteMetadata) => Effect.Effect<AppendResult>
   readonly exists: (name: string) => Effect.Effect<boolean>
   readonly list: () => Effect.Effect<ReadonlyArray<string>>
 }
@@ -113,7 +108,12 @@ export const fileSystem = (
             Effect.map((files) =>
               files
                 .filter((file) => file.endsWith(".json"))
-                .map((file) => path.relative(directory, file).replace(/\\/g, "/").replace(/\.json$/, ""))
+                .map((file) =>
+                  path
+                    .relative(directory, file)
+                    .replace(/\\/g, "/")
+                    .replace(/\.json$/, ""),
+                )
                 .toSorted((a, b) => a.localeCompare(b)),
             ),
           ),

packages/http-recorder/src/effect.ts

@@ -11,7 +11,7 @@ import {
   UrlParams,
 } from "effect/unstable/http"
 import * as CassetteService from "./cassette"
-import { defaultMatcher, selectMatch, selectSequential, type RequestMatcher } from "./matching"
+import { defaultMatcher, selectSequential, type RequestMatcher } from "./matching"
 import { appendOrFail, makeReplayState, resolveAutoMode } from "./recorder"
 import { defaults, type Redactor } from "./redactor"
 import { redactUrl } from "./redaction"
@@ -24,7 +24,6 @@ export interface RecordReplayOptions {
   readonly directory?: string
   readonly metadata?: CassetteMetadata
   readonly redactor?: Redactor
-  readonly dispatch?: "match" | "sequential"
   readonly match?: RequestMatcher
 }
 
@@ -71,7 +70,6 @@ export const recordingLayer = (
       const match = options.match ?? defaultMatcher
       const requested = options.mode ?? "auto"
       const mode = requested === "auto" ? yield* resolveAutoMode(cassetteService, name) : requested
-      const sequential = options.dispatch === "sequential"
       const replay = yield* makeReplayState(cassetteService, name, httpInteractions)
 
       const snapshotRequest = (request: HttpClientRequest.HttpClientRequest) =>
@@ -119,14 +117,12 @@ export const recordingLayer = (
               transportError(request, `Fixture "${name}" not found. Run locally to record it (CI=true forces replay).`),
             ),
           )
-          const result = sequential
-            ? selectSequential(interactions, incoming, match, yield* replay.cursor)
-            : selectMatch(interactions, incoming, match)
+          const result = selectSequential(interactions, incoming, match, yield* replay.cursor)
           if (!result.interaction)
             return yield* Effect.fail(
               transportError(request, `Fixture "${name}" does not match the current request: ${result.detail}.`),
             )
-          if (sequential) yield* replay.advance
+          yield* replay.advance
           return HttpClientResponse.fromWeb(
             request,
             new Response(decodeResponseBody(result.interaction.response), result.interaction.response),

packages/http-recorder/src/matching.ts

@@ -92,24 +92,6 @@ export const requestDiff = (expected: RequestSnapshot, received: RequestSnapshot
   return lines
 }
 
-export const mismatchDetail = (interactions: ReadonlyArray<HttpInteraction>, incoming: RequestSnapshot): string => {
-  if (interactions.length === 0) return "cassette has no recorded HTTP interactions"
-  const ranked = interactions
-    .map((interaction, index) => ({ index, lines: requestDiff(interaction.request, incoming) }))
-    .toSorted((a, b) => a.lines.length - b.lines.length || a.index - b.index)
-  const best = ranked[0]
-  return ["no recorded interaction matched", `closest interaction: #${best.index + 1}`, ...best.lines].join("\n")
-}
-
-export const selectMatch = (
-  interactions: ReadonlyArray<HttpInteraction>,
-  incoming: RequestSnapshot,
-  match: RequestMatcher,
-): { readonly interaction: HttpInteraction | undefined; readonly detail: string } => {
-  const interaction = interactions.find((candidate) => match(incoming, candidate.request))
-  return { interaction, detail: interaction ? "" : mismatchDetail(interactions, incoming) }
-}
-
 export const selectSequential = (
   interactions: ReadonlyArray<HttpInteraction>,
   incoming: RequestSnapshot,

packages/http-recorder/src/recorder.ts

@@ -34,13 +34,13 @@ export const appendOrFail = (
   interaction: Interaction,
   metadata: CassetteMetadata | undefined,
 ): Effect.Effect<void, UnsafeCassetteError> =>
-  cassette.append(name, interaction, metadata).pipe(
-    Effect.flatMap(({ findings }) =>
-      findings.length === 0
-        ? Effect.void
-        : Effect.fail(new UnsafeCassetteError({ cassetteName: name, findings })),
-    ),
-  )
+  cassette
+    .append(name, interaction, metadata)
+    .pipe(
+      Effect.flatMap(({ findings }) =>
+        findings.length === 0 ? Effect.void : Effect.fail(new UnsafeCassetteError({ cassetteName: name, findings })),
+      ),
+    )
 
 export interface ReplayState<T> {
   readonly load: Effect.Effect<ReadonlyArray<T>, CassetteNotFoundError>

packages/http-recorder/src/redaction.ts

@@ -113,4 +113,3 @@ export const secretFindings = (value: unknown): ReadonlyArray<SecretFinding> =>
       .filter((item) => entry.value.includes(item.value))
       .map((item) => ({ path: entry.path, reason: `environment secret ${item.name}` })),
   ])
-

packages/http-recorder/test/record-replay.test.ts

@@ -14,10 +14,7 @@ const seedCassetteDirectory = (directory: string, name: string, interactions: Re
     Effect.gen(function* () {
       const cassette = yield* HttpRecorder.Cassette.Service
       yield* Effect.forEach(interactions, (interaction) => cassette.append(name, interaction))
-    }).pipe(
-      Effect.provide(HttpRecorder.Cassette.fileSystem({ directory })),
-      Effect.provide(NodeFileSystem.layer),
-    ),
+    }).pipe(Effect.provide(HttpRecorder.Cassette.fileSystem({ directory })), Effect.provide(NodeFileSystem.layer)),
   )
 
 const post = (url: string, body: object) =>
@@ -233,19 +230,10 @@ describe("http-recorder", () => {
     )
   })
 
-  test("default matcher dispatches multi-interaction cassettes by request shape", async () => {
-    await run(
-      Effect.gen(function* () {
-        expect(yield* post("https://example.test/echo", { step: 2 })).toBe('{"reply":"second"}')
-        expect(yield* post("https://example.test/echo", { step: 1 })).toBe('{"reply":"first"}')
-      }),
-    )
-  })
-
-  test("sequential dispatch returns recorded responses in order for identical requests", async () => {
+  test("replay returns recorded responses in order for identical requests", async () => {
     await runWith(
       "record-replay/retry",
-      { dispatch: "sequential" },
+      {},
       Effect.gen(function* () {
         expect(yield* post("https://example.test/poll", { id: "job_1" })).toBe('{"status":"pending"}')
         expect(yield* post("https://example.test/poll", { id: "job_1" })).toBe('{"status":"complete"}')
@@ -253,21 +241,8 @@ describe("http-recorder", () => {
     )
   })
 
-  test("default matcher returns the first match for identical requests", async () => {
-    await runWith(
-      "record-replay/retry",
-      {},
-      Effect.gen(function* () {
-        expect(yield* post("https://example.test/poll", { id: "job_1" })).toBe('{"status":"pending"}')
-        expect(yield* post("https://example.test/poll", { id: "job_1" })).toBe('{"status":"pending"}')
-      }),
-    )
-  })
-
-  test("sequential dispatch reports cursor exhaustion when more requests are made than recorded", async () => {
-    await runWith(
-      "record-replay/multi-step",
-      { dispatch: "sequential" },
+  test("replay reports cursor exhaustion when more requests are made than recorded", async () => {
+    await run(
       Effect.gen(function* () {
         yield* post("https://example.test/echo", { step: 1 })
         yield* post("https://example.test/echo", { step: 2 })
@@ -277,10 +252,8 @@ describe("http-recorder", () => {
     )
   })
 
-  test("sequential dispatch still validates each recorded request", async () => {
-    await runWith(
-      "record-replay/multi-step",
-      { dispatch: "sequential" },
+  test("replay validates each recorded request in order", async () => {
+    await run(
       Effect.gen(function* () {
         yield* post("https://example.test/echo", { step: 1 })
         const exit = yield* Effect.exit(post("https://example.test/echo", { step: 3 }))
@@ -334,14 +307,13 @@ describe("http-recorder", () => {
     }
   })
 
-  test("mismatch diagnostics show closest redacted request differences", async () => {
+  test("mismatch diagnostics show redacted request differences against the expected interaction", async () => {
     await run(
       Effect.gen(function* () {
         const exit = yield* Effect.exit(
           post("https://example.test/echo?api_key=secret-value", { step: 3, token: "sk-123456789012345678901234" }),
         )
         const message = failureText(exit)
-        expect(message).toContain("closest interaction: #1")
         expect(message).toContain("url:")
         expect(message).toContain("https://example.test/echo?api_key=%5BREDACTED%5D")
         expect(message).toContain("body:")

packages/llm/AGENTS.md

@@ -289,6 +289,6 @@ Filters apply in replay and record mode. Combine them with `RECORD=true` when re
 
 **Binary response bodies.** Most providers stream text (SSE, JSON). AWS Bedrock streams binary AWS event-stream frames whose CRC32 fields would be mangled by a UTF-8 round-trip — those bodies are stored as base64 with `bodyEncoding: "base64"` on the response snapshot. Detection is by `Content-Type` in `@opencode-ai/http-recorder` (currently `application/vnd.amazon.eventstream` and `application/octet-stream`); cassettes for SSE/JSON routes omit the field and decode as text.
 
-**Matching strategies.** Replay defaults to structural matching, which finds an interaction by comparing method, URL, allow-listed headers, and the canonical JSON body. This is the right choice for tool loops because each round's request differs (the message history grows). For scenarios where successive requests are byte-identical and expect different responses (retries, polling), pass `dispatch: "sequential"` in `RecordReplayOptions` — replay then walks the cassette in record order via an internal cursor. `scriptedResponses` (in `test/lib/http.ts`) is the deterministic counterpart for tests that don't need a live provider; it scripts response bodies in order without reading from disk.
+**Matching strategy.** Replay walks the cassette in record order via an internal cursor: the Nth runtime request is served by the Nth recorded interaction, and each one is validated by comparing method, URL, allow-listed headers, and the canonical JSON body. This handles tool loops (each round's request differs as history grows) and retry/polling scenarios (successive byte-identical requests with different responses) uniformly. If a test reorders its requests, re-record the cassette. `scriptedResponses` (in `test/lib/http.ts`) is the deterministic counterpart for tests that don't need a live provider; it scripts response bodies in order without reading from disk.
 
 Do not blanket re-record an entire test file when adding one cassette. `RECORD=true` rewrites every recorded case that runs, and provider streams contain volatile IDs, timestamps, fingerprints, and obfuscation fields. Prefer deleting the one cassette you intend to refresh, or run a focused test pattern that only registers the scenario you want to record. Keep stable existing cassettes unchanged unless their request shape or expected behavior changed.

packages/llm/README.md

@@ -0,0 +1,129 @@
+# @opencode-ai/llm
+
+Schema-first LLM core for opencode. One typed request, response, event, and tool language; provider quirks live in adapters, not in calling code.
+
+```ts
+import { Effect } from "effect"
+import { LLM, LLMClient } from "@opencode-ai/llm"
+import { OpenAI } from "@opencode-ai/llm/providers"
+
+const model = OpenAI.model("gpt-4o-mini", { apiKey: process.env.OPENAI_API_KEY })
+
+const request = LLM.request({
+  model,
+  system: "You are concise.",
+  prompt: "Say hello in one short sentence.",
+  generation: { maxTokens: 40 },
+})
+
+const program = Effect.gen(function* () {
+  const response = yield* LLMClient.generate(request)
+  console.log(response.text)
+})
+```
+
+Run `LLMClient.stream(request)` instead of `generate` when you want incremental `LLMEvent`s. The event stream is provider-neutral — same shape across OpenAI Chat, OpenAI Responses, Anthropic Messages, Gemini, Bedrock Converse, and any OpenAI-compatible deployment.
+
+## Public API
+
+- **`LLM.request({...})`** — build a provider-neutral `LLMRequest`. Accepts ergonomic inputs (`system: string`, `prompt: string`) that normalize into the canonical Schema classes.
+- **`LLM.generate` / `LLM.stream`** — re-exported from `LLMClient` for one-import use.
+- **`LLM.user(...)` / `LLM.assistant(...)` / `LLM.toolMessage(...)`** — message constructors.
+- **`LLM.toolCall(...)` / `LLM.toolResult(...)` / `LLM.toolDefinition(...)`** — tool-related parts.
+- **`LLMClient.prepare(request)`** — compile a request through protocol body construction, validation, and HTTP preparation without sending. Useful for inspection and testing.
+- **`LLMEvent.is.*`** — typed guards (`is.text`, `is.toolCall`, `is.requestFinish`, …) for filtering streams.
+
+## Caching
+
+Prompt caching is **on by default**. Every `LLMRequest` resolves to `cache: "auto"` unless the caller opts out with `cache: "none"`. Each protocol translates `CacheHint`s to its wire format (`cache_control` on Anthropic, `cachePoint` on Bedrock; OpenAI and Gemini do implicit caching server-side and don't need inline markers — auto is a no-op there).
+
+### Auto placement
+
+`"auto"` places three breakpoints — last tool definition, last system part, latest user message. The last-user-message boundary is the load-bearing detail: in a tool-use loop, a single user turn expands into many assistant/tool round-trips, all sharing that prefix. Caching at that boundary lets every intra-turn API call hit.
+
+The math justifies the default: Anthropic's 5-minute cache write is 1.25× base, read is 0.1×, so a single reuse within 5 minutes already wins. One-shot completions below the per-model minimum-cacheable-token threshold silently no-op on the wire, so the worst case is harmless.
+
+### Opting out
+
+```ts
+LLM.request({
+  model,
+  system,
+  prompt: "one-off question",
+  cache: "none",
+})
+```
+
+### Granular policy
+
+```ts
+cache: {
+  tools?: boolean,
+  system?: boolean,
+  messages?: "latest-user-message" | "latest-assistant" | { tail: number },
+  ttlSeconds?: number,         // ≥ 3600 → 1h on Anthropic/Bedrock; else 5m
+}
+```
+
+### Manual hints
+
+Inline `CacheHint` on any text / system / tool / tool-result part overrides automatic placement. The auto policy preserves manual hints; it only fills gaps.
+
+```ts
+LLM.request({
+  model,
+  system: [
+    { type: "text", text: "stable system prompt", cache: { type: "ephemeral" } },
+  ],
+  ...
+})
+```
+
+### Provider behavior table
+
+| Protocol                | `cache: "auto"`                                                           |
+| ----------------------- | ------------------------------------------------------------------------- |
+| Anthropic Messages      | emits up to 3 `cache_control` markers (4-breakpoint cap enforced)         |
+| Bedrock Converse        | emits up to 3 `cachePoint` blocks (4-breakpoint cap enforced)             |
+| OpenAI Chat / Responses | no-op (implicit caching above 1024 tokens)                                |
+| Gemini                  | no-op (implicit caching on 2.5+; explicit `CachedContent` is out-of-band) |
+
+Normalized cache usage is read back into `response.usage.cacheReadInputTokens` and `cacheWriteInputTokens` across every provider.
+
+## Providers
+
+Each provider exports a `model(...)` helper that records identity, protocol, capabilities, auth, and defaults.
+
+```ts
+import { Anthropic } from "@opencode-ai/llm/providers"
+
+const model = Anthropic.model("claude-sonnet-4-6", {
+  apiKey: process.env.ANTHROPIC_API_KEY,
+})
+```
+
+Included providers: OpenAI, Anthropic, Google (Gemini), Amazon Bedrock, Azure OpenAI, Cloudflare, GitHub Copilot, OpenRouter, xAI, plus generic OpenAI-compatible helpers for DeepSeek, Cerebras, Groq, Fireworks, Together, etc.
+
+## Provider options & HTTP overlays
+
+Three escape hatches in order of stability:
+
+1. **`generation`** — portable knobs (`maxTokens`, `temperature`, `topP`, `topK`, penalties, seed, stop).
+2. **`providerOptions: { <provider>: {...} }`** — typed-at-the-facade provider-specific knobs (OpenAI `promptCacheKey`, Anthropic `thinking`, Gemini `thinkingConfig`, OpenRouter routing).
+3. **`http: { body, headers, query }`** — last-resort serializable overlays merged into the final HTTP request. Reach for this only when a stable typed path doesn't yet exist.
+
+Model-level defaults are overridden by request-level values for each axis.
+
+## Routes
+
+Adding a new model or deployment is usually 5–15 lines using `Route.make({ protocol, transport, ... })`. The four orthogonal pieces are protocol (body construction + stream parsing), transport (endpoint + auth + framing + encoding), defaults, and capabilities. See `AGENTS.md` for the architectural detail.
+
+## Effect
+
+This package is built on Effect. Public methods return `Effect` or `Stream`; provide `LLMClient.layer` (the default registers every shipped route) for runtime dispatch. The example at `example/tutorial.ts` is a runnable walkthrough.
+
+## See also
+
+- `AGENTS.md` — architecture, route construction, contributor guide
+- `example/tutorial.ts` — runnable end-to-end walkthrough
+- `test/provider/*.test.ts` — fixture-first protocol tests; `*.recorded.test.ts` files cover live cassettes

packages/llm/example/tutorial.ts

@@ -184,7 +184,7 @@ const FakeProtocol = Protocol.make<FakeBody, string, string, void>({
   stream: {
     event: Schema.String,
     initial: () => undefined,
-    step: (_, frame) => Effect.succeed([undefined, [{ type: "text-delta", text: frame }]] as const),
+    step: (_, frame) => Effect.succeed([undefined, [{ type: "text-delta", id: "text-0", text: frame }]] as const),
     onHalt: () => [{ type: "request-finish", reason: "stop" }],
   },
 })

packages/llm/package.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "name": "@opencode-ai/llm",
   "type": "module",
   "license": "MIT",

