refactor(session): consume native LLM events

packages/opencode/test/session/compaction.test.ts

@@ -28,6 +28,7 @@ import { testEffect } from "../lib/effect"
 import { CrossSpawnSpawner } from "@opencode-ai/core/cross-spawn-spawner"
 import { TestConfig } from "../fixture/config"
 import { SyncEvent } from "@/sync"
+import { LLMEvent, Usage } from "@opencode-ai/llm"
 
 void Log.init({ print: false })
 
@@ -45,6 +46,10 @@ const ref = {
   modelID: ModelID.make("test-model"),
 }
 
+const usage = (input: ConstructorParameters<typeof Usage>[0]) => new Usage(input)
+
+const basicUsage = () => usage({ inputTokens: 1, outputTokens: 1, totalTokens: 2 })
+
 afterEach(() => {
   mock.restore()
 })
@@ -289,11 +294,11 @@ function readCompactionPart(sessionID: SessionID) {
 
 function llm() {
   const queue: Array<
-    Stream.Stream<LLM.Event, unknown> | ((input: LLM.StreamInput) => Stream.Stream<LLM.Event, unknown>)
+    Stream.Stream<LLMEvent, unknown> | ((input: LLM.StreamInput) => Stream.Stream<LLMEvent, unknown>)
   > = []
 
