5 Commits

Author SHA1 Message Date
teamchong de0483cf51 fix(transform): wrap relocated env text in <system-reminder> for correct attribution 2026-07-03 01:50:08 -04:00
teamchong 06a8b70c01 fix(transform): relocate volatile env text behind ALL cache breakpoints
Volatile env/context text (git status, cwd, date) rode in req.system —
BEFORE the slab anchor in Anthropic's prefix order (tools → system →
messages) — so any git-state change cold-restarted the entire anchored
prefix. Telemetry attribution (events.jsonl 2026-06-26..07-02): 48.8%
of cold-create waste, ~2.6k tokens/session vs ~200 saved by imaging it.

Now appended as a trailing text block on the LAST user message — the
per-turn live tail that re-caches incrementally anyway — placed AFTER
history collapse (never baked into frozen chunks) and after tool_result
compression. Session-stable billingLine/sysRemainder stay in system;
fallback keeps env in system when no user message exists to carry it.
New telemetry field: envRelocatedChars.
2026-07-02 23:17:13 -04:00
teamchong 0ad3c68664 fix(transform): cross-session slab stability + drop compressSchemas knob
Fix #1 (cross-session cache bust): strip the volatile "# Environment"
markdown section (working dir, git status, platform, model ID) out of
the system text BEFORE the static/dynamic split. It carries no XML
wrapper so splitStaticDynamic couldn't catch it, and its git-status
lines change across sessions - baking them into the slab PNG busted
the cross-session cache (system_sha8 717f1fce -> 5efaa4bb for a
one-file edit). Parallel to stripBillingLine: the section re-enters
the system tail as plain text.

Schema handling simplified, knob removed: the Anthropic text-reference
path now renders prose-only tool docs (schemas stay untouched in
tools[] - duplicating them in the reference was pure bloat), while the
GPT imaged path always carries the compact schema (there it IS the
compression). Deletes the {type:'object'} stub + SCHEMA_STRUCTURAL_KEYS
machinery, stripSchemaDescriptions import, and the COMPRESS_SCHEMAS env.

Tests: e2e regressions for anchor relocation onto byte-frozen carry-over
history images and first-collapse anchor staying on the slab; history
relocation fixture grown past the collapseChunk=50 snap (56 messages ->
cutoff 50, frozen windows 11/21/31/41, collapsedTurns 49 - the old
35-message fixture floored to one window and never froze a chunk).

tsc clean; 615/615 tests pass.
2026-07-02 17:36:18 -04:00
teamchong 6d1c46dd20 fix(cache): pin prefix-cache anchor to the byte-stable history image
Intermittent prefix-cache busts on long sessions (#11): the single cache
breakpoint relocated onto the LAST history image — the still-growing chunk
whose bytes change every window advance — so the slab+history prefix
re-created at 1.25x instead of being read.

- relocateAnchorToHistoryImage pins collapseHistory's carry-over ordinal
  (the last byte-stable chunk) instead of the last image.
- demoteProtectedHeadText keeps pxpipe's slab scaffolding (images +
  fact-sheet + the [End of rendered context.] boundary) verbatim and only
  quarantines the user's stale opening turn, so the relocator can still
  find the slab anchor (#14 demotion had clobbered the boundary it keys on).
- cachePrefixDigest joins parts with \x00 so block-boundary differences
  can't hash-collide into false prefix-stability.

520/520 green, tsc clean.
2026-06-26 00:17:10 -04:00
teamchong 8417380061 test: e2e guards for cache stability, dashboard honesty, and design behaviors
Test-only (no src changes). Adds fake-upstream + real-createProxy suites:

- cache-stability-e2e: byte-identical cacheable prefix across growth, marker
  conservation + relocation onto the history image, tool-pair integrity.
- savings-math-e2e: gate never images at a net loss; baselineImagedTokens
  cross-checked against the real o200k tokenizer; baselineTokens wired from
  the count_tokens probe.
- savings-honesty: categorical no-overclaim invariants (warm <= cold,
  saved==delta*weight, zero credit on passthrough) + per-model pricing guards
  (Anthropic 1.25/0.1 shared policy; GPT 0.1x model-gated).
- design-behavior-e2e: system-prompt imaging, history collapse (old imaged /
  recent legible), tools-in-recent-turns kept usable.
- dashboard-api: GPT cold-cache parity (full text delta, not 0.1x warm rate).

Each guard mutation-verified (broke source, confirmed red, reverted).
2026-06-23 12:30:16 -04:00