packages/llm/src/cache-policy.ts

@@ -0,0 +1,111 @@
+// Apply an `LLMRequest.cache` policy by injecting `CacheHint`s onto the parts
+// the policy designates. Runs once at compile time, before the per-protocol
+// body builder, so the existing inline-hint lowering path handles the rest.
+//
+// The default `"auto"` shape places one breakpoint at the last tool definition,
+// one at the last system part, and one at the latest user message. This
+// matches what production agent harnesses (LangChain's caching middleware,
+// kern-ai's 10x cost-reduction playbook) converge on for tool-use loops: the
+// latest user message stays put while a single turn explodes into many
+// assistant/tool round-trips, so caching at that boundary lets every
+// intra-turn API call hit the prefix.
+//
+// Manual `cache: CacheHint` placements on individual parts are preserved —
+// this function only fills gaps the caller left empty.
+import { CacheHint, type CachePolicy, type CachePolicyObject } from "./schema/options"
+import { LLMRequest, Message, ToolDefinition, type ContentPart } from "./schema/messages"
+
+const AUTO: CachePolicyObject = {
+  tools: true,
+  system: true,
+  messages: "latest-user-message",
+}
+
+const NONE: CachePolicyObject = {}
+
+// Resolution rules:
+//   - undefined   → "auto" — caching is on by default. The math favors it:
+//                   Anthropic 5m-cache write is 1.25x base, read is 0.1x,
+//                   so a single reuse within 5 minutes already wins.
+//   - "auto"      → tools + system + latest user msg.
+//   - "none"      → no auto placement; manual `CacheHint`s still flow.
+//   - object form → exactly what the caller asked for.
+const resolve = (policy: CachePolicy | undefined): CachePolicyObject => {
+  if (policy === undefined || policy === "auto") return AUTO
+  if (policy === "none") return NONE
+  return policy
+}
+
+// Protocols whose wire format ignores inline cache markers (OpenAI's implicit
+// prefix caching, Gemini's implicit + out-of-band CachedContent). Skip the
+// whole policy pass for these — emitting hints would be harmless but pointless.
+const RESPECTS_INLINE_HINTS = new Set(["anthropic-messages", "bedrock-converse"])
+
+const makeHint = (ttlSeconds: number | undefined): CacheHint =>
+  ttlSeconds !== undefined ? new CacheHint({ type: "ephemeral", ttlSeconds }) : new CacheHint({ type: "ephemeral" })
+
+const markLastTool = (tools: ReadonlyArray<ToolDefinition>, hint: CacheHint): ReadonlyArray<ToolDefinition> => {
+  if (tools.length === 0) return tools
+  const last = tools.length - 1
+  if (tools[last]!.cache) return tools
+  return tools.map((tool, i) => (i === last ? new ToolDefinition({ ...tool, cache: hint }) : tool))
+}
+
+const markLastSystem = (system: LLMRequest["system"], hint: CacheHint): LLMRequest["system"] => {
+  if (system.length === 0) return system
+  const last = system.length - 1
+  if (system[last]!.cache) return system
+  return system.map((part, i) => (i === last ? { ...part, cache: hint } : part))
+}
+
+const lastIndexOfRole = (messages: ReadonlyArray<Message>, role: Message["role"]): number =>
+  messages.findLastIndex((m) => m.role === role)
+
+// Mark the last text part of `messages[index]`. If no text part exists, mark
+// the last content part regardless of type — that's the breakpoint position
+// in tool-result-only messages too.
+const markMessageAt = (messages: ReadonlyArray<Message>, index: number, hint: CacheHint): ReadonlyArray<Message> => {
+  if (index < 0 || index >= messages.length) return messages
+  const target = messages[index]!
+  if (target.content.length === 0) return messages
+  const lastTextIndex = target.content.findLastIndex((part) => part.type === "text")
+  const markAt = lastTextIndex >= 0 ? lastTextIndex : target.content.length - 1
+  const existing = target.content[markAt]!
+  if ("cache" in existing && existing.cache) return messages
+  const nextContent = target.content.map((part, i) => (i === markAt ? ({ ...part, cache: hint } as ContentPart) : part))
+  const next = new Message({ ...target, content: nextContent })
+  // Single pass over `messages`, substituting the one updated entry. Long
+  // conversations call this on every request, so avoid `.map()` here — its
+  // closure dispatch and identity copies show up in profiling.
+  const result = messages.slice()
+  result[index] = next
+  return result
+}
+
+const markMessages = (
+  messages: ReadonlyArray<Message>,
+  strategy: NonNullable<CachePolicyObject["messages"]>,
+  hint: CacheHint,
+): ReadonlyArray<Message> => {
+  if (messages.length === 0) return messages
+  if (strategy === "latest-user-message") return markMessageAt(messages, lastIndexOfRole(messages, "user"), hint)
+  if (strategy === "latest-assistant") return markMessageAt(messages, lastIndexOfRole(messages, "assistant"), hint)
+  const start = Math.max(0, messages.length - strategy.tail)
+  let next = messages
+  for (let i = start; i < messages.length; i++) next = markMessageAt(next, i, hint)
+  return next
+}
+
+export const applyCachePolicy = (request: LLMRequest): LLMRequest => {
+  if (!RESPECTS_INLINE_HINTS.has(request.model.route)) return request
+  const policy = resolve(request.cache)
+  if (!policy.tools && !policy.system && !policy.messages) return request
+
+  const hint = makeHint(policy.ttlSeconds)
+  const tools = policy.tools ? markLastTool(request.tools, hint) : request.tools
+  const system = policy.system ? markLastSystem(request.system, hint) : request.system
+  const messages = policy.messages ? markMessages(request.messages, policy.messages, hint) : request.messages
+
+  if (tools === request.tools && system === request.system && messages === request.messages) return request
+  return LLMRequest.update(request, { tools, system, messages })
+}

packages/llm/src/protocols/anthropic-messages.ts

@@ -5,10 +5,10 @@ import { Endpoint } from "../route/endpoint"
 import { Framing } from "../route/framing"
 import { Protocol } from "../route/protocol"
 import {
+  LLMEvent,
   Usage,
   type CacheHint,
   type FinishReason,
-  type LLMEvent,
   type LLMRequest,
   type ProviderMetadata,
   type ToolCallPart,
@@ -16,6 +16,7 @@ import {
   type ToolResultPart,
 } from "../schema"
 import { JsonObject, optionalArray, optionalNull, ProviderShared } from "./shared"
+import * as Cache from "./utils/cache"
 import { ToolStream } from "./utils/tool-stream"
 
 const ADAPTER = "anthropic-messages"
@@ -25,7 +26,10 @@ export const PATH = "/messages"
 // =============================================================================
 // Request Body Schema
 // =============================================================================
-const AnthropicCacheControl = Schema.Struct({ type: Schema.tag("ephemeral") })
+const AnthropicCacheControl = Schema.Struct({
+  type: Schema.tag("ephemeral"),
+  ttl: Schema.optional(Schema.Literals(["5m", "1h"])),
+})
 
 const AnthropicTextBlock = Schema.Struct({
   type: Schema.tag("text"),
@@ -193,8 +197,24 @@ const invalid = ProviderShared.invalidRequest
 // =============================================================================
 // Request Lowering
 // =============================================================================
-const cacheControl = (cache: CacheHint | undefined) =>
-  cache?.type === "ephemeral" ? { type: "ephemeral" as const } : undefined
+// Anthropic accepts at most 4 explicit cache_control breakpoints per request,
+// across `tools`, `system`, and `messages`. Beyond the cap the API returns a
+// 400 — so the lowering layer counts emitted markers and silently drops any
+// that exceed it.
+const ANTHROPIC_BREAKPOINT_CAP = 4
+
+const EPHEMERAL_5M = { type: "ephemeral" as const }
+const EPHEMERAL_1H = { type: "ephemeral" as const, ttl: "1h" as const }
+
+const cacheControl = (breakpoints: Cache.Breakpoints, cache: CacheHint | undefined) => {
+  if (cache?.type !== "ephemeral" && cache?.type !== "persistent") return undefined
+  if (breakpoints.remaining <= 0) {
+    breakpoints.dropped += 1
+    return undefined
+  }
+  breakpoints.remaining -= 1
+  return Cache.ttlBucket(cache.ttlSeconds) === "1h" ? EPHEMERAL_1H : EPHEMERAL_5M
+}
 
 const anthropicMetadata = (metadata: Record<string, unknown>): ProviderMetadata => ({ anthropic: metadata })
 
@@ -204,10 +224,11 @@ const signatureFromMetadata = (metadata: ProviderMetadata | undefined): string |
   return typeof anthropic.signature === "string" ? anthropic.signature : undefined
 }
 
-const lowerTool = (tool: ToolDefinition): AnthropicTool => ({
+const lowerTool = (breakpoints: Cache.Breakpoints, tool: ToolDefinition): AnthropicTool => ({
   name: tool.name,
   description: tool.description,
   input_schema: tool.inputSchema,
+  cache_control: cacheControl(breakpoints, tool.cache),
 })
 
 const lowerToolChoice = (toolChoice: NonNullable<LLMRequest["toolChoice"]>) =>
@@ -249,7 +270,10 @@ const lowerServerToolResult = Effect.fn("AnthropicMessages.lowerServerToolResult
   return { type: wireType, tool_use_id: part.id, content: part.result.value } satisfies AnthropicServerToolResultBlock
 })
 
-const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (request: LLMRequest) {
+const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (
+  request: LLMRequest,
+  breakpoints: Cache.Breakpoints,
+) {
   const messages: AnthropicMessage[] = []
 
   for (const message of request.messages) {
@@ -258,7 +282,7 @@ const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (re
       for (const part of message.content) {
         if (!ProviderShared.supportsContent(part, ["text"]))
           return yield* ProviderShared.unsupportedContent("Anthropic Messages", "user", ["text"])
-        content.push({ type: "text", text: part.text, cache_control: cacheControl(part.cache) })
+        content.push({ type: "text", text: part.text, cache_control: cacheControl(breakpoints, part.cache) })
       }
       messages.push({ role: "user", content })
       continue
@@ -268,7 +292,7 @@ const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (re
       const content: AnthropicAssistantBlock[] = []
       for (const part of message.content) {
         if (part.type === "text") {
-          content.push({ type: "text", text: part.text, cache_control: cacheControl(part.cache) })
+          content.push({ type: "text", text: part.text, cache_control: cacheControl(breakpoints, part.cache) })
           continue
         }
         if (part.type === "reasoning") {
@@ -304,6 +328,7 @@ const lowerMessages = Effect.fn("AnthropicMessages.lowerMessages")(function* (re
         tool_use_id: part.id,
         content: ProviderShared.toolResultText(part),
         is_error: part.result.type === "error" ? true : undefined,
+        cache_control: cacheControl(breakpoints, part.cache),
       })
     }
     messages.push({ role: "user", content })
@@ -330,18 +355,33 @@ const lowerThinking = Effect.fn("AnthropicMessages.lowerThinking")(function* (re
 const fromRequest = Effect.fn("AnthropicMessages.fromRequest")(function* (request: LLMRequest) {
   const toolChoice = request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined
   const generation = request.generation
+  // Allocate the 4-breakpoint budget in invalidation order: tools → system →
+  // messages. Tools live highest in the cache hierarchy, so when callers
+  // over-mark we keep their tool hints and shed the message-tail ones first.
+  const breakpoints = Cache.newBreakpoints(ANTHROPIC_BREAKPOINT_CAP)
+  const tools =
+    request.tools.length === 0 || request.toolChoice?.type === "none"
+      ? undefined
+      : request.tools.map((tool) => lowerTool(breakpoints, tool))
+  const system =
+    request.system.length === 0
+      ? undefined
+      : request.system.map((part) => ({
+          type: "text" as const,
+          text: part.text,
+          cache_control: cacheControl(breakpoints, part.cache),
+        }))
+  const messages = yield* lowerMessages(request, breakpoints)
+  if (breakpoints.dropped > 0) {
+    yield* Effect.logWarning(
+      `Anthropic Messages: dropped ${breakpoints.dropped} cache breakpoint(s); the API allows at most ${ANTHROPIC_BREAKPOINT_CAP} per request.`,
+    )
+  }
   return {
     model: request.model.id,
-    system:
-      request.system.length === 0
-        ? undefined
-        : request.system.map((part) => ({
-            type: "text" as const,
-            text: part.text,
-            cache_control: cacheControl(part.cache),
-          })),
-    messages: yield* lowerMessages(request),
-    tools: request.tools.length === 0 || request.toolChoice?.type === "none" ? undefined : request.tools.map(lowerTool),
+    system,
+    messages,
+    tools,
     tool_choice: toolChoice,
     stream: true as const,
     max_tokens: generation?.maxTokens ?? request.model.limits.output ?? 4096,
@@ -364,34 +404,56 @@ const mapFinishReason = (reason: string | null | undefined): FinishReason => {
   return "unknown"
 }
 
+// Anthropic reports the non-overlapping breakdown natively — its
+// `input_tokens` is the *non-cached* count per the Messages API docs, with
+// cache reads and writes as separate fields. We sum them to derive the
+// inclusive `inputTokens` the rest of the contract expects. Extended
+// thinking tokens are *not* broken out by Anthropic — they're billed as
+// part of `output_tokens`, so `reasoningTokens` stays `undefined` and
+// `outputTokens` carries the combined total.
 const mapUsage = (usage: AnthropicUsage | undefined): Usage | undefined => {
   if (!usage) return undefined
+  const nonCached = usage.input_tokens
+  const cacheRead = usage.cache_read_input_tokens ?? undefined
+  const cacheWrite = usage.cache_creation_input_tokens ?? undefined
+  const inputTokens = ProviderShared.sumTokens(nonCached, cacheRead, cacheWrite)
   return new Usage({
-    inputTokens: usage.input_tokens,
+    inputTokens,
     outputTokens: usage.output_tokens,
-    cacheReadInputTokens: usage.cache_read_input_tokens ?? undefined,
-    cacheWriteInputTokens: usage.cache_creation_input_tokens ?? undefined,
-    totalTokens: ProviderShared.totalTokens(usage.input_tokens, usage.output_tokens, undefined),
-    native: usage,
+    nonCachedInputTokens: nonCached,
+    cacheReadInputTokens: cacheRead,
+    cacheWriteInputTokens: cacheWrite,
+    totalTokens: ProviderShared.totalTokens(inputTokens, usage.output_tokens, undefined),
+    providerMetadata: { anthropic: usage },
   })
 }
 
 // Anthropic emits usage on `message_start` and again on `message_delta` — the
 // final delta carries the authoritative totals. Right-biased merge: each
-// field prefers `right` when defined, falls back to `left`. `totalTokens` is
-// recomputed from the merged input/output to stay consistent.
+// field prefers `right` when defined, falls back to `left`. `inputTokens` is
+// recomputed from the merged breakdown so the inclusive total stays
+// consistent with `nonCached + cacheRead + cacheWrite`.
 const mergeUsage = (left: Usage | undefined, right: Usage | undefined) => {
   if (!left) return right
   if (!right) return left
-  const inputTokens = right.inputTokens ?? left.inputTokens
+  const nonCachedInputTokens = right.nonCachedInputTokens ?? left.nonCachedInputTokens
+  const cacheReadInputTokens = right.cacheReadInputTokens ?? left.cacheReadInputTokens
+  const cacheWriteInputTokens = right.cacheWriteInputTokens ?? left.cacheWriteInputTokens
+  const inputTokens = ProviderShared.sumTokens(nonCachedInputTokens, cacheReadInputTokens, cacheWriteInputTokens)
   const outputTokens = right.outputTokens ?? left.outputTokens
   return new Usage({
     inputTokens,
     outputTokens,
-    cacheReadInputTokens: right.cacheReadInputTokens ?? left.cacheReadInputTokens,
-    cacheWriteInputTokens: right.cacheWriteInputTokens ?? left.cacheWriteInputTokens,
+    nonCachedInputTokens,
+    cacheReadInputTokens,
+    cacheWriteInputTokens,
     totalTokens: ProviderShared.totalTokens(inputTokens, outputTokens, undefined),
-    native: { ...left.native, ...right.native },
+    providerMetadata: {
+      anthropic: {
+        ...(left.providerMetadata?.["anthropic"] ?? {}),
+        ...(right.providerMetadata?.["anthropic"] ?? {}),
+      },
+    },
   })
 }
 
@@ -415,14 +477,13 @@ const serverToolResultEvent = (block: NonNullable<AnthropicEvent["content_block"
       ? String((block.content as Record<string, unknown>).type)
       : ""
   const isError = errorPayload.endsWith("_tool_result_error")
-  return {
-    type: "tool-result",
+  return LLMEvent.toolResult({
     id: block.tool_use_id ?? "",
     name: SERVER_TOOL_RESULT_NAMES[block.type],
     result: isError ? { type: "error", value: block.content } : { type: "json", value: block.content },
     providerExecuted: true,
     providerMetadata: anthropicMetadata({ blockType: block.type }),
-  }
+  })
 }
 
 type StepResult = readonly [ParserState, ReadonlyArray<LLMEvent>]
@@ -453,18 +514,17 @@ const onContentBlockStart = (state: ParserState, event: AnthropicEvent): StepRes
   }
 
   if (block.type === "text" && block.text) {
-    return [state, [{ type: "text-delta", text: block.text }]]
+    return [state, [LLMEvent.textDelta({ id: `text-${event.index ?? 0}`, text: block.text })]]
   }
 
   if (block.type === "thinking" && block.thinking) {
     return [
       state,
       [
-        {
-          type: "reasoning-delta",
+        LLMEvent.reasoningDelta({
+          id: `reasoning-${event.index ?? 0}`,
           text: block.thinking,
-          ...(block.signature ? { providerMetadata: anthropicMetadata({ signature: block.signature }) } : {}),
-        },
+        }),
       ],
     ]
   }
@@ -480,17 +540,25 @@ const onContentBlockDelta = Effect.fn("AnthropicMessages.onContentBlockDelta")(f
   const delta = event.delta
 
   if (delta?.type === "text_delta" && delta.text) {
-    return [state, [{ type: "text-delta", text: delta.text }]] satisfies StepResult
+    return [state, [LLMEvent.textDelta({ id: `text-${event.index ?? 0}`, text: delta.text })]] satisfies StepResult
   }
 
   if (delta?.type === "thinking_delta" && delta.thinking) {
-    return [state, [{ type: "reasoning-delta", text: delta.thinking }]] satisfies StepResult
+    return [
+      state,
+      [LLMEvent.reasoningDelta({ id: `reasoning-${event.index ?? 0}`, text: delta.thinking })],
+    ] satisfies StepResult
   }
 
   if (delta?.type === "signature_delta" && delta.signature) {
     return [
       state,
-      [{ type: "reasoning-delta", text: "", providerMetadata: anthropicMetadata({ signature: delta.signature }) }],
+      [
+        LLMEvent.reasoningEnd({
+          id: `reasoning-${event.index ?? 0}`,
+          providerMetadata: anthropicMetadata({ signature: delta.signature }),
+        }),
+      ],
     ] satisfies StepResult
   }
 
@@ -524,21 +592,20 @@ const onMessageDelta = (state: ParserState, event: AnthropicEvent): StepResult =
   return [
     { ...state, usage },
     [
-      {
-        type: "request-finish",
+      LLMEvent.requestFinish({
         reason: mapFinishReason(event.delta?.stop_reason),
         usage,
-        ...(event.delta?.stop_sequence
-          ? { providerMetadata: anthropicMetadata({ stopSequence: event.delta.stop_sequence }) }
-          : {}),
-      },
+        providerMetadata: event.delta?.stop_sequence
+          ? anthropicMetadata({ stopSequence: event.delta.stop_sequence })
+          : undefined,
+      }),
     ],
   ]
 }
 
 const onError = (state: ParserState, event: AnthropicEvent): StepResult => [
   state,
-  [{ type: "provider-error", message: event.error?.message ?? "Anthropic Messages stream error" }],
+  [LLMEvent.providerError({ message: event.error?.message ?? "Anthropic Messages stream error" })],
 ]
 
 const step = (state: ParserState, event: AnthropicEvent) => {

packages/llm/src/protocols/bedrock-converse.ts

@@ -3,10 +3,10 @@ import { Route, type RouteModelInput } from "../route/client"
 import { Endpoint } from "../route/endpoint"
 import { Protocol } from "../route/protocol"
 import {
+  LLMEvent,
   Usage,
   type CacheHint,
   type FinishReason,
-  type LLMEvent,
   type LLMRequest,
   type ToolCallPart,
   type ToolDefinition,
@@ -108,7 +108,7 @@ type BedrockMessage = Schema.Schema.Type<typeof BedrockMessage>
 const BedrockSystemBlock = Schema.Union([BedrockTextBlock, BedrockCache.CachePointBlock])
 type BedrockSystemBlock = Schema.Schema.Type<typeof BedrockSystemBlock>
 
-const BedrockTool = Schema.Struct({
+const BedrockToolSpec = Schema.Struct({
   toolSpec: Schema.Struct({
     name: Schema.String,
     description: Schema.String,
@@ -117,6 +117,9 @@ const BedrockTool = Schema.Struct({
     }),
   }),
 })
+type BedrockToolSpec = Schema.Schema.Type<typeof BedrockToolSpec>
+
+const BedrockTool = Schema.Union([BedrockToolSpec, BedrockCache.CachePointBlock])
 type BedrockTool = Schema.Schema.Type<typeof BedrockTool>
 
 const BedrockToolChoice = Schema.Union([
@@ -214,7 +217,7 @@ type BedrockEvent = Schema.Schema.Type<typeof BedrockEvent>
 // =============================================================================
 // Request Lowering
 // =============================================================================
-const lowerTool = (tool: ToolDefinition): BedrockTool => ({
+const lowerToolSpec = (tool: ToolDefinition): BedrockToolSpec => ({
   toolSpec: {
     name: tool.name,
     description: tool.description,
@@ -222,11 +225,22 @@ const lowerTool = (tool: ToolDefinition): BedrockTool => ({
   },
 })
 