   return {
-    push(stream: Stream.Stream<LLM.Event, unknown> | ((input: LLM.StreamInput) => Stream.Stream<LLM.Event, unknown>)) {
+    push(stream: Stream.Stream<LLMEvent, unknown> | ((input: LLM.StreamInput) => Stream.Stream<LLMEvent, unknown>)) {
       queue.push(stream)
     },
     layer: Layer.succeed(
@@ -312,54 +317,22 @@ function llm() {
 function reply(
   text: string,
   capture?: (input: LLM.StreamInput) => void,
-): (input: LLM.StreamInput) => Stream.Stream<LLM.Event, unknown> {
+): (input: LLM.StreamInput) => Stream.Stream<LLMEvent, unknown> {
   return (input) => {
     capture?.(input)
     return Stream.make(
-      { type: "start" } satisfies LLM.Event,
-      { type: "text-start", id: "txt-0" } satisfies LLM.Event,
-      { type: "text-delta", id: "txt-0", delta: text, text } as LLM.Event,
-      { type: "text-end", id: "txt-0" } satisfies LLM.Event,
-      {
-        type: "finish-step",
-        finishReason: "stop",
-        rawFinishReason: "stop",
-        response: { id: "res", modelId: "test-model", timestamp: new Date() },
-        providerMetadata: undefined,
-        usage: {
-          inputTokens: 1,
-          outputTokens: 1,
-          totalTokens: 2,
-          inputTokenDetails: {
-            noCacheTokens: undefined,
-            cacheReadTokens: undefined,
-            cacheWriteTokens: undefined,
-          },
-          outputTokenDetails: {
-            textTokens: undefined,
-            reasoningTokens: undefined,
-          },
-        },
-      } satisfies LLM.Event,
-      {
-        type: "finish",
-        finishReason: "stop",
-        rawFinishReason: "stop",
-        totalUsage: {
-          inputTokens: 1,
-          outputTokens: 1,
-          totalTokens: 2,
-          inputTokenDetails: {
-            noCacheTokens: undefined,
-            cacheReadTokens: undefined,
-            cacheWriteTokens: undefined,
-          },
-          outputTokenDetails: {
-            textTokens: undefined,
-            reasoningTokens: undefined,
-          },
-        },
-      } satisfies LLM.Event,
+      LLMEvent.textStart({ id: "txt-0" }),
+      LLMEvent.textDelta({ id: "txt-0", text }),
+      LLMEvent.textEnd({ id: "txt-0" }),
+      LLMEvent.stepFinish({
+        index: 0,
+        reason: "stop",
+        usage: basicUsage(),
+      }),
+      LLMEvent.requestFinish({
+        reason: "stop",
+        usage: basicUsage(),
+      }),
     )
   }
 }
@@ -1198,7 +1171,7 @@ describe("session.compaction.process", () => {
         Stream.fromAsyncIterable(
           {
             async *[Symbol.asyncIterator]() {
-              yield { type: "start" } as LLM.Event
+              yield LLMEvent.stepStart({ index: 0 })
               throw new APICallError({
                 message: "boom",
                 url: "https://example.com/v1/chat/completions",
@@ -1290,49 +1263,16 @@ describe("session.compaction.process", () => {
       const stub = llm()
       stub.push(
         Stream.make(
-          { type: "start" } satisfies LLM.Event,
-          { type: "tool-input-start", id: "call-1", toolName: "_noop" } satisfies LLM.Event,
-          { type: "tool-call", toolCallId: "call-1", toolName: "_noop", input: {} } satisfies LLM.Event,
-          {
-            type: "finish-step",
-            finishReason: "tool-calls",
-            rawFinishReason: "tool_calls",
-            response: { id: "res", modelId: "test-model", timestamp: new Date() },
-            providerMetadata: undefined,
-            usage: {
-              inputTokens: 1,
-              outputTokens: 1,
-              totalTokens: 2,
-              inputTokenDetails: {
-                noCacheTokens: undefined,
-                cacheReadTokens: undefined,
-                cacheWriteTokens: undefined,
-              },
-              outputTokenDetails: {
-                textTokens: undefined,
-                reasoningTokens: undefined,
-              },
-            },
-          } satisfies LLM.Event,
-          {
-            type: "finish",
-            finishReason: "tool-calls",
-            rawFinishReason: "tool_calls",
-            totalUsage: {
-              inputTokens: 1,
-              outputTokens: 1,
-              totalTokens: 2,
-              inputTokenDetails: {
-                noCacheTokens: undefined,
-                cacheReadTokens: undefined,
-                cacheWriteTokens: undefined,
-              },
-              outputTokenDetails: {
-                textTokens: undefined,
-                reasoningTokens: undefined,
-              },
-            },
-          } satisfies LLM.Event,
+          LLMEvent.toolCall({ id: "call-1", name: "_noop", input: {} }),
+          LLMEvent.stepFinish({
+            index: 0,
+            reason: "tool-calls",
+            usage: basicUsage(),
+          }),
+          LLMEvent.requestFinish({
+            reason: "tool-calls",
+            usage: basicUsage(),
+          }),
         ),
       )
       return Effect.gen(function* () {
@@ -1543,20 +1483,7 @@ describe("SessionNs.getUsage", () => {
     const model = createModel({ context: 100_000, output: 32_000 })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1000,
-        outputTokens: 500,
-        totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: undefined,
-          cacheReadTokens: undefined,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 1000, outputTokens: 500, totalTokens: 1500 }),
     })
 
     expect(result.tokens.input).toBe(1000)
@@ -1570,20 +1497,7 @@ describe("SessionNs.getUsage", () => {
     const model = createModel({ context: 100_000, output: 32_000 })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1000,
-        outputTokens: 500,
-        totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: 800,
-          cacheReadTokens: 200,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 1000, outputTokens: 500, totalTokens: 1500, cacheReadInputTokens: 200 }),
     })
 
     expect(result.tokens.input).toBe(800)
@@ -1594,20 +1508,7 @@ describe("SessionNs.getUsage", () => {
     const model = createModel({ context: 100_000, output: 32_000 })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1000,
-        outputTokens: 500,
-        totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: undefined,
-          cacheReadTokens: undefined,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 1000, outputTokens: 500, totalTokens: 1500 }),
       metadata: {
         anthropic: {
           cacheCreationInputTokens: 300,
@@ -1623,20 +1524,7 @@ describe("SessionNs.getUsage", () => {
     // AI SDK v6 normalizes inputTokens to include cached tokens for all providers
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1000,
-        outputTokens: 500,
-        totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: 800,
-          cacheReadTokens: 200,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 1000, outputTokens: 500, totalTokens: 1500, cacheReadInputTokens: 200 }),
       metadata: {
         anthropic: {},
       },
@@ -1650,20 +1538,7 @@ describe("SessionNs.getUsage", () => {
     const model = createModel({ context: 100_000, output: 32_000 })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1000,
-        outputTokens: 500,
-        totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: undefined,
-          cacheReadTokens: undefined,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: 400,
-          reasoningTokens: 100,
-        },
-      },
+      usage: usage({ inputTokens: 1000, outputTokens: 500, reasoningTokens: 100, totalTokens: 1500 }),
     })
 