+const lowerTools = (breakpoints: BedrockCache.Breakpoints, tools: ReadonlyArray<ToolDefinition>): BedrockTool[] => {
+  const result: BedrockTool[] = []
+  for (const tool of tools) {
+    result.push(lowerToolSpec(tool))
+    const cachePoint = BedrockCache.block(breakpoints, tool.cache)
+    if (cachePoint) result.push(cachePoint)
+  }
+  return result
+}
+
 const textWithCache = (
+  breakpoints: BedrockCache.Breakpoints,
   text: string,
   cache: CacheHint | undefined,
 ): Array<BedrockTextBlock | BedrockCache.CachePointBlock> => {
-  const cachePoint = BedrockCache.block(cache)
+  const cachePoint = BedrockCache.block(breakpoints, cache)
   return cachePoint ? [{ text }, cachePoint] : [{ text }]
 }
 
@@ -257,7 +271,10 @@ const lowerToolResult = (part: ToolResultPart): BedrockToolResultBlock => ({
   },
 })
 
-const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (request: LLMRequest) {
+const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (
+  request: LLMRequest,
+  breakpoints: BedrockCache.Breakpoints,
+) {
   const messages: BedrockMessage[] = []
 
   for (const message of request.messages) {
@@ -267,7 +284,7 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (requ
         if (!ProviderShared.supportsContent(part, ["text", "media"]))
           return yield* ProviderShared.unsupportedContent("Bedrock Converse", "user", ["text", "media"])
         if (part.type === "text") {
-          content.push(...textWithCache(part.text, part.cache))
+          content.push(...textWithCache(breakpoints, part.text, part.cache))
           continue
         }
         if (part.type === "media") {
@@ -289,7 +306,7 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (requ
             "tool-call",
           ])
         if (part.type === "text") {
-          content.push(...textWithCache(part.text, part.cache))
+          content.push(...textWithCache(breakpoints, part.text, part.cache))
           continue
         }
         if (part.type === "reasoning") {
@@ -309,11 +326,13 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (requ
       continue
     }
 
-    const content: BedrockToolResultBlock[] = []
+    const content: BedrockUserBlock[] = []
     for (const part of message.content) {
       if (!ProviderShared.supportsContent(part, ["tool-result"]))
         return yield* ProviderShared.unsupportedContent("Bedrock Converse", "tool", ["tool-result"])
       content.push(lowerToolResult(part))
+      const cachePoint = BedrockCache.block(breakpoints, part.cache)
+      if (cachePoint) content.push(cachePoint)
     }
     messages.push({ role: "user", content })
   }
@@ -323,16 +342,32 @@ const lowerMessages = Effect.fn("BedrockConverse.lowerMessages")(function* (requ
 
 // System prompts share the cache-point convention: emit the text block, then
 // optionally a positional `cachePoint` marker.
-const lowerSystem = (system: ReadonlyArray<LLMRequest["system"][number]>): BedrockSystemBlock[] =>
-  system.flatMap((part) => textWithCache(part.text, part.cache))
+const lowerSystem = (
+  breakpoints: BedrockCache.Breakpoints,
+  system: ReadonlyArray<LLMRequest["system"][number]>,
+): BedrockSystemBlock[] => system.flatMap((part) => textWithCache(breakpoints, part.text, part.cache))
 
 const fromRequest = Effect.fn("BedrockConverse.fromRequest")(function* (request: LLMRequest) {
   const toolChoice = request.toolChoice ? yield* lowerToolChoice(request.toolChoice) : undefined
   const generation = request.generation
+  // Bedrock-Claude shares Anthropic's 4-breakpoint cap. Spend the budget in
+  // tools → system → messages order to favour the highest-impact prefixes.
+  const breakpoints = BedrockCache.breakpoints()
+  const toolConfig =
+    request.tools.length > 0 && request.toolChoice?.type !== "none"
+      ? { tools: lowerTools(breakpoints, request.tools), toolChoice }
+      : undefined
+  const system = request.system.length === 0 ? undefined : lowerSystem(breakpoints, request.system)
+  const messages = yield* lowerMessages(request, breakpoints)
+  if (breakpoints.dropped > 0) {
+    yield* Effect.logWarning(
+      `Bedrock Converse: dropped ${breakpoints.dropped} cache breakpoint(s); the API allows at most ${BedrockCache.BEDROCK_BREAKPOINT_CAP} per request.`,
+    )
+  }
   return {
     modelId: request.model.id,
-    messages: yield* lowerMessages(request),
-    system: request.system.length === 0 ? undefined : lowerSystem(request.system),
+    messages,
+    system,
     inferenceConfig:
       generation?.maxTokens === undefined &&
       generation?.temperature === undefined &&
@@ -345,10 +380,7 @@ const fromRequest = Effect.fn("BedrockConverse.fromRequest")(function* (request:
             topP: generation?.topP,
             stopSequences: generation?.stop,
           },
-    toolConfig:
-      request.tools.length > 0 && request.toolChoice?.type !== "none"
-        ? { tools: request.tools.map(lowerTool), toolChoice }
-        : undefined,
+    toolConfig,
   }
 })
 
@@ -363,15 +395,22 @@ const mapFinishReason = (reason: string): FinishReason => {
   return "unknown"
 }
 
+// AWS Bedrock Converse reports `inputTokens` (inclusive total) with
+// `cacheReadInputTokens` and `cacheWriteInputTokens` as subsets. Pass
+// the total through and derive the non-cached breakdown. Bedrock does
+// not break reasoning out of `outputTokens` for any current model.
 const mapUsage = (usage: BedrockUsageSchema | undefined): Usage | undefined => {
   if (!usage) return undefined
+  const cacheTotal = (usage.cacheReadInputTokens ?? 0) + (usage.cacheWriteInputTokens ?? 0)
+  const nonCached = ProviderShared.subtractTokens(usage.inputTokens, cacheTotal)
   return new Usage({
     inputTokens: usage.inputTokens,
     outputTokens: usage.outputTokens,
-    totalTokens: ProviderShared.totalTokens(usage.inputTokens, usage.outputTokens, usage.totalTokens),
+    nonCachedInputTokens: nonCached,
     cacheReadInputTokens: usage.cacheReadInputTokens,
     cacheWriteInputTokens: usage.cacheWriteInputTokens,
-    native: usage,
+    totalTokens: ProviderShared.totalTokens(usage.inputTokens, usage.outputTokens, usage.totalTokens),
+    providerMetadata: { bedrock: usage },
   })
 }
 
@@ -400,13 +439,26 @@ const step = (state: ParserState, event: BedrockEvent) =>
     }
 
     if (event.contentBlockDelta?.delta?.text) {
-      return [state, [{ type: "text-delta" as const, text: event.contentBlockDelta.delta.text }]] as const
+      return [
+        state,
+        [
+          LLMEvent.textDelta({
+            id: `text-${event.contentBlockDelta.contentBlockIndex}`,
+            text: event.contentBlockDelta.delta.text,
+          }),
+        ],
+      ] as const
     }
 
     if (event.contentBlockDelta?.delta?.reasoningContent?.text) {
       return [
         state,
-        [{ type: "reasoning-delta" as const, text: event.contentBlockDelta.delta.reasoningContent.text }],
+        [
+          LLMEvent.reasoningDelta({
+            id: `reasoning-${event.contentBlockDelta.contentBlockIndex}`,
+            text: event.contentBlockDelta.delta.reasoningContent.text,
+          }),
+        ],
       ] as const
     }
 
@@ -449,16 +501,13 @@ const step = (state: ParserState, event: BedrockEvent) =>
         event.modelStreamErrorException?.message ??
         event.serviceUnavailableException?.message ??
         "Bedrock Converse stream error"
-      return [state, [{ type: "provider-error" as const, message, retryable: true }]] as const
+      return [state, [LLMEvent.providerError({ message, retryable: true })]] as const
     }
 
     if (event.validationException || event.throttlingException) {
       const message =
         event.validationException?.message ?? event.throttlingException?.message ?? "Bedrock Converse error"
-      return [
-        state,
-        [{ type: "provider-error" as const, message, retryable: event.throttlingException !== undefined }],
-      ] as const
+      return [state, [LLMEvent.providerError({ message, retryable: event.throttlingException !== undefined })]] as const
     }
 
     return [state, []] as const
@@ -468,7 +517,7 @@ const framing = BedrockEventStream.framing(ADAPTER)
 
 const onHalt = (state: ParserState): ReadonlyArray<LLMEvent> =>
   state.pendingFinish
-    ? [{ type: "request-finish", reason: state.pendingFinish.reason, usage: state.pendingFinish.usage }]
+    ? [LLMEvent.requestFinish({ reason: state.pendingFinish.reason, usage: state.pendingFinish.usage })]
     : []
 
 // =============================================================================

packages/llm/src/protocols/gemini.ts

@@ -5,9 +5,9 @@ import { Endpoint } from "../route/endpoint"
 import { Framing } from "../route/framing"
 import { Protocol } from "../route/protocol"
 import {
+  LLMEvent,
   Usage,
   type FinishReason,
-  type LLMEvent,
   type LLMRequest,
   type MediaPart,
   type TextPart,
@@ -281,15 +281,28 @@ const fromRequest = Effect.fn("Gemini.fromRequest")(function* (request: LLMReque
 // =============================================================================
 // Stream Parsing
 // =============================================================================
+// Gemini reports `promptTokenCount` (inclusive total) with a
+// `cachedContentTokenCount` subset. `candidatesTokenCount` is *exclusive*
+// of `thoughtsTokenCount` — visible-only, not a total — so we sum the two
+// to produce the inclusive `outputTokens` the rest of the contract expects.
 const mapUsage = (usage: GeminiUsage | undefined) => {
   if (!usage) return undefined
+  const cached = usage.cachedContentTokenCount
+  const nonCached = ProviderShared.subtractTokens(usage.promptTokenCount, cached)
+  // `candidatesTokenCount` is visible-only; sum with thoughts to produce the
+  // inclusive `outputTokens` the contract expects. Only compute the total
+  // when the visible component is reported — otherwise we'd fabricate an
+  // inclusive number from a partial breakdown.
+  const outputTokens =
+    usage.candidatesTokenCount !== undefined ? usage.candidatesTokenCount + (usage.thoughtsTokenCount ?? 0) : undefined
   return new Usage({
     inputTokens: usage.promptTokenCount,
-    outputTokens: usage.candidatesTokenCount,
+    outputTokens,
+    nonCachedInputTokens: nonCached,
+    cacheReadInputTokens: cached,
     reasoningTokens: usage.thoughtsTokenCount,
-    cacheReadInputTokens: usage.cachedContentTokenCount,
-    totalTokens: ProviderShared.totalTokens(usage.promptTokenCount, usage.candidatesTokenCount, usage.totalTokenCount),
-    native: usage,
+    totalTokens: ProviderShared.totalTokens(usage.promptTokenCount, outputTokens, usage.totalTokenCount),
+    providerMetadata: { google: usage },
   })
 }
 
@@ -311,7 +324,7 @@ const mapFinishReason = (finishReason: string | undefined, hasToolCalls: boolean
 
 const finish = (state: ParserState): ReadonlyArray<LLMEvent> =>
   state.finishReason || state.usage
-    ? [{ type: "request-finish", reason: mapFinishReason(state.finishReason, state.hasToolCalls), usage: state.usage }]
+    ? [LLMEvent.requestFinish({ reason: mapFinishReason(state.finishReason, state.hasToolCalls), usage: state.usage })]
     : []
 
 const step = (state: ParserState, event: GeminiEvent) => {
@@ -332,14 +345,18 @@ const step = (state: ParserState, event: GeminiEvent) => {
 
   for (const part of candidate.content.parts) {
     if ("text" in part && part.text.length > 0) {
-      events.push({ type: part.thought ? "reasoning-delta" : "text-delta", text: part.text })
+      events.push(
+        part.thought
+          ? LLMEvent.reasoningDelta({ id: "reasoning-0", text: part.text })
+          : LLMEvent.textDelta({ id: "text-0", text: part.text }),
+      )
       continue
     }
 
     if ("functionCall" in part) {
       const input = part.functionCall.args
       const id = `tool_${nextToolCallId++}`
-      events.push({ type: "tool-call", id, name: part.functionCall.name, input })
+      events.push(LLMEvent.toolCall({ id, name: part.functionCall.name, input }))
       hasToolCalls = true
     }
   }

packages/llm/src/protocols/openai-chat.ts

@@ -6,9 +6,9 @@ import { Framing } from "../route/framing"
 import { HttpTransport } from "../route/transport"
 import { Protocol } from "../route/protocol"
 import {
+  LLMEvent,
   Usage,
   type FinishReason,
-  type LLMEvent,
   type LLMRequest,
   type TextPart,
   type ToolCallPart,
@@ -290,15 +290,24 @@ const mapFinishReason = (reason: string | null | undefined): FinishReason => {
   return "unknown"
 }
 
+// OpenAI Chat reports `prompt_tokens` (inclusive total) with a
+// `cached_tokens` subset, and `completion_tokens` (inclusive total) with
+// a `reasoning_tokens` subset. We pass the inclusive totals through and
+// derive the non-cached breakdown so the `LLM.Usage` contract is
+// satisfied on both sides.
 const mapUsage = (usage: OpenAIChatEvent["usage"]): Usage | undefined => {
   if (!usage) return undefined
+  const cached = usage.prompt_tokens_details?.cached_tokens
+  const reasoning = usage.completion_tokens_details?.reasoning_tokens
+  const nonCached = ProviderShared.subtractTokens(usage.prompt_tokens, cached)
   return new Usage({
     inputTokens: usage.prompt_tokens,
     outputTokens: usage.completion_tokens,
-    reasoningTokens: usage.completion_tokens_details?.reasoning_tokens,
-    cacheReadInputTokens: usage.prompt_tokens_details?.cached_tokens,
+    nonCachedInputTokens: nonCached,
+    cacheReadInputTokens: cached,
+    reasoningTokens: reasoning,
     totalTokens: ProviderShared.totalTokens(usage.prompt_tokens, usage.completion_tokens, usage.total_tokens),
-    native: usage,
+    providerMetadata: { openai: usage },
   })
 }
 
@@ -312,7 +321,7 @@ const step = (state: ParserState, event: OpenAIChatEvent) =>
     const toolDeltas = delta?.tool_calls ?? []
     let tools = state.tools
 
-    if (delta?.content) events.push({ type: "text-delta", text: delta.content })
+    if (delta?.content) events.push(LLMEvent.textDelta({ id: "text-0", text: delta.content }))
 
     for (const tool of toolDeltas) {
       const result = ToolStream.appendOrStart(
@@ -348,10 +357,7 @@ const step = (state: ParserState, event: OpenAIChatEvent) =>
 const finishEvents = (state: ParserState): ReadonlyArray<LLMEvent> => {
   const hasToolCalls = state.toolCallEvents.length > 0
   const reason = state.finishReason === "stop" && hasToolCalls ? "tool-calls" : state.finishReason
-  return [
-    ...state.toolCallEvents,
-    ...(reason ? ([{ type: "request-finish", reason, usage: state.usage }] satisfies ReadonlyArray<LLMEvent>) : []),
-  ]
+  return [...state.toolCallEvents, ...(reason ? [LLMEvent.requestFinish({ reason, usage: state.usage })] : [])]
 }
 
 // =============================================================================

packages/llm/src/protocols/openai-responses.ts

@@ -6,9 +6,9 @@ import { Framing } from "../route/framing"
 import { HttpTransport, WebSocketTransport } from "../route/transport"
 import { Protocol } from "../route/protocol"
 import {
+  LLMEvent,
   Usage,
   type FinishReason,
-  type LLMEvent,
   type LLMRequest,
   type ProviderMetadata,
   type TextPart,
@@ -276,15 +276,23 @@ const fromRequest = Effect.fn("OpenAIResponses.fromRequest")(function* (request:
 // =============================================================================
 // Stream Parsing
 // =============================================================================
+// OpenAI Responses reports `input_tokens` (inclusive total) with a
+// `cached_tokens` subset, and `output_tokens` (inclusive total) with a
+// `reasoning_tokens` subset. Pass the totals through and derive the
+// non-cached breakdown.
 const mapUsage = (usage: OpenAIResponsesUsage | null | undefined) => {
   if (!usage) return undefined
+  const cached = usage.input_tokens_details?.cached_tokens
+  const reasoning = usage.output_tokens_details?.reasoning_tokens
+  const nonCached = ProviderShared.subtractTokens(usage.input_tokens, cached)
   return new Usage({
     inputTokens: usage.input_tokens,
     outputTokens: usage.output_tokens,
-    reasoningTokens: usage.output_tokens_details?.reasoning_tokens,
-    cacheReadInputTokens: usage.input_tokens_details?.cached_tokens,
+    nonCachedInputTokens: nonCached,
+    cacheReadInputTokens: cached,
+    reasoningTokens: reasoning,
     totalTokens: ProviderShared.totalTokens(usage.input_tokens, usage.output_tokens, usage.total_tokens),
-    native: usage,
+    providerMetadata: { openai: usage },
   })
 }
 
@@ -348,22 +356,20 @@ const hostedToolEvents = (
   const tool = HOSTED_TOOLS[item.type]
   const providerMetadata = openaiMetadata({ itemId: item.id })
   return [
-    {
-      type: "tool-call",
+    LLMEvent.toolCall({
       id: item.id,
       name: tool.name,
       input: tool.input(item),
       providerExecuted: true,
       providerMetadata,
-    },
-    {
-      type: "tool-result",
+    }),
+    LLMEvent.toolResult({
       id: item.id,
       name: tool.name,
       result: hostedToolResult(item),
       providerExecuted: true,
       providerMetadata,
-    },
+    }),
   ]
 }
 
@@ -379,17 +385,7 @@ const TERMINAL_TYPES = new Set(["response.completed", "response.incomplete", "re
 
 const onOutputTextDelta = (state: ParserState, event: OpenAIResponsesEvent): StepResult => {
   if (!event.delta) return [state, NO_EVENTS]
-  return [
-    state,
-    [
-      {
-        type: "text-delta",
-        id: event.item_id,
-        text: event.delta,
-        ...(event.item_id ? { providerMetadata: openaiMetadata({ itemId: event.item_id }) } : {}),
-      },
-    ],
-  ]
+  return [state, [LLMEvent.textDelta({ id: event.item_id ?? "text-0", text: event.delta })]]
 }
 
 const onOutputItemAdded = (state: ParserState, event: OpenAIResponsesEvent): StepResult => {
@@ -458,30 +454,28 @@ const onOutputItemDone = Effect.fn("OpenAIResponses.onOutputItemDone")(function*
 const onResponseFinish = (state: ParserState, event: OpenAIResponsesEvent): StepResult => [
   state,
   [
-    {
-      type: "request-finish",
+    LLMEvent.requestFinish({
       reason: mapFinishReason(event, state.hasFunctionCall),
       usage: mapUsage(event.response?.usage),
-      ...(event.response?.id || event.response?.service_tier
-        ? {
-            providerMetadata: openaiMetadata({
+      providerMetadata:
+        event.response?.id || event.response?.service_tier
+          ? openaiMetadata({
               responseId: event.response.id,
               serviceTier: event.response.service_tier,
-            }),
-          }
-        : {}),
-    },
+            })
+          : undefined,
+    }),
   ],
 ]
 
 const onResponseFailed = (state: ParserState, event: OpenAIResponsesEvent): StepResult => [
   state,
-  [{ type: "provider-error", message: event.message ?? event.code ?? "OpenAI Responses response failed" }],
+  [LLMEvent.providerError({ message: event.message ?? event.code ?? "OpenAI Responses response failed" })],
 ]
 
 const onError = (state: ParserState, event: OpenAIResponsesEvent): StepResult => [
   state,
-  [{ type: "provider-error", message: event.message ?? event.code ?? "OpenAI Responses stream error" }],
+  [LLMEvent.providerError({ message: event.message ?? event.code ?? "OpenAI Responses stream error" })],
 ]
 
 const step = (state: ParserState, event: OpenAIResponsesEvent) => {

packages/llm/src/protocols/shared.ts

@@ -42,6 +42,13 @@ export interface ToolAccumulator {
  * supplied total; otherwise falls back to `inputTokens + outputTokens` only
  * when at least one is defined. Returns `undefined` when neither input nor
  * output is known so routes don't publish a misleading `0`.
+ *
+ * Under the additive `LLM.Usage` contract, `inputTokens` and `outputTokens`
+ * are the non-cached input and visible output only. The provider-supplied
+ * `total` is the source of truth when present; the computed fallback
+ * under-counts cache and reasoning by design and exists mainly so
+ * Anthropic-style providers (which don't surface a total) still get a
+ * sensible aggregate on the input + output axes.
  */
 export const totalTokens = (
   inputTokens: number | undefined,
@@ -53,6 +60,35 @@ export const totalTokens = (
   return (inputTokens ?? 0) + (outputTokens ?? 0)
 }
 
+/**
+ * Subtract `subtrahend` from `total`, clamping to zero if the provider
+ * reports a non-sensical breakdown (e.g. `cached_tokens > prompt_tokens`).
+ * Used by protocol mappers when deriving a non-overlapping breakdown field
+ * from a provider's inclusive total — `nonCachedInputTokens` from
+ * `inputTokens - cacheReadInputTokens - cacheWriteInputTokens`.
+ *
+ * If `total` is `undefined`, returns `undefined` (we don't fabricate
+ * counts). If `subtrahend` is `undefined`, returns `total` unchanged. The
+ * provider-native breakdown stays available on `Usage.native` for debugging.
+ */
+export const subtractTokens = (total: number | undefined, subtrahend: number | undefined): number | undefined => {
+  if (total === undefined) return undefined
+  if (subtrahend === undefined) return total
+  return Math.max(0, total - subtrahend)
+}
+
+/**
+ * Sum a list of optional token counts, returning `undefined` only when
+ * every value is `undefined` (so we don't fabricate a `0`). Used by
+ * protocol mappers to derive the inclusive `inputTokens` total from a
+ * provider that natively reports a non-overlapping breakdown
+ * (e.g. Anthropic, whose `input_tokens` is already non-cached only).
+ */
+export const sumTokens = (...values: ReadonlyArray<number | undefined>): number | undefined => {
+  if (values.every((value) => value === undefined)) return undefined
+  return values.reduce<number>((acc, value) => acc + (value ?? 0), 0)
+}
+
 export const eventError = (route: string, message: string, raw?: string) =>
   new LLMError({
     module: "ProviderShared",

packages/llm/src/protocols/utils/bedrock-cache.ts

@@ -1,20 +1,37 @@
 import { Schema } from "effect"
 import type { CacheHint } from "../../schema"
+import { newBreakpoints, ttlBucket, type Breakpoints } from "./cache"
 
 // Bedrock cache markers are positional: emit a `cachePoint` block immediately
-// after the content the caller wants treated as a cacheable prefix.
+// after the content the caller wants treated as a cacheable prefix. Bedrock
+// accepts optional `ttl: "5m" | "1h"` on cachePoint, mirroring Anthropic.
 export const CachePointBlock = Schema.Struct({
-  cachePoint: Schema.Struct({ type: Schema.tag("default") }),
+  cachePoint: Schema.Struct({
+    type: Schema.tag("default"),
+    ttl: Schema.optional(Schema.Literals(["5m", "1h"])),
+  }),
 })
 export type CachePointBlock = Schema.Schema.Type<typeof CachePointBlock>
 
-// Bedrock recently added optional `ttl: "5m" | "1h"` on cachePoint. Map
-// `CacheHint.ttlSeconds` here once a recorded cassette validates the wire shape.
-const DEFAULT: CachePointBlock = { cachePoint: { type: "default" } }
+// Bedrock-Claude enforces the same 4-breakpoint cap as the Anthropic Messages
+// API. Callers pass a shared counter through every `block()` call site so the
+// budget is respected across `system`, `messages`, and `tools`.
+export const BEDROCK_BREAKPOINT_CAP = 4
 
-export const block = (cache: CacheHint | undefined): CachePointBlock | undefined => {
+export type { Breakpoints } from "./cache"
+export const breakpoints = () => newBreakpoints(BEDROCK_BREAKPOINT_CAP)
+
+const DEFAULT_5M: CachePointBlock = { cachePoint: { type: "default" } }
+const DEFAULT_1H: CachePointBlock = { cachePoint: { type: "default", ttl: "1h" } }
+
+export const block = (breakpoints: Breakpoints, cache: CacheHint | undefined): CachePointBlock | undefined => {
   if (cache?.type !== "ephemeral" && cache?.type !== "persistent") return undefined
-  return DEFAULT
+  if (breakpoints.remaining <= 0) {
+    breakpoints.dropped += 1
+    return undefined
+  }
+  breakpoints.remaining -= 1
+  return ttlBucket(cache.ttlSeconds) === "1h" ? DEFAULT_1H : DEFAULT_5M
 }
 
 export * as BedrockCache from "./bedrock-cache"

packages/llm/src/protocols/utils/cache.ts

@@ -0,0 +1,16 @@
+// Shared helpers for provider cache-marker lowering. Anthropic and Bedrock
+// both enforce a 4-breakpoint cap per request and accept the same `5m`/`1h`
+// TTL buckets, so the counter and TTL mapping live here.
+
+export interface Breakpoints {
+  remaining: number
+  dropped: number
+}
+
+export const newBreakpoints = (cap: number): Breakpoints => ({ remaining: cap, dropped: 0 })
+
+// Returns `"1h"` for any `ttlSeconds >= 3600`, otherwise `undefined` (the
+// provider default 5m). Anthropic & Bedrock both treat anything shorter than
+// an hour as 5m.
+export const ttlBucket = (ttlSeconds: number | undefined): "1h" | undefined =>
+  ttlSeconds !== undefined && ttlSeconds >= 3600 ? "1h" : undefined

packages/llm/src/protocols/utils/tool-stream.ts

@@ -1,5 +1,5 @@
 import { Effect } from "effect"
-import { LLMError, type ProviderMetadata, type ToolCall, type ToolInputDelta } from "../../schema"
+import { LLMError, LLMEvent, type ProviderMetadata, type ToolCall, type ToolInputDelta } from "../../schema"
 import { eventError, parseToolInput, type ToolAccumulator } from "../shared"
 
 type StreamKey = string | number
@@ -49,34 +49,24 @@ const withoutTool = <K extends StreamKey>(tools: State<K>, key: K): State<K> =>
   return next
 }
 
-const inputDelta = (tool: PendingTool, text: string): ToolInputDelta => ({
-  type: "tool-input-delta",
-  id: tool.id,
-  name: tool.name,
-  text,
-  ...(tool.providerMetadata ? { providerMetadata: tool.providerMetadata } : {}),
-})
+const inputDelta = (tool: PendingTool, text: string): ToolInputDelta =>
+  LLMEvent.toolInputDelta({
+    id: tool.id,
+    name: tool.name,
+    text,
+  })
 
 const toolCall = (route: string, tool: PendingTool, inputOverride?: string) =>
   parseToolInput(route, tool.name, inputOverride ?? tool.input).pipe(
     Effect.map(
       (input): ToolCall =>
-        tool.providerExecuted
-          ? {
-              type: "tool-call",
-              id: tool.id,
-              name: tool.name,
-              input,
-              providerExecuted: true,
-              ...(tool.providerMetadata ? { providerMetadata: tool.providerMetadata } : {}),
-            }
-          : {
-              type: "tool-call",
-              id: tool.id,
-              name: tool.name,
-              input,
-              ...(tool.providerMetadata ? { providerMetadata: tool.providerMetadata } : {}),
-            },
+        LLMEvent.toolCall({
+          id: tool.id,
+          name: tool.name,
+          input,
+          providerExecuted: tool.providerExecuted ? true : undefined,
+          providerMetadata: tool.providerMetadata,
+        }),
     ),
   )
 

packages/llm/src/route/client.ts

@@ -8,6 +8,7 @@ import type { Transport, TransportRuntime } from "./transport"
 import { WebSocketExecutor } from "./transport"
 import type { Service as WebSocketExecutorService } from "./transport/websocket"
 import type { Protocol } from "./protocol"
+import { applyCachePolicy } from "../cache-policy"
 import * as ProviderShared from "../protocols/shared"
 import * as ToolRuntime from "../tool-runtime"
 import type { Tools } from "../tool"
@@ -400,7 +401,7 @@ export function make<Body, Prepared, Frame, Event, State>(
 // validated provider body plus transport-private prepared data, but does not
 // execute transport.
 const compile = Effect.fn("LLM.compile")(function* (request: LLMRequest) {
-  const resolved = resolveRequestOptions(request)
+  const resolved = applyCachePolicy(resolveRequestOptions(request))
   const route = registeredRoute(resolved.model.route)
   if (!route) return yield* noRoute(resolved.model)
 

packages/llm/src/schema/events.ts

@@ -1,73 +1,155 @@
 import { Schema } from "effect"
-import { FinishReason, ProtocolID, ProviderMetadata, RouteID } from "./ids"
+import { ContentBlockID, FinishReason, ProtocolID, ProviderMetadata, ResponseID, RouteID, ToolCallID } from "./ids"
 import { ModelRef } from "./options"
 import { ToolResultValue } from "./messages"
 
+/**
+ * Token usage reported by an LLM provider.
+ *
+ * **Inclusive totals** (match AI SDK / OpenAI / LangChain convention — a
+ * reader from any of those ecosystems sees the number they expect):
+ *
+ * - `inputTokens` — total prompt tokens, *including* cached reads/writes.
+ * - `outputTokens` — total output tokens, *including* reasoning.
+ * - `totalTokens` — provider-supplied total, or `inputTokens + outputTokens`.
+ *
+ * **Non-overlapping breakdown** (every field is independently meaningful;
+ * consumers never have to subtract):
+ *
+ * - `nonCachedInputTokens` — the "fresh" portion of the prompt.
+ * - `cacheReadInputTokens` — input tokens served from cache.
+ * - `cacheWriteInputTokens` — input tokens written to cache.
+ * - `reasoningTokens` — subset of `outputTokens` spent on hidden reasoning.
+ *
+ * **Invariant**: `nonCachedInputTokens + cacheReadInputTokens +
+ * cacheWriteInputTokens = inputTokens`, and `reasoningTokens ≤ outputTokens`.
+ * Each protocol mapper computes whichever side it doesn't get natively,
+ * with `Math.max(0, …)` clamping for defense against provider bugs. Because
+ * every breakdown field is stored independently, downstream consumers can
+ * read whatever they need (cost-by-category, context-pressure, AI-SDK-style
+ * inclusive total) without ever subtracting — eliminating the underflow
+ * class of bug where a clamped difference would silently store the wrong
+ * value.
+ *
+ * **Semantics by provider**:
+ *
+ * - OpenAI Chat / Responses / Gemini / Bedrock: provider reports inclusive
+ *   `inputTokens` and an inclusive `outputTokens`; mapper subtracts to
+ *   derive the breakdown.
+ * - Anthropic: provider reports the breakdown natively (`input_tokens` is
+ *   non-cached only); mapper sums to derive the inclusive `inputTokens`.
+ *   Anthropic does *not* break extended-thinking out of `output_tokens`, so
+ *   `reasoningTokens` is `undefined` and `outputTokens` carries the
+ *   combined total — a documented limitation of the Anthropic API.
+ *
+ * `providerMetadata` always carries the provider's raw usage payload —
+ * keyed by provider name (`{ openai: ... }`, `{ anthropic: ... }`, etc.)
+ * — for fields we don't normalize and for billing-level audit trails.
+ * Matches the same escape-hatch field on `LLMEvent`.
+ */
 export class Usage extends Schema.Class<Usage>("LLM.Usage")({
   inputTokens: Schema.optional(Schema.Number),
   outputTokens: Schema.optional(Schema.Number),
-  reasoningTokens: Schema.optional(Schema.Number),
+  nonCachedInputTokens: Schema.optional(Schema.Number),
   cacheReadInputTokens: Schema.optional(Schema.Number),
   cacheWriteInputTokens: Schema.optional(Schema.Number),
+  reasoningTokens: Schema.optional(Schema.Number),
   totalTokens: Schema.optional(Schema.Number),
-  native: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
-}) {}
+  providerMetadata: Schema.optional(ProviderMetadata),
+}) {
+  /**
+   * Visible output tokens — `outputTokens` minus `reasoningTokens`, clamped
+   * to zero. The one place subtraction happens in this contract; the clamp
+   * means a provider reporting `reasoningTokens > outputTokens` produces a
+   * harmless zero rather than a negative that crashes downstream schemas.
+   */
+  get visibleOutputTokens() {
+    return Math.max(0, (this.outputTokens ?? 0) - (this.reasoningTokens ?? 0))
+  }
+}
 
 export const RequestStart = Schema.Struct({
-  type: Schema.Literal("request-start"),
-  id: Schema.String,
+  type: Schema.tag("request-start"),
+  id: ResponseID,
   model: ModelRef,
 }).annotate({ identifier: "LLM.Event.RequestStart" })
 export type RequestStart = Schema.Schema.Type<typeof RequestStart>
 
 export const StepStart = Schema.Struct({
-  type: Schema.Literal("step-start"),
+  type: Schema.tag("step-start"),
   index: Schema.Number,
 }).annotate({ identifier: "LLM.Event.StepStart" })
 export type StepStart = Schema.Schema.Type<typeof StepStart>
 
 export const TextStart = Schema.Struct({
-  type: Schema.Literal("text-start"),
-  id: Schema.String,
+  type: Schema.tag("text-start"),
+  id: ContentBlockID,
   providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Event.TextStart" })
 export type TextStart = Schema.Schema.Type<typeof TextStart>
 
 export const TextDelta = Schema.Struct({
-  type: Schema.Literal("text-delta"),
-  id: Schema.optional(Schema.String),
+  type: Schema.tag("text-delta"),
+  id: ContentBlockID,
   text: Schema.String,
-  providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Event.TextDelta" })
 export type TextDelta = Schema.Schema.Type<typeof TextDelta>
 
 export const TextEnd = Schema.Struct({
-  type: Schema.Literal("text-end"),
-  id: Schema.String,
+  type: Schema.tag("text-end"),
+  id: ContentBlockID,
   providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Event.TextEnd" })
 export type TextEnd = Schema.Schema.Type<typeof TextEnd>
 
-export const ReasoningDelta = Schema.Struct({
-  type: Schema.Literal("reasoning-delta"),
-  id: Schema.optional(Schema.String),
-  text: Schema.String,
+export const ReasoningStart = Schema.Struct({
+  type: Schema.tag("reasoning-start"),
+  id: ContentBlockID,
   providerMetadata: Schema.optional(ProviderMetadata),
+}).annotate({ identifier: "LLM.Event.ReasoningStart" })
+export type ReasoningStart = Schema.Schema.Type<typeof ReasoningStart>
+
+export const ReasoningDelta = Schema.Struct({
+  type: Schema.tag("reasoning-delta"),
+  id: ContentBlockID,
+  text: Schema.String,
 }).annotate({ identifier: "LLM.Event.ReasoningDelta" })
 export type ReasoningDelta = Schema.Schema.Type<typeof ReasoningDelta>
 
+export const ReasoningEnd = Schema.Struct({
+  type: Schema.tag("reasoning-end"),
+  id: ContentBlockID,
+  providerMetadata: Schema.optional(ProviderMetadata),
+}).annotate({ identifier: "LLM.Event.ReasoningEnd" })
+export type ReasoningEnd = Schema.Schema.Type<typeof ReasoningEnd>
+
+export const ToolInputStart = Schema.Struct({
+  type: Schema.tag("tool-input-start"),
+  id: ToolCallID,
+  name: Schema.String,
+  providerMetadata: Schema.optional(ProviderMetadata),
+}).annotate({ identifier: "LLM.Event.ToolInputStart" })
+export type ToolInputStart = Schema.Schema.Type<typeof ToolInputStart>
+
 export const ToolInputDelta = Schema.Struct({
-  type: Schema.Literal("tool-input-delta"),
-  id: Schema.String,
+  type: Schema.tag("tool-input-delta"),
+  id: ToolCallID,
   name: Schema.String,
   text: Schema.String,
-  providerMetadata: Schema.optional(ProviderMetadata),
 }).annotate({ identifier: "LLM.Event.ToolInputDelta" })
 export type ToolInputDelta = Schema.Schema.Type<typeof ToolInputDelta>
 
+export const ToolInputEnd = Schema.Struct({
+  type: Schema.tag("tool-input-end"),
+  id: ToolCallID,
+  name: Schema.String,
+  providerMetadata: Schema.optional(ProviderMetadata),
+}).annotate({ identifier: "LLM.Event.ToolInputEnd" })
+export type ToolInputEnd = Schema.Schema.Type<typeof ToolInputEnd>
+
 export const ToolCall = Schema.Struct({
-  type: Schema.Literal("tool-call"),
-  id: Schema.String,
+  type: Schema.tag("tool-call"),
+  id: ToolCallID,
   name: Schema.String,
   input: Schema.Unknown,
   providerExecuted: Schema.optional(Schema.Boolean),
@@ -76,8 +158,8 @@ export const ToolCall = Schema.Struct({
 export type ToolCall = Schema.Schema.Type<typeof ToolCall>
 
 export const ToolResult = Schema.Struct({
-  type: Schema.Literal("tool-result"),
-  id: Schema.String,
+  type: Schema.tag("tool-result"),
+  id: ToolCallID,
   name: Schema.String,
   result: ToolResultValue,
   providerExecuted: Schema.optional(Schema.Boolean),
@@ -86,8 +168,8 @@ export const ToolResult = Schema.Struct({
 export type ToolResult = Schema.Schema.Type<typeof ToolResult>
 
 export const ToolError = Schema.Struct({
-  type: Schema.Literal("tool-error"),
-  id: Schema.String,
+  type: Schema.tag("tool-error"),
+  id: ToolCallID,
   name: Schema.String,
   message: Schema.String,
   providerMetadata: Schema.optional(ProviderMetadata),
@@ -95,7 +177,7 @@ export const ToolError = Schema.Struct({
 export type ToolError = Schema.Schema.Type<typeof ToolError>
 
 export const StepFinish = Schema.Struct({
-  type: Schema.Literal("step-finish"),
+  type: Schema.tag("step-finish"),
   index: Schema.Number,
   reason: FinishReason,
   usage: Schema.optional(Usage),
@@ -104,7 +186,7 @@ export const StepFinish = Schema.Struct({
 export type StepFinish = Schema.Schema.Type<typeof StepFinish>
 
 export const RequestFinish = Schema.Struct({
-  type: Schema.Literal("request-finish"),
+  type: Schema.tag("request-finish"),
   reason: FinishReason,
   usage: Schema.optional(Usage),
   providerMetadata: Schema.optional(ProviderMetadata),
@@ -112,7 +194,7 @@ export const RequestFinish = Schema.Struct({
 export type RequestFinish = Schema.Schema.Type<typeof RequestFinish>
 
 export const ProviderErrorEvent = Schema.Struct({
-  type: Schema.Literal("provider-error"),
+  type: Schema.tag("provider-error"),
   message: Schema.String,
   retryable: Schema.optional(Schema.Boolean),
   providerMetadata: Schema.optional(ProviderMetadata),
@@ -125,8 +207,12 @@ const llmEventTagged = Schema.Union([
   TextStart,
   TextDelta,
   TextEnd,
+  ReasoningStart,
   ReasoningDelta,
+  ReasoningEnd,
+  ToolInputStart,
   ToolInputDelta,
+  ToolInputEnd,
   ToolCall,
   ToolResult,
   ToolError,
@@ -135,20 +221,52 @@ const llmEventTagged = Schema.Union([
   ProviderErrorEvent,
 ]).pipe(Schema.toTaggedUnion("type"))
 