     expect(result.tokens.input).toBe(1000)
@@ -1684,20 +1559,7 @@ describe("SessionNs.getUsage", () => {
     })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 0,
-        outputTokens: 1_000_000,
-        totalTokens: 1_000_000,
-        inputTokenDetails: {
-          noCacheTokens: undefined,
-          cacheReadTokens: undefined,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: 750_000,
-          reasoningTokens: 250_000,
-        },
-      },
+      usage: usage({ inputTokens: 0, outputTokens: 1_000_000, reasoningTokens: 250_000, totalTokens: 1_000_000 }),
     })
 
     expect(result.tokens.output).toBe(750_000)
@@ -1709,20 +1571,7 @@ describe("SessionNs.getUsage", () => {
     const model = createModel({ context: 100_000, output: 32_000 })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 0,
-        outputTokens: 0,
-        totalTokens: 0,
-        inputTokenDetails: {
-          noCacheTokens: undefined,
-          cacheReadTokens: undefined,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 0, outputTokens: 0, totalTokens: 0 }),
     })
 
     expect(result.tokens.input).toBe(0)
@@ -1745,20 +1594,7 @@ describe("SessionNs.getUsage", () => {
     })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1_000_000,
-        outputTokens: 100_000,
-        totalTokens: 1_100_000,
-        inputTokenDetails: {
-          noCacheTokens: undefined,
-          cacheReadTokens: undefined,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 1_000_000, outputTokens: 100_000, totalTokens: 1_100_000 }),
     })
 
     expect(result.cost).toBe(3 + 1.5)
@@ -1769,24 +1605,16 @@ describe("SessionNs.getUsage", () => {
     (npm) => {
       const model = createModel({ context: 100_000, output: 32_000, npm })
       // AI SDK v6: inputTokens includes cached tokens for all providers
-      const usage = {
+      const item = usage({
         inputTokens: 1000,
         outputTokens: 500,
         totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: 800,
-          cacheReadTokens: 200,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      }
+        cacheReadInputTokens: 200,
+      })
       if (npm === "@ai-sdk/amazon-bedrock") {
         const result = SessionNs.getUsage({
           model,
-          usage,
+          usage: item,
           metadata: {
             bedrock: {
               usage: {
@@ -1807,7 +1635,7 @@ describe("SessionNs.getUsage", () => {
 
       const result = SessionNs.getUsage({
         model,
-        usage,
+        usage: item,
         metadata: {
           anthropic: {
             cacheCreationInputTokens: 300,
@@ -1828,20 +1656,7 @@ describe("SessionNs.getUsage", () => {
     const model = createModel({ context: 100_000, output: 32_000, npm: "@ai-sdk/google-vertex/anthropic" })
     const result = SessionNs.getUsage({
       model,
-      usage: {
-        inputTokens: 1000,
-        outputTokens: 500,
-        totalTokens: 1500,
-        inputTokenDetails: {
-          noCacheTokens: 800,
-          cacheReadTokens: 200,
-          cacheWriteTokens: undefined,
-        },
-        outputTokenDetails: {
-          textTokens: undefined,
-          reasoningTokens: undefined,
-        },
-      },
+      usage: usage({ inputTokens: 1000, outputTokens: 500, totalTokens: 1500, cacheReadInputTokens: 200 }),
       metadata: {
         vertex: {
           cacheCreationInputTokens: 300,