+type WithID<Event extends { readonly id: unknown }, ID> = Omit<Event, "type" | "id"> & { readonly id: ID | string }
+
+const responseID = (value: ResponseID | string) => ResponseID.make(value)
+const contentBlockID = (value: ContentBlockID | string) => ContentBlockID.make(value)
+const toolCallID = (value: ToolCallID | string) => ToolCallID.make(value)
+
 /**
  * camelCase aliases for `LLMEvent.guards` (provided by `Schema.toTaggedUnion`).
  * Lets consumers write `events.filter(LLMEvent.is.toolCall)` instead of
  * `events.filter(LLMEvent.guards["tool-call"])`.
  */
 export const LLMEvent = Object.assign(llmEventTagged, {
+  requestStart: (input: WithID<RequestStart, ResponseID>) => RequestStart.make({ ...input, id: responseID(input.id) }),
+  stepStart: StepStart.make,
+  textStart: (input: WithID<TextStart, ContentBlockID>) => TextStart.make({ ...input, id: contentBlockID(input.id) }),
+  textDelta: (input: WithID<TextDelta, ContentBlockID>) => TextDelta.make({ ...input, id: contentBlockID(input.id) }),
+  textEnd: (input: WithID<TextEnd, ContentBlockID>) => TextEnd.make({ ...input, id: contentBlockID(input.id) }),
+  reasoningStart: (input: WithID<ReasoningStart, ContentBlockID>) =>
+    ReasoningStart.make({ ...input, id: contentBlockID(input.id) }),
+  reasoningDelta: (input: WithID<ReasoningDelta, ContentBlockID>) =>
+    ReasoningDelta.make({ ...input, id: contentBlockID(input.id) }),
+  reasoningEnd: (input: WithID<ReasoningEnd, ContentBlockID>) =>
+    ReasoningEnd.make({ ...input, id: contentBlockID(input.id) }),
+  toolInputStart: (input: WithID<ToolInputStart, ToolCallID>) =>
+    ToolInputStart.make({ ...input, id: toolCallID(input.id) }),
+  toolInputDelta: (input: WithID<ToolInputDelta, ToolCallID>) =>
+    ToolInputDelta.make({ ...input, id: toolCallID(input.id) }),
+  toolInputEnd: (input: WithID<ToolInputEnd, ToolCallID>) => ToolInputEnd.make({ ...input, id: toolCallID(input.id) }),
+  toolCall: (input: WithID<ToolCall, ToolCallID>) => ToolCall.make({ ...input, id: toolCallID(input.id) }),
+  toolResult: (input: WithID<ToolResult, ToolCallID>) => ToolResult.make({ ...input, id: toolCallID(input.id) }),
+  toolError: (input: WithID<ToolError, ToolCallID>) => ToolError.make({ ...input, id: toolCallID(input.id) }),
+  stepFinish: StepFinish.make,
+  requestFinish: RequestFinish.make,
+  providerError: ProviderErrorEvent.make,
   is: {
     requestStart: llmEventTagged.guards["request-start"],
     stepStart: llmEventTagged.guards["step-start"],
     textStart: llmEventTagged.guards["text-start"],
     textDelta: llmEventTagged.guards["text-delta"],
     textEnd: llmEventTagged.guards["text-end"],
+    reasoningStart: llmEventTagged.guards["reasoning-start"],
     reasoningDelta: llmEventTagged.guards["reasoning-delta"],
+    reasoningEnd: llmEventTagged.guards["reasoning-end"],
+    toolInputStart: llmEventTagged.guards["tool-input-start"],
     toolInputDelta: llmEventTagged.guards["tool-input-delta"],
+    toolInputEnd: llmEventTagged.guards["tool-input-end"],
     toolCall: llmEventTagged.guards["tool-call"],
     toolResult: llmEventTagged.guards["tool-result"],
     toolError: llmEventTagged.guards["tool-error"],

packages/llm/src/schema/ids.ts

@@ -14,6 +14,15 @@ export type ModelID = typeof ModelID.Type
 export const ProviderID = Schema.String.pipe(Schema.brand("LLM.ProviderID"))
 export type ProviderID = typeof ProviderID.Type
 
+export const ResponseID = Schema.String
+export type ResponseID = Schema.Schema.Type<typeof ResponseID>
+
+export const ContentBlockID = Schema.String
+export type ContentBlockID = Schema.Schema.Type<typeof ContentBlockID>
+
+export const ToolCallID = Schema.String
+export type ToolCallID = Schema.Schema.Type<typeof ToolCallID>
+
 export const ReasoningEfforts = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] as const
 export const ReasoningEffort = Schema.Literals(ReasoningEfforts)
 export type ReasoningEffort = Schema.Schema.Type<typeof ReasoningEffort>

packages/llm/src/schema/messages.ts

@@ -1,6 +1,6 @@
 import { Schema } from "effect"
 import { JsonSchema, MessageRole, ProviderMetadata } from "./ids"
-import { CacheHint, GenerationOptions, HttpOptions, ModelRef, ProviderOptions } from "./options"
+import { CacheHint, CachePolicy, GenerationOptions, HttpOptions, ModelRef, ProviderOptions } from "./options"
 
 const isRecord = (value: unknown): value is Record<string, unknown> =>
   typeof value === "object" && value !== null && !Array.isArray(value)
@@ -79,6 +79,7 @@ export const ToolResultPart = Object.assign(
     name: Schema.String,
     result: ToolResultValue,
     providerExecuted: Schema.optional(Schema.Boolean),
+    cache: Schema.optional(CacheHint),
     metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
     providerMetadata: Schema.optional(ProviderMetadata),
   }).annotate({ identifier: "LLM.Content.ToolResult" }),
@@ -94,6 +95,7 @@ export const ToolResultPart = Object.assign(
       name: input.name,
       result: ToolResultValue.make(input.result, input.resultType),
       providerExecuted: input.providerExecuted,
+      cache: input.cache,
       metadata: input.metadata,
       providerMetadata: input.providerMetadata,
     }),
@@ -151,6 +153,7 @@ export class ToolDefinition extends Schema.Class<ToolDefinition>("LLM.ToolDefini
   name: Schema.String,
   description: Schema.String,
   inputSchema: JsonSchema,
+  cache: Schema.optional(CacheHint),
   metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
   native: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
 }) {}
@@ -203,6 +206,7 @@ export class LLMRequest extends Schema.Class<LLMRequest>("LLM.Request")({
   providerOptions: Schema.optional(ProviderOptions),
   http: Schema.optional(HttpOptions),
   responseFormat: Schema.optional(ResponseFormat),
+  cache: Schema.optional(CachePolicy),
   metadata: Schema.optional(Schema.Record(Schema.String, Schema.Unknown)),
 }) {}
 
@@ -220,6 +224,7 @@ export namespace LLMRequest {
     providerOptions: request.providerOptions,
     http: request.http,
     responseFormat: request.responseFormat,
+    cache: request.cache,
     metadata: request.metadata,
   })
 

packages/llm/src/schema/options.ts

@@ -200,3 +200,31 @@ export class CacheHint extends Schema.Class<CacheHint>("LLM.CacheHint")({
   type: Schema.Literals(["ephemeral", "persistent"]),
   ttlSeconds: Schema.optional(Schema.Number),
 }) {}
+
+// Auto-placement policy for prompt caching. The protocol-neutral lowering step
+// reads this and injects `CacheHint`s at the configured boundaries; the
+// per-protocol body builders then translate those hints into wire markers as
+// usual. `"auto"` is the recommended default for agent loops — it places one
+// breakpoint at the last tool definition, one at the last system part, and one
+// at the latest user message. The combination of provider invalidation
+// hierarchy (tools → system → messages) and Anthropic/Bedrock's 20-block
+// lookback means three trailing breakpoints reliably cover the static prefix.
+//
+// Pass `"none"` to opt out entirely (the legacy behavior). Pass the granular
+// object form to override individual choices.
+export const CachePolicyObject = Schema.Struct({
+  tools: Schema.optional(Schema.Boolean),
+  system: Schema.optional(Schema.Boolean),
+  messages: Schema.optional(
+    Schema.Union([
+      Schema.Literal("latest-user-message"),
+      Schema.Literal("latest-assistant"),
+      Schema.Struct({ tail: Schema.Number }),
+    ]),
+  ),
+  ttlSeconds: Schema.optional(Schema.Number),
+})
+export type CachePolicyObject = Schema.Schema.Type<typeof CachePolicyObject>
+
+export const CachePolicy = Schema.Union([Schema.Literal("auto"), Schema.Literal("none"), CachePolicyObject])
+export type CachePolicy = Schema.Schema.Type<typeof CachePolicy>

packages/llm/src/tool-runtime.ts

@@ -4,7 +4,7 @@ import {
   type ContentPart,
   type FinishReason,
   type LLMError,
-  type LLMEvent,
+  LLMEvent,
   LLMRequest,
   Message,
   type ProviderMetadata,
@@ -115,11 +115,19 @@ interface StepState {
 
 const accumulate = (state: StepState, event: LLMEvent) => {
   if (event.type === "text-delta") {
-    appendStreamingText(state, "text", event.text, event.providerMetadata)
+    appendStreamingText(state, "text", event.text, undefined)
     return
   }
   if (event.type === "reasoning-delta") {
-    appendStreamingText(state, "reasoning", event.text, event.providerMetadata)
+    appendStreamingText(state, "reasoning", event.text, undefined)
+    return
+  }
+  if (event.type === "reasoning-end") {
+    appendStreamingText(state, "reasoning", "", event.providerMetadata)
+    return
+  }
+  if (event.type === "text-end") {
+    appendStreamingText(state, "text", "", event.providerMetadata)
     return
   }
   if (event.type === "tool-call") {
@@ -219,10 +227,10 @@ const decodeAndExecute = (tool: AnyTool, input: unknown): Effect.Effect<ToolResu
 const emitEvents = (call: ToolCallPart, result: ToolResultValue): ReadonlyArray<LLMEvent> =>
   result.type === "error"
     ? [
-        { type: "tool-error", id: call.id, name: call.name, message: String(result.value) },
-        { type: "tool-result", id: call.id, name: call.name, result },
+        LLMEvent.toolError({ id: call.id, name: call.name, message: String(result.value) }),
+        LLMEvent.toolResult({ id: call.id, name: call.name, result }),
       ]
-    : [{ type: "tool-result", id: call.id, name: call.name, result }]
+    : [LLMEvent.toolResult({ id: call.id, name: call.name, result })]
 
 const followUpRequest = (
   request: LLMRequest,

packages/llm/test/adapter.test.ts

@@ -50,7 +50,9 @@ const request = LLM.request({
 })
 
 const raiseEvent = (event: FakeEvent): import("../src/schema").LLMEvent =>
-  event.type === "finish" ? { type: "request-finish", reason: event.reason } : { type: "text-delta", text: event.text }
+  event.type === "finish"
+    ? { type: "request-finish", reason: event.reason }
+    : { type: "text-delta", id: "text-0", text: event.text }
 
 const fakeProtocol = Protocol.make<FakeBody, FakeEvent, FakeEvent, void>({
   id: "fake",

packages/llm/test/cache-policy.test.ts

@@ -0,0 +1,262 @@
+import { describe, expect, test } from "bun:test"
+import { Effect } from "effect"
+import { CacheHint, LLM } from "../src"
+import { LLMClient } from "../src/route"
+import * as AnthropicMessages from "../src/protocols/anthropic-messages"
+import * as BedrockConverse from "../src/protocols/bedrock-converse"
+import * as Gemini from "../src/protocols/gemini"
+import * as OpenAIChat from "../src/protocols/openai-chat"
+import { applyCachePolicy } from "../src/cache-policy"
+import { it } from "./lib/effect"
+
+const anthropicModel = AnthropicMessages.model({
+  id: "claude-sonnet-4-5",
+  baseURL: "https://api.anthropic.test/v1/",
+  headers: { "x-api-key": "test" },
+})
+
+const bedrockModel = BedrockConverse.model({
+  id: "anthropic.claude-3-5-sonnet-20241022-v2:0",
+  credentials: { region: "us-east-1", accessKeyId: "fixture", secretAccessKey: "fixture" },
+})
+
+const openaiModel = OpenAIChat.model({
+  id: "gpt-4o-mini",
+  baseURL: "https://api.openai.test/v1/",
+  headers: { authorization: "Bearer test" },
+})
+
+const geminiModel = Gemini.model({
+  id: "gemini-2.5-flash",
+  baseURL: "https://generativelanguage.test/v1beta/",
+  headers: { "x-goog-api-key": "test" },
+})
+
+describe("applyCachePolicy", () => {
+  it.effect("undefined cache resolves to 'auto' (the recommended default)", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          system: "You are concise.",
+          prompt: "hi",
+        }),
+      )
+
+      // No explicit cache field → auto policy fires → last system part + latest
+      // user message both get cache_control markers.
+      expect(prepared.body).toMatchObject({
+        system: [{ type: "text", text: "You are concise.", cache_control: { type: "ephemeral" } }],
+        messages: [{ role: "user", content: [{ type: "text", text: "hi", cache_control: { type: "ephemeral" } }] }],
+      })
+    }),
+  )
+
+  it.effect("'auto' marks the last tool, last system part, and latest user message on Anthropic", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          system: "Sys A",
+          tools: [{ name: "t1", description: "t1", inputSchema: { type: "object", properties: {} } }],
+          messages: [LLM.user("first user"), LLM.assistant("assistant reply"), LLM.user("latest user message")],
+          cache: "auto",
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        tools: [{ name: "t1", cache_control: { type: "ephemeral" } }],
+        system: [{ type: "text", text: "Sys A", cache_control: { type: "ephemeral" } }],
+        messages: [
+          { role: "user", content: [{ type: "text", text: "first user" }] },
+          { role: "assistant", content: [{ type: "text", text: "assistant reply" }] },
+          {
+            role: "user",
+            content: [{ type: "text", text: "latest user message", cache_control: { type: "ephemeral" } }],
+          },
+        ],
+      })
+    }),
+  )
+
+  it.effect("'auto' is a no-op on OpenAI (implicit caching protocol)", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: openaiModel,
+          system: "Sys",
+          prompt: "hi",
+          cache: "auto",
+        }),
+      )
+
+      const body = prepared.body as { messages: Array<{ content: unknown }> }
+      // OpenAI doesn't accept cache_control on messages — policy must skip.
+      const flat = JSON.stringify(body)
+      expect(flat).not.toContain("cache_control")
+      expect(flat).not.toContain("cachePoint")
+    }),
+  )
+
+  it.effect("'auto' is a no-op on Gemini (out-of-band caching protocol)", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: geminiModel,
+          system: "Sys",
+          prompt: "hi",
+          cache: "auto",
+        }),
+      )
+
+      const flat = JSON.stringify(prepared.body)
+      expect(flat).not.toContain("cache_control")
+      expect(flat).not.toContain("cachePoint")
+    }),
+  )
+
+  it.effect("'auto' on Bedrock emits cachePoint markers in the right places", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: bedrockModel,
+          system: "Sys",
+          tools: [{ name: "t1", description: "t1", inputSchema: { type: "object", properties: {} } }],
+          messages: [LLM.user("first user"), LLM.assistant("reply"), LLM.user("latest user")],
+          cache: "auto",
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        toolConfig: {
+          tools: [{ toolSpec: { name: "t1" } }, { cachePoint: { type: "default" } }],
+        },
+        system: [{ text: "Sys" }, { cachePoint: { type: "default" } }],
+        messages: [
+          { role: "user", content: [{ text: "first user" }] },
+          { role: "assistant", content: [{ text: "reply" }] },
+          { role: "user", content: [{ text: "latest user" }, { cachePoint: { type: "default" } }] },
+        ],
+      })
+    }),
+  )
+
+  it.effect("'none' disables auto placement even when manual hints exist", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          system: "Sys",
+          tools: [{ name: "t1", description: "t1", inputSchema: { type: "object", properties: {} } }],
+          prompt: "hi",
+          cache: "none",
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        tools: [{ name: "t1", cache_control: undefined }],
+        system: [{ type: "text", text: "Sys", cache_control: undefined }],
+      })
+    }),
+  )
+
+  it.effect("granular object form: tools-only marks just tools", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          system: "Sys",
+          tools: [{ name: "t1", description: "t1", inputSchema: { type: "object", properties: {} } }],
+          prompt: "hi",
+          cache: { tools: true },
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        tools: [{ name: "t1", cache_control: { type: "ephemeral" } }],
+        system: [{ type: "text", text: "Sys", cache_control: undefined }],
+      })
+    }),
+  )
+
+  it.effect("auto policy preserves manual CacheHints on other parts", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          system: [
+            { type: "text", text: "first system", cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3600 }) },
+            { type: "text", text: "last system" },
+          ],
+          prompt: "hi",
+          cache: "auto",
+        }),
+      )
+
+      const body = prepared.body as { system: Array<{ text: string; cache_control?: unknown }> }
+      expect(body.system[0]?.cache_control).toEqual({ type: "ephemeral", ttl: "1h" })
+      expect(body.system[1]?.cache_control).toEqual({ type: "ephemeral" })
+    }),
+  )
+
+  it.effect("ttlSeconds in the policy flows through to wire markers", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          system: "Sys",
+          prompt: "hi",
+          cache: { system: true, ttlSeconds: 3600 },
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        system: [{ type: "text", text: "Sys", cache_control: { type: "ephemeral", ttl: "1h" } }],
+      })
+    }),
+  )
+
+  it.effect("messages: { tail: 2 } marks the last 2 message boundaries", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          messages: [LLM.user("u1"), LLM.assistant("a1"), LLM.user("u2"), LLM.assistant("a2")],
+          cache: { messages: { tail: 2 } },
+        }),
+      )
+
+      const body = prepared.body as { messages: Array<{ content: Array<{ cache_control?: unknown }> }> }
+      expect(body.messages[0]?.content[0]?.cache_control).toBeUndefined()
+      expect(body.messages[1]?.content[0]?.cache_control).toBeUndefined()
+      expect(body.messages[2]?.content[0]?.cache_control).toEqual({ type: "ephemeral" })
+      expect(body.messages[3]?.content[0]?.cache_control).toEqual({ type: "ephemeral" })
+    }),
+  )
+
+  it.effect("'latest-assistant' marks the last assistant message", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model: anthropicModel,
+          messages: [LLM.user("u1"), LLM.assistant("a1"), LLM.user("u2")],
+          cache: { messages: "latest-assistant" },
+        }),
+      )
+
+      const body = prepared.body as { messages: Array<{ content: Array<{ cache_control?: unknown }> }> }
+      expect(body.messages[0]?.content[0]?.cache_control).toBeUndefined()
+      expect(body.messages[1]?.content[0]?.cache_control).toEqual({ type: "ephemeral" })
+      expect(body.messages[2]?.content[0]?.cache_control).toBeUndefined()
+    }),
+  )
+
+  test("returns the same request reference when policy is a no-op (pure function)", () => {
+    const request = LLM.request({
+      model: anthropicModel,
+      prompt: "hi",
+      cache: "none",
+    })
+    expect(applyCachePolicy(request)).toBe(request)
+  })
+})

packages/llm/test/llm.test.ts

@@ -126,7 +126,7 @@ describe("llm constructors", () => {
     expect(
       LLMResponse.text({
         events: [
-          { type: "text-delta", text: "hi" },
+          { type: "text-delta", id: "text-0", text: "hi" },
           { type: "request-finish", reason: "stop" },
         ],
       }),

packages/llm/test/provider/anthropic-messages-cache.recorded.test.ts

@@ -0,0 +1,55 @@
+import { Redactor } from "@opencode-ai/http-recorder"
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { CacheHint, LLM } from "../../src"
+import { LLMClient } from "../../src/route"
+import * as AnthropicMessages from "../../src/protocols/anthropic-messages"
+import { LARGE_CACHEABLE_SYSTEM } from "../recorded-scenarios"
+import { recordedTests } from "../recorded-test"
+
+const model = AnthropicMessages.model({
+  id: "claude-haiku-4-5-20251001",
+  apiKey: process.env.ANTHROPIC_API_KEY ?? "fixture",
+})
+
+// Two identical generations in a row. The first call writes the prefix into
+// Anthropic's cache; the second should report a cache read against the same
+// prefix. Cassette captures both interactions in order.
+const cacheRequest = LLM.request({
+  id: "recorded_anthropic_cache",
+  model,
+  system: [{ type: "text", text: LARGE_CACHEABLE_SYSTEM, cache: new CacheHint({ type: "ephemeral" }) }],
+  prompt: "Say hi.",
+  // Manual hint on the system part is the only marker we want here — skip the
+  // auto-policy's latest-user-message breakpoint so the cassette body matches.
+  cache: "none",
+  generation: { maxTokens: 16, temperature: 0 },
+})
+
+const recorded = recordedTests({
+  prefix: "anthropic-messages-cache",
+  provider: "anthropic",
+  protocol: "anthropic-messages",
+  requires: ["ANTHROPIC_API_KEY"],
+  // Two identical requests in one cassette — replay walks the cassette in
+  // recording order so the second call replays the cached-hit interaction.
+  options: {
+    redactor: Redactor.defaults({ requestHeaders: { allow: ["content-type", "anthropic-version"] } }),
+  },
+})
+
+describe("Anthropic Messages cache recorded", () => {
+  recorded.effect.with("writes then reads cache_control on identical second call", { tags: ["cache"] }, () =>
+    Effect.gen(function* () {
+      const first = yield* LLMClient.generate(cacheRequest)
+      // The first call may write the cache (cacheWriteInputTokens > 0) or it
+      // may be a fresh miss (both fields 0) depending on whether the prefix is
+      // already warm on Anthropic's side. The assertion that matters is that
+      // the SECOND call reports a non-zero cache read.
+      expect(first.usage?.cacheReadInputTokens ?? 0).toBeGreaterThanOrEqual(0)
+
+      const second = yield* LLMClient.generate(cacheRequest)
+      expect(second.usage?.cacheReadInputTokens ?? 0).toBeGreaterThan(0)
+    }),
+  )
+})

packages/llm/test/provider/anthropic-messages.test.ts

@@ -1,6 +1,6 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
-import { CacheHint, LLM, LLMError } from "../../src"
+import { CacheHint, LLM, LLMError, Usage } from "../../src"
 import { LLMClient } from "../../src/route"
 import * as AnthropicMessages from "../../src/protocols/anthropic-messages"
 import { it } from "../lib/effect"
@@ -18,6 +18,9 @@ const request = LLM.request({
   model,
   system: { type: "text", text: "You are concise.", cache: new CacheHint({ type: "ephemeral" }) },
   prompt: "Say hello.",
+  // This fixture predates the `cache: "auto"` default; pin the policy off so
+  // existing wire-shape assertions only see the manual hint on the system part.
+  cache: "none",
   generation: { maxTokens: 20, temperature: 0 },
 })
 
@@ -48,6 +51,7 @@ describe("Anthropic Messages route", () => {
             LLM.assistant([LLM.toolCall({ id: "call_1", name: "lookup", input: { query: "weather" } })]),
             LLM.toolMessage({ id: "call_1", name: "lookup", result: { forecast: "sunny" } }),
           ],
+          cache: "none",
         }),
       )
 
@@ -110,12 +114,13 @@ describe("Anthropic Messages route", () => {
       expect(response.text).toBe("Hello!")
       expect(response.reasoning).toBe("thinking")
       expect(response.usage).toMatchObject({
-        inputTokens: 5,
+        inputTokens: 6,
         outputTokens: 2,
+        nonCachedInputTokens: 5,
         cacheReadInputTokens: 1,
-        totalTokens: 7,
+        totalTokens: 8,
       })
-      expect(response.events.find((event) => event.type === "reasoning-delta" && event.text === "")).toMatchObject({
+      expect(response.events.find((event) => event.type === "reasoning-end")).toMatchObject({
         providerMetadata: { anthropic: { signature: "sig_1" } },
       })
       expect(response.events.at(-1)).toMatchObject({
@@ -152,7 +157,13 @@ describe("Anthropic Messages route", () => {
         {
           type: "request-finish",
           reason: "tool-calls",
-          usage: { inputTokens: 5, outputTokens: 1, totalTokens: 6, native: { input_tokens: 5, output_tokens: 1 } },
+          usage: new Usage({
+            inputTokens: 5,
+            outputTokens: 1,
+            nonCachedInputTokens: 5,
+            totalTokens: 6,
+            providerMetadata: { anthropic: { input_tokens: 5, output_tokens: 1 } },
+          }),
         },
       ])
     }),
@@ -374,4 +385,134 @@ describe("Anthropic Messages route", () => {
       expect(error.message).toContain("Anthropic Messages user messages only support text content for now")
     }),
   )
+
+  it.effect("maps ttlSeconds >= 3600 to cache_control ttl: '1h'", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          system: { type: "text", text: "system", cache: new CacheHint({ type: "ephemeral", ttlSeconds: 3600 }) },
+          prompt: "hi",
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        system: [{ type: "text", text: "system", cache_control: { type: "ephemeral", ttl: "1h" } }],
+      })
+    }),
+  )
+
+  it.effect("emits cache_control on tool definitions and tool-result blocks", () =>
+    Effect.gen(function* () {
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          tools: [
+            {
+              name: "lookup",
+              description: "lookup tool",
+              inputSchema: { type: "object", properties: {} },
+              cache: new CacheHint({ type: "ephemeral" }),
+            },
+          ],
+          messages: [
+            LLM.user("What's the weather?"),
+            LLM.assistant([LLM.toolCall({ id: "call_1", name: "lookup", input: {} })]),
+            LLM.toolMessage({
+              id: "call_1",
+              name: "lookup",
+              result: { temp: 72 },
+              cache: new CacheHint({ type: "ephemeral" }),
+            }),
+          ],
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        tools: [{ name: "lookup", cache_control: { type: "ephemeral" } }],
+        messages: [
+          { role: "user", content: [{ type: "text", text: "What's the weather?" }] },
+          { role: "assistant", content: [{ type: "tool_use", id: "call_1", name: "lookup" }] },
+          {
+            role: "user",
+            content: [{ type: "tool_result", tool_use_id: "call_1", cache_control: { type: "ephemeral" } }],
+          },
+        ],
+      })
+    }),
+  )
+
+  it.effect("drops cache_control breakpoints past the 4-per-request cap", () =>
+    Effect.gen(function* () {
+      const hint = new CacheHint({ type: "ephemeral" })
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          system: [
+            { type: "text", text: "a", cache: hint },
+            { type: "text", text: "b", cache: hint },
+            { type: "text", text: "c", cache: hint },
+            { type: "text", text: "d", cache: hint },
+            { type: "text", text: "e", cache: hint },
+            { type: "text", text: "f", cache: hint },
+          ],
+          prompt: "hi",
+        }),
+      )
+
+      const system = (prepared.body as { system: Array<{ cache_control?: unknown }> }).system
+      const marked = system.filter((part) => part.cache_control !== undefined)
+      expect(marked).toHaveLength(4)
+      expect(system[4]?.cache_control).toBeUndefined()
+      expect(system[5]?.cache_control).toBeUndefined()
+    }),
+  )
+
+  it.effect("spends breakpoint budget on tools before system before messages", () =>
+    Effect.gen(function* () {
+      const hint = new CacheHint({ type: "ephemeral" })
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          tools: [
+            {
+              name: "t1",
+              description: "t1",
+              inputSchema: { type: "object", properties: {} },
+              cache: hint,
+            },
+            {
+              name: "t2",
+              description: "t2",
+              inputSchema: { type: "object", properties: {} },
+              cache: hint,
+            },
+            {
+              name: "t3",
+              description: "t3",
+              inputSchema: { type: "object", properties: {} },
+              cache: hint,
+            },
+            {
+              name: "t4",
+              description: "t4",
+              inputSchema: { type: "object", properties: {} },
+              cache: hint,
+            },
+          ],
+          system: [{ type: "text", text: "system-tail", cache: hint }],
+          messages: [LLM.user([{ type: "text", text: "message-tail", cache: hint }])],
+        }),
+      )
+
+      const body = prepared.body as {
+        tools: Array<{ cache_control?: unknown }>
+        system: Array<{ cache_control?: unknown }>
+        messages: Array<{ content: Array<{ cache_control?: unknown }> }>
+      }
+      expect(body.tools.every((t) => t.cache_control !== undefined)).toBe(true)
+      expect(body.system[0]?.cache_control).toBeUndefined()
+      expect(body.messages[0]?.content[0]?.cache_control).toBeUndefined()
+    }),
+  )
 })

packages/llm/test/provider/bedrock-converse-cache.recorded.test.ts

@@ -0,0 +1,55 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { CacheHint, LLM } from "../../src"
+import { LLMClient } from "../../src/route"
+import * as BedrockConverse from "../../src/protocols/bedrock-converse"
+import { LARGE_CACHEABLE_SYSTEM } from "../recorded-scenarios"
+import { recordedTests } from "../recorded-test"
+
+const RECORDING_REGION = process.env.BEDROCK_RECORDING_REGION ?? "us-east-1"
+
+// Use a Claude model on Bedrock — Nova has automatic prefix caching that
+// doesn't reliably surface `cacheRead`/`cacheWrite` in usage, so the second
+// call wouldn't deterministically prove cache mapping works. Override with
+// BEDROCK_CACHE_MODEL_ID if your account has access elsewhere.
+const model = BedrockConverse.model({
+  id: process.env.BEDROCK_CACHE_MODEL_ID ?? "us.anthropic.claude-haiku-4-5-20251001-v1:0",
+  credentials: {
+    region: RECORDING_REGION,
+    accessKeyId: process.env.AWS_ACCESS_KEY_ID ?? "fixture",
+    secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY ?? "fixture",
+    sessionToken: process.env.AWS_SESSION_TOKEN,
+  },
+})
+
+const cacheRequest = LLM.request({
+  id: "recorded_bedrock_cache",
+  model,
+  system: [{ type: "text", text: LARGE_CACHEABLE_SYSTEM, cache: new CacheHint({ type: "ephemeral" }) }],
+  prompt: "Say hi.",
+  // Manual hint on the system part is the only marker we want here — skip the
+  // auto-policy's latest-user-message breakpoint so the cassette body matches.
+  cache: "none",
+  generation: { maxTokens: 16, temperature: 0 },
+})
+
+const recorded = recordedTests({
+  prefix: "bedrock-converse-cache",
+  provider: "amazon-bedrock",
+  protocol: "bedrock-converse",
+  requires: ["AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY"],
+  // Two identical requests in one cassette — replay walks the cassette in
+  // recording order so the second call replays the cached-hit interaction.
+})
+
+describe("Bedrock Converse cache recorded", () => {
+  recorded.effect.with("writes then reads cachePoint on identical second call", { tags: ["cache"] }, () =>
+    Effect.gen(function* () {
+      const first = yield* LLMClient.generate(cacheRequest)
+      expect(first.usage?.cacheReadInputTokens ?? 0).toBeGreaterThanOrEqual(0)
+
+      const second = yield* LLMClient.generate(cacheRequest)
+      expect(second.usage?.cacheReadInputTokens ?? 0).toBeGreaterThan(0)
+    }),
+  )
+})

packages/llm/test/provider/bedrock-converse.test.ts

@@ -63,6 +63,9 @@ const baseRequest = LLM.request({
   model,
   system: "You are concise.",
   prompt: "Say hello.",
+  // Wire-shape assertions in this file predate the `cache: "auto"` default;
+  // pin the policy off so they only exercise the lowering path itself.
+  cache: "none",
   generation: { maxTokens: 64, temperature: 0 },
 })
 
@@ -125,6 +128,7 @@ describe("Bedrock Converse route", () => {
             LLM.assistant([LLM.toolCall({ id: "tool_1", name: "lookup", input: { query: "weather" } })]),
             LLM.toolMessage({ id: "tool_1", name: "lookup", result: { forecast: "sunny" } }),
           ],
+          cache: "none",
         }),
       )
 
@@ -339,6 +343,7 @@ describe("Bedrock Converse route", () => {
               { type: "media", mediaType: "image/webp", data: "DDDD" },
             ]),
           ],
+          cache: "none",
         }),
       )
 
@@ -440,6 +445,78 @@ describe("Bedrock Converse route", () => {
       expect(error.message).toContain("Bedrock Converse does not support media type application/x-tar")
     }),
   )
+
+  it.effect("maps ttlSeconds >= 3600 to cachePoint ttl: '1h'", () =>
+    Effect.gen(function* () {
+      const cache = new CacheHint({ type: "ephemeral", ttlSeconds: 3600 })
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          system: [{ type: "text", text: "system", cache }],
+          prompt: "hi",
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        system: [{ text: "system" }, { cachePoint: { type: "default", ttl: "1h" } }],
+      })
+    }),
+  )
+
+  it.effect("appends cachePoint after marked tool definitions and tool-result blocks", () =>
+    Effect.gen(function* () {
+      const cache = new CacheHint({ type: "ephemeral" })
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          tools: [{ name: "lookup", description: "lookup", inputSchema: { type: "object", properties: {} }, cache }],
+          messages: [
+            LLM.user("What's the weather?"),
+            LLM.assistant([LLM.toolCall({ id: "call_1", name: "lookup", input: {} })]),
+            LLM.toolMessage({ id: "call_1", name: "lookup", result: { temp: 72 }, cache }),
+          ],
+          cache: "none",
+        }),
+      )
+
+      expect(prepared.body).toMatchObject({
+        toolConfig: {
+          tools: [{ toolSpec: { name: "lookup" } }, { cachePoint: { type: "default" } }],
+        },
+        messages: [
+          { role: "user", content: [{ text: "What's the weather?" }] },
+          { role: "assistant", content: [{ toolUse: { toolUseId: "call_1" } }] },
+          {
+            role: "user",
+            content: [{ toolResult: { toolUseId: "call_1" } }, { cachePoint: { type: "default" } }],
+          },
+        ],
+      })
+    }),
+  )
+
+  it.effect("drops cachePoint markers past the 4-per-request cap", () =>
+    Effect.gen(function* () {
+      const cache = new CacheHint({ type: "ephemeral" })
+      const prepared = yield* LLMClient.prepare(
+        LLM.request({
+          model,
+          system: [
+            { type: "text", text: "a", cache },
+            { type: "text", text: "b", cache },
+            { type: "text", text: "c", cache },
+            { type: "text", text: "d", cache },
+            { type: "text", text: "e", cache },
+            { type: "text", text: "f", cache },
+          ],
+          prompt: "hi",
+        }),
+      )
+
+      const system = (prepared.body as { system: Array<{ cachePoint?: unknown }> }).system
+      expect(system.filter((part) => "cachePoint" in part)).toHaveLength(4)
+    }),
+  )
 })
 
 // Live recorded integration tests. Run with `RECORD=true AWS_ACCESS_KEY_ID=...
@@ -484,6 +561,7 @@ describe("Bedrock Converse recorded", () => {
           model: recordedModel(),
           system: "Reply with the single word 'Hello'.",
           prompt: "Say hello.",
+          cache: "none",
           generation: { maxTokens: 16, temperature: 0 },
         }),
       )
@@ -506,6 +584,7 @@ describe("Bedrock Converse recorded", () => {
           prompt: "Call get_weather with city exactly Paris.",
           tools: [weatherTool],
           toolChoice: LLM.toolChoice(weatherTool),
+          cache: "none",
           generation: { maxTokens: 80, temperature: 0 },
         }),
       )

packages/llm/test/provider/gemini-cache.recorded.test.ts

@@ -0,0 +1,49 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { LLM } from "../../src"
+import { LLMClient } from "../../src/route"
+import * as Gemini from "../../src/protocols/gemini"
+import { LARGE_CACHEABLE_SYSTEM } from "../recorded-scenarios"
+import { recordedTests } from "../recorded-test"
+
+const model = Gemini.model({
+  id: "gemini-2.5-flash",
+  apiKey: process.env.GOOGLE_GENERATIVE_AI_API_KEY ?? process.env.GEMINI_API_KEY ?? "fixture",
+})
+
+// Gemini does implicit prefix caching on 2.5+ models above ~1024 tokens. The
+// `CacheHint` is currently a no-op for Gemini (the explicit `CachedContent`
+// API is out-of-band and intentionally not wired up). This test exists to
+// pin the usage-parsing path: `cachedContentTokenCount` should surface as
+// `cacheReadInputTokens` on the second identical call.
+const cacheRequest = LLM.request({
+  id: "recorded_gemini_cache",
+  model,
+  system: LARGE_CACHEABLE_SYSTEM,
+  prompt: "Say hi.",
+  generation: { maxTokens: 16, temperature: 0 },
+})
+
+const recorded = recordedTests({
+  prefix: "gemini-cache",
+  provider: "google",
+  protocol: "gemini",
+  requires: ["GOOGLE_GENERATIVE_AI_API_KEY"],
+  // Two identical requests in one cassette — replay walks the cassette in
+  // recording order so the second call replays the cached-hit interaction.
+})
+
+describe("Gemini cache recorded", () => {
+  recorded.effect.with("reports cachedContentTokenCount on identical second call", { tags: ["cache"] }, () =>
+    Effect.gen(function* () {
+      const first = yield* LLMClient.generate(cacheRequest)
+      expect(first.usage?.cacheReadInputTokens ?? 0).toBeGreaterThanOrEqual(0)
+
+      const second = yield* LLMClient.generate(cacheRequest)
+      // Implicit caching is best-effort on Gemini's side; we assert the field
+      // is at least populated and non-negative. When re-recording, verify the
+      // cassette shows > 0 in the second response's usage.
+      expect(second.usage?.cacheReadInputTokens ?? 0).toBeGreaterThanOrEqual(0)
+    }),
+  )
+})

packages/llm/test/provider/gemini.test.ts

@@ -1,6 +1,6 @@
 import { describe, expect } from "bun:test"
 import { Effect } from "effect"
-import { LLM, LLMError } from "../../src"
+import { LLM, LLMError, Usage } from "../../src"
 import { LLMClient } from "../../src/route"
 import * as Gemini from "../../src/protocols/gemini"
 import { it } from "../lib/effect"
@@ -198,32 +198,36 @@ describe("Gemini route", () => {
       expect(response.reasoning).toBe("thinking")
       expect(response.usage).toMatchObject({
         inputTokens: 5,
-        outputTokens: 2,
-        reasoningTokens: 1,
+        outputTokens: 3,
+        nonCachedInputTokens: 4,
         cacheReadInputTokens: 1,
+        reasoningTokens: 1,
         totalTokens: 7,
       })
       expect(response.events).toEqual([
-        { type: "reasoning-delta", text: "thinking" },
-        { type: "text-delta", text: "Hello" },
-        { type: "text-delta", text: "!" },
+        { type: "reasoning-delta", id: "reasoning-0", text: "thinking" },
+        { type: "text-delta", id: "text-0", text: "Hello" },
+        { type: "text-delta", id: "text-0", text: "!" },
         {
           type: "request-finish",
           reason: "stop",
-          usage: {
+          usage: new Usage({
             inputTokens: 5,
-            outputTokens: 2,
-            reasoningTokens: 1,
+            outputTokens: 3,
+            nonCachedInputTokens: 4,
             cacheReadInputTokens: 1,
+            reasoningTokens: 1,
             totalTokens: 7,
-            native: {
-              promptTokenCount: 5,
-              candidatesTokenCount: 2,
-              totalTokenCount: 7,
-              thoughtsTokenCount: 1,
-              cachedContentTokenCount: 1,
+            providerMetadata: {
+              google: {
+                promptTokenCount: 5,
+                candidatesTokenCount: 2,
+                totalTokenCount: 7,
+                thoughtsTokenCount: 1,
+                cachedContentTokenCount: 1,
+              },
             },
-          },
+          }),
         },
       ])
     }),
@@ -257,12 +261,13 @@ describe("Gemini route", () => {
         {
           type: "request-finish",
           reason: "tool-calls",
-          usage: {
+          usage: new Usage({
             inputTokens: 5,
             outputTokens: 1,
+            nonCachedInputTokens: 5,
             totalTokens: 6,
-            native: { promptTokenCount: 5, candidatesTokenCount: 1 },
-          },
+            providerMetadata: { google: { promptTokenCount: 5, candidatesTokenCount: 1 } },
+          }),
         },
       ])
     }),

packages/llm/test/provider/openai-chat.test.ts

@@ -1,7 +1,7 @@
 import { describe, expect } from "bun:test"
 import { Effect, Schema, Stream } from "effect"
 import { HttpClientRequest } from "effect/unstable/http"
-import { LLM, LLMError } from "../../src"
+import { LLM, LLMError, Usage } from "../../src"
 import * as Azure from "../../src/providers/azure"
 import * as OpenAI from "../../src/providers/openai"
 import * as OpenAIChat from "../../src/protocols/openai-chat"
@@ -225,25 +225,28 @@ describe("OpenAI Chat route", () => {
 
       expect(response.text).toBe("Hello!")
       expect(response.events).toEqual([
-        { type: "text-delta", text: "Hello" },
-        { type: "text-delta", text: "!" },
+        { type: "text-delta", id: "text-0", text: "Hello" },
+        { type: "text-delta", id: "text-0", text: "!" },
         {
           type: "request-finish",
           reason: "stop",
-          usage: {
+          usage: new Usage({
             inputTokens: 5,
             outputTokens: 2,
-            reasoningTokens: 0,
+            nonCachedInputTokens: 4,
             cacheReadInputTokens: 1,
+            reasoningTokens: 0,
             totalTokens: 7,
-            native: {
-              prompt_tokens: 5,
-              completion_tokens: 2,
-              total_tokens: 7,
-              prompt_tokens_details: { cached_tokens: 1 },
-              completion_tokens_details: { reasoning_tokens: 0 },
+            providerMetadata: {
+              openai: {
+                prompt_tokens: 5,
+                completion_tokens: 2,
+                total_tokens: 7,
+                prompt_tokens_details: { cached_tokens: 1 },
+                completion_tokens_details: { reasoning_tokens: 0 },
+              },
             },
-          },
+          }),
         },
       ])
     }),

packages/llm/test/provider/openai-responses-cache.recorded.test.ts

@@ -0,0 +1,47 @@
+import { describe, expect } from "bun:test"
+import { Effect } from "effect"
+import { LLM } from "../../src"
+import { LLMClient } from "../../src/route"
+import * as OpenAIResponses from "../../src/protocols/openai-responses"
+import { LARGE_CACHEABLE_SYSTEM } from "../recorded-scenarios"
+import { recordedTests } from "../recorded-test"
+
+const model = OpenAIResponses.model({
+  id: "gpt-4.1-mini",
+  apiKey: process.env.OPENAI_API_KEY ?? "fixture",
+})
+
+// OpenAI caches prefixes automatically once they cross the 1024-token threshold;
+// `CacheHint` is a no-op for the wire body. The stable signal is the
+// `prompt_cache_key` routing hint, which keeps repeated calls on the same shard
+// so cache hits are observable.
+const cacheRequest = LLM.request({
+  id: "recorded_openai_responses_cache",
+  model,
+  system: LARGE_CACHEABLE_SYSTEM,
+  prompt: "Say hi.",
+  generation: { maxTokens: 16, temperature: 0 },
+  providerOptions: { openai: { promptCacheKey: "recorded-cache-test" } },
+})
+
+const recorded = recordedTests({
+  prefix: "openai-responses-cache",
+  provider: "openai",
+  protocol: "openai-responses",
+  requires: ["OPENAI_API_KEY"],
+  // Two identical requests in one cassette — replay walks the cassette in
+  // recording order so the second call replays the cached-hit interaction,
+  // not the cold-miss one.
+})
+
+describe("OpenAI Responses cache recorded", () => {
+  recorded.effect.with("reports cached_tokens on identical second call", { tags: ["cache"] }, () =>
+    Effect.gen(function* () {
+      const first = yield* LLMClient.generate(cacheRequest)
+      expect(first.usage?.cacheReadInputTokens ?? 0).toBeGreaterThanOrEqual(0)
+
+      const second = yield* LLMClient.generate(cacheRequest)
+      expect(second.usage?.cacheReadInputTokens ?? 0).toBeGreaterThan(0)
+    }),
+  )
+})

packages/llm/test/provider/openai-responses.test.ts

@@ -1,7 +1,7 @@
 import { describe, expect } from "bun:test"
 import { ConfigProvider, Effect, Layer, Stream } from "effect"
 import { Headers, HttpClientRequest } from "effect/unstable/http"
-import { LLM, LLMError } from "../../src"
+import { LLM, LLMError, Usage } from "../../src"
 import { Auth, LLMClient, RequestExecutor, WebSocketExecutor } from "../../src/route"
 import * as Azure from "../../src/providers/azure"
 import * as OpenAI from "../../src/providers/openai"
@@ -336,26 +336,29 @@ describe("OpenAI Responses route", () => {
 
       expect(response.text).toBe("Hello!")
       expect(response.events).toEqual([
-        { type: "text-delta", id: "msg_1", text: "Hello", providerMetadata: { openai: { itemId: "msg_1" } } },
-        { type: "text-delta", id: "msg_1", text: "!", providerMetadata: { openai: { itemId: "msg_1" } } },
+        { type: "text-delta", id: "msg_1", text: "Hello" },
+        { type: "text-delta", id: "msg_1", text: "!" },
         {
           type: "request-finish",
           reason: "stop",
           providerMetadata: { openai: { responseId: "resp_1", serviceTier: "default" } },
-          usage: {
+          usage: new Usage({
             inputTokens: 5,
             outputTokens: 2,
-            reasoningTokens: 0,
+            nonCachedInputTokens: 4,
             cacheReadInputTokens: 1,
+            reasoningTokens: 0,
             totalTokens: 7,
-            native: {
-              input_tokens: 5,
-              output_tokens: 2,
-              total_tokens: 7,
-              input_tokens_details: { cached_tokens: 1 },
-              output_tokens_details: { reasoning_tokens: 0 },
+            providerMetadata: {
+              openai: {
+                input_tokens: 5,
+                output_tokens: 2,
+                total_tokens: 7,
+                input_tokens_details: { cached_tokens: 1 },
+                output_tokens_details: { reasoning_tokens: 0 },
+              },
             },
-          },
+          }),
         },
       ])
     }),
@@ -394,14 +397,12 @@ describe("OpenAI Responses route", () => {
           id: "call_1",
           name: "lookup",
           text: '{"query"',
-          providerMetadata: { openai: { itemId: "item_1" } },
         },
         {
           type: "tool-input-delta",
           id: "call_1",
           name: "lookup",
           text: ':"weather"}',
-          providerMetadata: { openai: { itemId: "item_1" } },
         },
         {
           type: "tool-call",
@@ -413,7 +414,13 @@ describe("OpenAI Responses route", () => {
         {
           type: "request-finish",
           reason: "tool-calls",
-          usage: { inputTokens: 5, outputTokens: 1, totalTokens: 6, native: { input_tokens: 5, output_tokens: 1 } },
+          usage: new Usage({
+            inputTokens: 5,
+            outputTokens: 1,
+            nonCachedInputTokens: 5,
+            totalTokens: 6,
+            providerMetadata: { openai: { input_tokens: 5, output_tokens: 1 } },
+          }),
         },
       ])
     }),

packages/llm/test/recorded-scenarios.ts

@@ -6,6 +6,18 @@ import { tool } from "../src/tool"
 
 export const weatherToolName = "get_weather"
 
+// A deterministic system prompt long enough to clear every supported provider's
+// minimum cacheable-prefix threshold (Anthropic Haiku 3.5: 2048 tokens; Anthropic
+// Opus/Haiku 4.5: 4096 tokens; OpenAI/Gemini/Bedrock: lower). Built by repeating
+// a fixed sentence — the cassette replays bit-for-bit, so the exact text matters
+// only when re-recording with `RECORD=true`.
+export const LARGE_CACHEABLE_SYSTEM = (() => {
+  const sentence = "You are a concise, factual assistant. Answer precisely and avoid filler. Cite numbers when known. "
+  // ~100 chars per sentence × 250 repeats ≈ 25,000 chars ≈ 5k+ tokens, safely
+  // above every provider's threshold.
+  return sentence.repeat(250)
+})()
+
 export const weatherTool = LLM.toolDefinition({
   name: weatherToolName,
   description: "Get current weather for a city.",
@@ -39,6 +51,7 @@ export const textRequest = (input: {
     model: input.model,
     system: "You are concise.",
     prompt: input.prompt ?? "Reply with exactly: Hello!",
+    cache: "none",
     generation:
       input.temperature === false
         ? { maxTokens: input.maxTokens ?? 20 }
@@ -58,6 +71,7 @@ export const weatherToolRequest = (input: {
     prompt: "Call get_weather with city exactly Paris.",
     tools: [weatherTool],
     toolChoice: LLM.toolChoice(weatherTool),
+    cache: "none",
     generation:
       input.temperature === false
         ? { maxTokens: input.maxTokens ?? 80 }
@@ -76,6 +90,7 @@ export const weatherToolLoopRequest = (input: {
     model: input.model,
     system: input.system ?? "Use the get_weather tool, then answer in one short sentence.",
     prompt: "What is the weather in Paris?",
+    cache: "none",
     generation:
       input.temperature === false
         ? { maxTokens: input.maxTokens ?? 80 }

packages/llm/test/schema.test.ts

@@ -1,6 +1,7 @@
 import { describe, expect, test } from "bun:test"
 import { Schema } from "effect"
-import { ContentPart, LLMEvent, LLMRequest, ModelID, ModelLimits, ModelRef, ProviderID } from "../src/schema"
+import { ContentPart, LLMEvent, LLMRequest, ModelID, ModelLimits, ModelRef, ProviderID, Usage } from "../src/schema"
+import { ProviderShared } from "../src/protocols/shared"
 
 const model = new ModelRef({
   id: ModelID.make("fake-model"),
@@ -48,3 +49,30 @@ describe("llm schema", () => {
     expect(ContentPart.guards.media({ type: "text", text: "hi" })).toBe(false)
   })
 })
+
+describe("LLM.Usage", () => {
+  test("subtractTokens clamps non-sensical breakdowns to zero", () => {
+    // Defense against a provider reporting cached_tokens > prompt_tokens or
+    // reasoning_tokens > completion_tokens — the negative would otherwise
+    // round-trip through the pipeline and crash strict downstream schemas.
+    expect(ProviderShared.subtractTokens(5, 3)).toBe(2)
+    expect(ProviderShared.subtractTokens(5, 10)).toBe(0)
+    expect(ProviderShared.subtractTokens(5, undefined)).toBe(5)
+    expect(ProviderShared.subtractTokens(undefined, 3)).toBeUndefined()
+    expect(ProviderShared.subtractTokens(undefined, undefined)).toBeUndefined()
+  })
+
+  test("sumTokens returns undefined only when every input is undefined", () => {
+    expect(ProviderShared.sumTokens(1, 2, 3)).toBe(6)
+    expect(ProviderShared.sumTokens(1, undefined, 3)).toBe(4)
+    expect(ProviderShared.sumTokens(undefined, undefined, undefined)).toBeUndefined()
+    expect(ProviderShared.sumTokens()).toBeUndefined()
+  })
+
+  test("visibleOutputTokens clamps reasoning > output to zero", () => {
+    expect(new Usage({ outputTokens: 10, reasoningTokens: 4 }).visibleOutputTokens).toBe(6)
+    expect(new Usage({ outputTokens: 10 }).visibleOutputTokens).toBe(10)
+    expect(new Usage({ outputTokens: 4, reasoningTokens: 10 }).visibleOutputTokens).toBe(0)
+    expect(new Usage({}).visibleOutputTokens).toBe(0)
+  })
+})

packages/opencode/AGENTS.md

@@ -9,6 +9,13 @@
 - **Output**: creates `migration/<timestamp>_<slug>/migration.sql` and `snapshot.json`.
 - **Tests**: migration tests should read the per-folder layout (no `_journal.json`).
 
+## Development server
+
+- Running `bun dev` from `packages/opencode` starts the live interactive TUI. Do not run it as a blocking foreground command when you need to inspect the result.
+- Start it in `tmux` instead: `tmux new-session -d -s opencode-dev 'bun dev'`.
+- Capture the current TUI output with: `tmux capture-pane -pt opencode-dev`.
+- Stop the session explicitly when done: `tmux kill-session -t opencode-dev`.
+
 # Module shape
 
 Do not use `export namespace Foo { ... }` for module organization. It is not

packages/opencode/migration/20260511000411_data_migration_state/migration.sql

@@ -0,0 +1,4 @@
+CREATE TABLE `data_migration` (
+	`name` text PRIMARY KEY,
+	`time_completed` integer NOT NULL
+);

packages/opencode/migration/20260511143053_drop_project_sandboxes/migration.sql

@@ -0,0 +1 @@
+ALTER TABLE `project` DROP COLUMN `sandboxes`;

packages/opencode/package.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
-  "version": "1.14.46",
+  "version": "1.14.48",
   "name": "opencode",
   "type": "module",
   "license": "MIT",

packages/opencode/parsers-config.ts

@@ -286,5 +286,15 @@ export default {
         ],
       },
     },
+    {
+      filetype: "diff",
+      aliases: ["udiff", "patch"],
+      wasm: "https://github.com/tree-sitter-grammars/tree-sitter-diff/releases/download/v0.1.0/tree-sitter-diff.wasm",
+      queries: {
+        highlights: [
+          "https://raw.githubusercontent.com/tree-sitter-grammars/tree-sitter-diff/master/queries/highlights.scm",
+        ],
+      },
+    },
   ],
 }

packages/opencode/src/agent/agent.ts

@@ -26,7 +26,6 @@ import * as Option from "effect/Option"
 import * as OtelTracer from "@effect/opentelemetry/Tracer"
 import { zod } from "@opencode-ai/core/effect-zod"
 import { withStatics, type DeepMutable } from "@opencode-ai/core/schema"
-import { Reference } from "@/reference/reference"
 
 export const Info = Schema.Struct({
   name: Schema.String,
@@ -301,76 +300,6 @@ export const layer = Layer.effect(
           item.permission = Permission.merge(item.permission, Permission.fromConfig(value.permission ?? {}))
         }
 
-        function referencePrompt(reference: Reference.Resolved) {
-          if (reference.kind === "local") {
-            return [
-              `You are configured reference @${reference.name}, a read-only research agent for external reference material.`,
-              `Local directory: ${reference.path}`,
-              `Inspect this directory as the primary reference source. Prefer repo_overview with path ${JSON.stringify(reference.path)} before broader searches. Do not edit files.`,
-              `Return exact absolute file paths for findings whenever possible.`,
-            ].join("\n\n")
-          }
-
-          if (reference.kind === "invalid") {
-            return [
-              `You are configured reference @${reference.name}, but this reference is not usable yet.`,
-              `Configured repository: ${reference.repository}`,
-              `Problem: ${reference.message}`,
-              `Explain this configuration problem if invoked. Do not edit files or attempt fallback clones.`,
-            ].join("\n\n")
-          }
-
-          return [
-            `You are configured reference @${reference.name}, a read-only research agent for external reference material.`,
-            `Repository: ${reference.repository}`,
-            ...(reference.branch ? [`Branch/ref: ${reference.branch}`] : []),
-            `Cached directory: ${reference.path}`,
-            `OpenCode materializes this configured repository before use. Do not call repo_clone for this reference.`,
-            `Inspect the cached directory as the primary reference source. Prefer repo_overview with path ${JSON.stringify(reference.path)} before broader searches, then use Glob, Grep, and Read inside that directory. Do not edit files.`,
-            `Return exact absolute file paths for findings whenever possible.`,
-          ].join("\n\n")
-        }
-
-        function referenceDescription(reference: Reference.Resolved) {
-          if (reference.kind === "local") return `Scout reference for local directory ${reference.path}`
-          if (reference.kind === "git") return `Scout reference for repository ${reference.repository}`
-          return `Invalid Scout reference for repository ${reference.repository}`
-        }
-
-        if (Flag.OPENCODE_EXPERIMENTAL_SCOUT) {
-          const resolvedReferences = Reference.resolveAll({
-            references: cfg.reference ?? {},
-            directory: ctx.directory,
-            worktree: ctx.worktree,
-          })
-          for (const resolved of resolvedReferences) {
-            if (agents[resolved.name]) continue
-            const localPath = resolved.kind === "invalid" ? undefined : resolved.path
-            agents[resolved.name] = {
-              name: resolved.name,
-              description: referenceDescription(resolved),
-              permission: Permission.merge(
-                agents.scout.permission,
-                Permission.fromConfig({
-                  repo_clone: "deny",
-                  ...(localPath
-                    ? {
-                        external_directory: {
-                          [localPath]: "allow",
-                          [path.join(localPath, "*")]: "allow",
-                        },
-                      }
-                    : {}),
-                }),
-              ),
-              prompt: referencePrompt(resolved),
-              options: { reference: cfg.reference?.[resolved.name], resolved },
-              mode: "subagent",
-              native: false,
-            }
-          }
-        }
-
         // Ensure Truncate.GLOB is allowed unless explicitly configured
         for (const name in agents) {
           const agent = agents[name]

packages/opencode/src/cli/cmd/tui/app.tsx

@@ -93,7 +93,6 @@ const appBindingCommands = [
   "theme.mode.lock",
   "help.show",
   "docs.open",
-  "app.exit",
   "app.debug",
   "app.console",
   "app.heap_snapshot",
@@ -648,11 +647,6 @@ function App(props: { onSnapshot?: () => Promise<string[]> }) {
         title: "Exit the app",
         slashName: "exit",
         slashAliases: ["quit", "q"],
-        enabled: () => {
-          const current = promptRef.current
-          if (!current?.focused) return true
-          return current.current.input === ""
-        },
         run: () => exit(),
         category: "System",
       },
@@ -785,6 +779,17 @@ function App(props: { onSnapshot?: () => Promise<string[]> }) {
     bindings: tuiConfig.keybinds.gather("app", appBindingCommands),
   }))
 
+  useBindings(() => ({
+    enabled: () => {
+      const ok = command.matcher.get()
+      if (!ok) return false
+      const current = promptRef.current
+      if (!current?.focused) return true
+      return current.current.input === ""
+    },
+    bindings: tuiConfig.keybinds.gather("app_exit", ["app.exit"]),
+  }))
+
   event.on(TuiEvent.CommandExecute.type, (evt) => {
     command.run(evt.properties.command)
   })

packages/opencode/src/cli/cmd/tui/component/prompt/autocomplete.tsx

@@ -18,6 +18,8 @@ import { Locale } from "@/util/locale"
 import type { PromptInfo } from "./history"
 import { useFrecency } from "./frecency"
 import { useBindings } from "../../keymap"
+import { Reference } from "@/reference/reference"
+import type { Config } from "@/config/config"
 
 function removeLineRange(input: string) {
   const hashIndex = input.lastIndexOf("#")
@@ -260,6 +262,87 @@ export function Autocomplete(props: {
     }
   }
 
+  function createReferenceFilePart(input: {
+    alias: string
+    root: string
+    item: string
+    lineRange?: { startLine: number; endLine?: number }
+  }) {
+    const filename = `${input.alias}/${
+      input.lineRange && !input.item.endsWith("/")
+        ? `${input.item}#${input.lineRange.startLine}${input.lineRange.endLine ? `-${input.lineRange.endLine}` : ""}`
+        : input.item
+    }`
+    const urlObj = pathToFileURL(path.join(input.root, input.item))
+
+    if (input.lineRange && !input.item.endsWith("/")) {
+      urlObj.searchParams.set("start", String(input.lineRange.startLine))
+      if (input.lineRange.endLine !== undefined) {
+        urlObj.searchParams.set("end", String(input.lineRange.endLine))
+      }
+    }
+
+    return {
+      filename,
+      url: urlObj.href,
+      part: {
+        type: "file" as const,
+        mime: input.item.endsWith("/") ? "application/x-directory" : "text/plain",
+        filename,
+        url: urlObj.href,
+        source: {
+          type: "file" as const,
+          text: {
+            start: 0,
+            end: 0,
+            value: "",
+          },
+          path: filename,
+        },
+      },
+    }
+  }
+
+  function referencePromptText(reference: Reference.Resolved) {
+    const problem = reference.kind === "invalid" ? reference.message : undefined
+    return [
+      `Referenced configured reference @${reference.name}.`,
+      ...(reference.kind === "local" ? ["Kind: local directory"] : []),
+      ...(reference.kind === "git" ? ["Kind: git repository"] : []),
+      ...(reference.kind === "invalid" ? [`Repository: ${reference.repository}`] : []),
+      ...(reference.kind === "git" ? [`Repository: ${reference.repository}`] : []),
+      ...(reference.kind === "git" && reference.branch ? [`Branch/ref: ${reference.branch}`] : []),
+      ...(reference.kind === "invalid" ? [] : [`Reference root: ${reference.path}`]),
+      ...(problem
+        ? [`Problem: ${problem}`]
+        : [
+            "For targeted context, inspect the reference path directly with Read, Glob, and Grep. For broader research, call the task tool with subagent scout and include this reference path.",
+          ]),
+    ].join("\n")
+  }
+
+  const references = createMemo(() =>
+    Reference.resolveAll({
+      references: (sync.data.config.reference ?? {}) as NonNullable<Config.Info["reference"]>,
+      directory: sync.path.directory || process.cwd(),
+      worktree: sync.path.worktree || sync.path.directory || process.cwd(),
+    }),
+  )
+
+  const referenceSearch = createMemo(() => {
+    if (!store.visible || store.visible === "/") return
+    const { lineRange, baseQuery } = extractLineRange(search())
+    const slash = baseQuery.indexOf("/")
+    if (slash === -1) return
+    const reference = references().find((item) => item.name === baseQuery.slice(0, slash))
+    if (!reference || reference.kind === "invalid") return
+    return {
+      reference,
+      query: baseQuery.slice(slash + 1),
+      lineRange,
+    }
+  })
+
   function normalizeMentionPath(filePath: string) {
     const baseDir = sync.path.directory || process.cwd()
     const absolute = path.resolve(filePath)
@@ -291,6 +374,7 @@ export function Autocomplete(props: {
     () => search(),
     async (query) => {
       if (!store.visible || store.visible === "/") return []
+      if (referenceSearch()) return []
 
       const { lineRange, baseQuery } = extractLineRange(query ?? "")
 
@@ -339,6 +423,43 @@ export function Autocomplete(props: {
     },
   )
 
+  const [referenceFiles] = createResource(
+    () => referenceSearch(),
+    async (match) => {
+      if (!match) return []
+
+      const result = await sdk.client.find.files({
+        directory: match.reference.path,
+        query: match.query,
+        limit: 50,
+      })
+
+      if (result.error || !result.data) return []
+
+      const width = props.anchor().width - 4
+      return result.data.map((item): AutocompleteOption => {
+        const { filename, part } = createReferenceFilePart({
+          alias: match.reference.name,
+          root: match.reference.path,
+          item,
+          lineRange: match.lineRange,
+        })
+        return {
+          display: Locale.truncateMiddle(filename, width),
+          value: filename,
+          isDirectory: item.endsWith("/"),
+          path: filename,
+          onSelect: () => {
+            insertPart(filename, part)
+          },
+        }
+      })
+    },
+    {
+      initialValue: [],
+    },
+  )
+
   const mcpResources = createMemo(() => {
     if (!store.visible || store.visible === "/") return []
 
@@ -397,6 +518,22 @@ export function Autocomplete(props: {
       )
   })
 
+  const referenceAliases = createMemo(() =>
+    references().map(
+      (reference): AutocompleteOption => ({
+        display: "@" + reference.name,
+        description: reference.kind === "invalid" ? reference.message : " configured reference",
+        onSelect: () => {
+          insertPart(reference.name, {
+            type: "text",
+            text: referencePromptText(reference),
+            synthetic: true,
+          })
+        },
+      }),
+    ),
+  )
+
   const commands = createMemo((): AutocompleteOption[] => {
     const results: AutocompleteOption[] = [...command.slashes()]
 
@@ -428,11 +565,18 @@ export function Autocomplete(props: {
 
   const options = createMemo((prev: AutocompleteOption[] | undefined) => {
     const filesValue = files()
+    const referenceFilesValue = referenceFiles()
+    const referenceSearchValue = referenceSearch()
     const agentsValue = agents()
+    const referenceAliasesValue = referenceAliases()
     const commandsValue = commands()
 
     const mixed: AutocompleteOption[] =
-      store.visible === "@" ? [...agentsValue, ...(filesValue || []), ...mcpResources()] : [...commandsValue]
+      store.visible === "@"
+        ? referenceSearchValue
+          ? referenceFilesValue || []
+          : [...referenceAliasesValue, ...agentsValue, ...(filesValue || []), ...mcpResources()]
+        : [...commandsValue]
 
     const searchValue = search()
 
@@ -440,7 +584,7 @@ export function Autocomplete(props: {
       return mixed
     }
 
-    if (files.loading && prev && prev.length > 0) {
+    if ((files.loading || referenceFiles.loading) && prev && prev.length > 0) {
       return prev
     }
 
@@ -505,7 +649,7 @@ export function Autocomplete(props: {
     const input = props.input()
     const currentCursorOffset = input.cursorOffset
 
-    const displayText = selected.display.trimEnd()
+    const displayText = (selected.value ?? selected.display).trimEnd()
     const path = displayText.startsWith("@") ? displayText.slice(1) : displayText
 
     input.cursorOffset = store.index

packages/opencode/src/cli/cmd/tui/component/prompt/index.tsx

@@ -712,7 +712,6 @@ export function Prompt(props: PromptProps) {
       ...input.traits,
       ...computePromptTraits({
         mode: store.mode,
-        disabled: !!props.disabled,
         autocompleteVisible: !!auto()?.visible,
       }),
     }
@@ -927,13 +926,7 @@ export function Prompt(props: PromptProps) {
       target: inputTarget,
       enabled: (() => {
         cursorVersion()
-        return (
-          inputTarget() !== undefined &&
-          !props.disabled &&
-          !auto()?.visible &&
-          input !== undefined &&
-          (input.cursorOffset === 0 || input.visualCursor.visualRow === 0)
-        )
+        return inputTarget() !== undefined && !props.disabled && !auto()?.visible && input !== undefined
       })(),
       commands: [
         {
@@ -942,12 +935,12 @@ export function Prompt(props: PromptProps) {
           category: "Prompt",
           run() {
             if (input.cursorOffset !== 0) {
-              input.cursorOffset = 0
-              return
+              if (input.scrollY + input.visualCursor.visualRow === 0) input.cursorOffset = 0
+              return false
             }
 
             const item = history.move(-1, input.plainText)
-            if (!item) return
+            if (!item) return false
             input.setText(item.input)
             setStore("prompt", item)
             setStore("mode", item.mode ?? "normal")
@@ -965,13 +958,7 @@ export function Prompt(props: PromptProps) {
       target: inputTarget,
       enabled: (() => {
         cursorVersion()
-        return (
-          inputTarget() !== undefined &&
-          !props.disabled &&
-          !auto()?.visible &&
-          input !== undefined &&
-          (input.cursorOffset === input.plainText.length || input.visualCursor.visualRow === input.height - 1)
-        )
+        return inputTarget() !== undefined && !props.disabled && !auto()?.visible && input !== undefined
       })(),
       commands: [
         {
@@ -980,12 +967,16 @@ export function Prompt(props: PromptProps) {
           category: "Prompt",
           run() {
             if (input.cursorOffset !== input.plainText.length) {
-              input.cursorOffset = input.plainText.length
-              return
+              if (
+                input.scrollY + input.visualCursor.visualRow ===
+                Math.max(0, input.editorView.getTotalVirtualLineCount() - 1)
+              )
+                input.cursorOffset = input.plainText.length
+              return false
             }
 
             const item = history.move(1, input.plainText)
-            if (!item) return
+            if (!item) return false
             input.setText(item.input)
             setStore("prompt", item)
             setStore("mode", item.mode ?? "normal")

packages/opencode/src/cli/cmd/tui/component/prompt/traits.ts

@@ -4,7 +4,6 @@ export type PromptMode = "normal" | "shell"
 
 export interface PromptTraitsInput {
   mode: PromptMode
-  disabled: boolean
   autocompleteVisible: boolean
 }
 
@@ -16,10 +15,9 @@ export type PromptTraits = EditorTraits & {
 /**
  * Compute the textarea editor traits for the prompt.
  *
- * `traits.suspend` gates the textarea's keybinding actions (backspace,
- * delete-word, arrow movement, undo/redo, etc.). Shell mode is an active
- * editing mode — only `disabled` should suspend the textarea, otherwise
- * users can type in shell mode but cannot delete or move the cursor.
+ * The OpenTUI managed textarea keymap owns `traits.suspend`. Prompt traits
+ * only expose capture/status metadata so focus changes cannot unsuspend the
+ * keymap-managed editor mappings.
  */
 export function computePromptTraits(input: PromptTraitsInput): PromptTraits {
   const capture =
@@ -30,7 +28,6 @@ export function computePromptTraits(input: PromptTraitsInput): PromptTraits {
       : undefined
   return {
     capture,
-    suspend: input.disabled,
     status: input.mode === "shell" ? "SHELL" : undefined,
     owner: "opencode",
     role: "prompt",

packages/opencode/src/cli/cmd/tui/keymap.tsx

@@ -8,9 +8,9 @@ import {
 import {
   KeymapProvider,
   reactiveMatcherFromSignal,
-  useBindings,
   useKeymap,
   useKeymapSelector,
+  useBindings,
 } from "@opentui/keymap/solid"
 import type { Accessor } from "solid-js"
 import type { TuiConfig } from "./config/tui"
@@ -26,6 +26,28 @@ export { reactiveMatcherFromSignal, useBindings, useKeymapSelector }
 
 export type OpenTuiKeymap = ReturnType<typeof useKeymap>
 
+const KEY_ALIASES = {
+  enter: "return",
+  esc: "escape",
+} as const
+
+function expandKeyAliases(input: string) {
+  const result = Object.entries(KEY_ALIASES).reduce(
+    (acc, [alias, key]) => acc.replace(new RegExp(`(^|[+,\\s>])${alias}(?=$|[+,\\s<])`, "gi"), `$1${key}`),
+    input,
+  )
+  if (result === input) return
+  return result
+}
+
+function registerKeyAliases(keymap: OpenTuiKeymap) {
+  return keymap.appendBindingExpander((ctx) => {
+    const key = expandKeyAliases(ctx.input)
+    if (!key) return
+    return [{ key, displays: ctx.displays }]
+  })
+}
+
 const inputCommands = [
   "input.move.left",
   "input.move.right",
@@ -98,8 +120,13 @@ export function formatKeyBindings(
   return formatCommandBindingsExtra(bindings, formatOptions(config))
 }
 
-export function registerOpencodeKeymap(keymap: OpenTuiKeymap, renderer: CliRenderer, config: TuiConfig.Resolved) {
+export function registerOpencodeKeymap(
+  keymap: OpenTuiKeymap,
+  renderer: CliRenderer,
+  config: Pick<TuiConfig.Resolved, "keybinds" | "leader_timeout">,
+) {
   const offCommaBindings = addons.registerCommaBindings(keymap)
+  const offAliasExpander = registerKeyAliases(keymap)
   const offBaseLayout = addons.registerBaseLayoutFallback(keymap)
   const offLeader = addons.registerTimedLeader(keymap, {
     trigger: config.keybinds.get(LEADER_TOKEN),
@@ -108,20 +135,17 @@ export function registerOpencodeKeymap(keymap: OpenTuiKeymap, renderer: CliRende
   })
   const offEscape = addons.registerEscapeClearsPendingSequence(keymap)
   const offBackspace = addons.registerBackspacePopsPendingSequence(keymap)
-  const offInputCommands = addons.registerEditBufferCommands(keymap, renderer)
-  const offInputSuspension = addons.registerTextareaMappingSuspension(keymap, renderer)
-  const offInputBindings = keymap.registerLayer({
+  const offInputBindings = addons.registerManagedTextareaLayer(keymap, renderer, {
     enabled: () => renderer.currentFocusedEditor !== null,
     bindings: config.keybinds.gather("input", inputCommands),
   })
 
   return () => {
     offInputBindings()
-    offInputSuspension()
-    offInputCommands()
     offBackspace()
     offEscape()
     offLeader()
+    offAliasExpander()
     offBaseLayout()
     offCommaBindings()
   }

packages/opencode/src/cli/cmd/tui/routes/session/index.tsx

@@ -746,7 +746,7 @@ export function Session() {
       title: "Line up",
       value: "session.line.up",
       category: "Session",
-      enabled: false,
+      hidden: true,
       run: () => {
         scroll.scrollBy(-1)
         dialog.clear()
@@ -756,7 +756,7 @@ export function Session() {
       title: "Line down",
       value: "session.line.down",
       category: "Session",
-      enabled: false,
+      hidden: true,
       run: () => {
         scroll.scrollBy(1)
         dialog.clear()

packages/opencode/src/cli/cmd/tui/validate-session.ts

@@ -1,5 +1,8 @@
 import { createOpencodeClient } from "@opencode-ai/sdk/v2"
 import { SessionID } from "@/session/schema"
+import { Schema } from "effect"
+
+const decodeSessionID = Schema.decodeUnknownSync(SessionID)
 
 export async function validateSession(input: {
   url: string
@@ -10,9 +13,11 @@ export async function validateSession(input: {
 }) {
   if (!input.sessionID) return
 
-  const result = SessionID.zod.safeParse(input.sessionID)
-  if (!result.success) {
-    throw new Error(`Invalid session ID: ${result.error.issues.at(0)?.message ?? "unknown error"}`)
+  let sessionID: SessionID
+  try {
+    sessionID = decodeSessionID(input.sessionID)
+  } catch (error) {
+    throw new Error(`Invalid session ID: ${error instanceof Error ? error.message : "unknown error"}`, { cause: error })
   }
 
   await createOpencodeClient({
@@ -20,5 +25,5 @@ export async function validateSession(input: {
     directory: input.directory,
     fetch: input.fetch,
     headers: input.headers,
-  }).session.get({ sessionID: result.data }, { throwOnError: true })
+  }).session.get({ sessionID }, { throwOnError: true })
 }

packages/opencode/src/config/command.ts

@@ -1,12 +1,10 @@
 export * as ConfigCommand from "./command"
 
 import * as Log from "@opencode-ai/core/util/log"
-import { Schema } from "effect"
+import { Cause, Exit, Schema } from "effect"
 import { NamedError } from "@opencode-ai/core/util/error"
 import { Glob } from "@opencode-ai/core/util/glob"
 import { Bus } from "@/bus"
-import { zod } from "@opencode-ai/core/effect-zod"
-import { withStatics } from "@opencode-ai/core/schema"
 import { configEntryNameFromPath } from "./entry-name"
 import { InvalidError } from "./error"
 import * as ConfigMarkdown from "./markdown"
@@ -20,10 +18,12 @@ export const Info = Schema.Struct({
   agent: Schema.optional(Schema.String),
   model: Schema.optional(ConfigModelID),
   subtask: Schema.optional(Schema.Boolean),
-}).pipe(withStatics((s) => ({ zod: zod(s) })))
+})
 
 export type Info = Schema.Schema.Type<typeof Info>
 
+const decodeInfo = Schema.decodeUnknownExit(Info)
+
 export async function load(dir: string) {
   const result: Record<string, Info> = {}
   for (const item of await Glob.scan("{command,commands}/**/*.md", {
@@ -51,12 +51,12 @@ export async function load(dir: string) {
       ...md.data,
       template: md.content.trim(),
     }
-    const parsed = Info.zod.safeParse(config)
-    if (parsed.success) {
-      result[config.name] = parsed.data
+    const parsed = decodeInfo(config, { errors: "all", propertyOrder: "original" })
+    if (Exit.isSuccess(parsed)) {
+      result[config.name] = parsed.value
       continue
     }
-    throw new InvalidError({ path: item, issues: parsed.error.issues }, { cause: parsed.error })
+    throw new InvalidError({ path: item, message: Cause.pretty(parsed.cause) }, { cause: Cause.squash(parsed.cause) })
   }
   return result
 }

packages/opencode/src/config/config.ts

@@ -145,7 +145,7 @@ export const Info = Schema.Struct({
   }),
   skills: Schema.optional(ConfigSkills.Info).annotate({ description: "Additional skill folder paths" }),
   reference: Schema.optional(ConfigReference.Info).annotate({
-    description: "Named git or local directory references that can be @ mentioned as Scout-backed subagents",
+    description: "Named git or local directory references that can be mentioned as @alias or @alias/path",
   }),
   watcher: Schema.optional(
     Schema.Struct({
@@ -273,10 +273,10 @@ export const Info = Schema.Struct({
       }),
       tail_turns: Schema.optional(NonNegativeInt).annotate({
         description:
-          "Number of recent user turns, including their following assistant/tool responses, to keep verbatim during compaction (default: 2)",
+          "Number of recent user turns, including their following assistant/tool responses, to serialize into the compaction summary (default: 2)",
       }),
       preserve_recent_tokens: Schema.optional(NonNegativeInt).annotate({
-        description: "Maximum number of tokens from recent turns to preserve verbatim after compaction",
+        description: "Maximum number of tokens from recent turns to serialize into the compaction summary",
       }),
       reserved: Schema.optional(NonNegativeInt).annotate({
         description: "Token buffer for compaction. Leaves enough window to avoid overflow during compaction.",

packages/opencode/src/control-plane/schema.ts

@@ -1,7 +1,6 @@
 import { Schema } from "effect"
 
 import { Identifier } from "@/id/id"
-import { zod } from "@opencode-ai/core/effect-zod"
 import { withStatics } from "@opencode-ai/core/schema"
 
 const workspaceIdSchema = Schema.String.check(Schema.isStartsWith("wrk")).pipe(Schema.brand("WorkspaceID"))
@@ -11,6 +10,5 @@ export type WorkspaceID = typeof workspaceIdSchema.Type
 export const WorkspaceID = workspaceIdSchema.pipe(
   withStatics((schema: typeof workspaceIdSchema) => ({
     ascending: (id?: string) => schema.make(Identifier.ascending("workspace", id)),
-    zod: zod(schema),
   })),
 )

packages/opencode/src/control-plane/workspace.ts

@@ -641,7 +641,7 @@ export const layer = Layer.effect(
 
             // "claim" this session so any future events coming from
             // the old workspace are ignored
-            SyncEvent.claim(input.sessionID, input.workspaceID ?? previous.projectID)
+            yield* sync.claim(input.sessionID, input.workspaceID ?? previous.projectID)
           }
         }
 
@@ -676,14 +676,12 @@ export const layer = Layer.effect(
         }
 
         if (input.workspaceID === null) {
-          yield* Effect.sync(() =>
-            SyncEvent.run(Session.Event.Updated, {
-              sessionID: input.sessionID,
-              info: {
-                workspaceID: null,
-              },
-            }),
-          )
+          yield* sync.run(Session.Event.Updated, {
+            sessionID: input.sessionID,
+            info: {
+              workspaceID: null,
+            },
+          })
 
           log.info("session warp complete", {
             workspaceID: input.workspaceID,

packages/opencode/src/data-migration.sql.ts

@@ -0,0 +1,6 @@
+import { integer, sqliteTable, text } from "drizzle-orm/sqlite-core"
+
+export const DataMigrationTable = sqliteTable("data_migration", {
+  name: text().primaryKey(),
+  time_completed: integer().notNull(),
+})