Repository navigation
release v0.1.4 - #14
Merged
Merged
Conversation
This was referenced Sep 2, 2026
antclt
pushed a commit
to antclt/billion-context
that referenced
this pull request
Sep 3, 2026
…c passthrough (ranxianglei#206 parity, issue ranxianglei#14)
This was referenced Sep 20, 2026
Closed
Merged
ranxianglei
pushed a commit
that referenced
this pull request
Oct 11, 2026
…tentional drift) The golden-google drive runs compress.injectTool+injectNudge, so after the fold at body #9 the proxy injects its ephemeral [ACP] status-marker line into the next request's history (#717/#1029). Pre-fix that line rode into the google lane's projection on every subsequent turn; post-fix prepareGoogle strips it before projection — which is exactly what #2798 changes, now observable in the pin set. Delta is exactly the strip effect and nothing else: outbound bodies #10-#14 each shrink by a constant 268 bytes (the fixed marker content removed from contents); stateDigest/sectionDigests(state,stats,metadata)/stateKeyDigests(messageRefs,tokenSnapshot,lastPassIds) move accordingly; blockContents digest, blocks/nudge/deadRefs/hiddenOrphanRefs/terminalStreak digests, foldAt/editedAt/summaryFrom and stateShape (incl. refs count) are unchanged. The other three wires' goldens regenerate byte-identical (no git diff for chat/anthropic/responses), confirming no cross-lane effect. Regenerated with node --import tsx scripts/update-storage-render-goldens.ts (#2798, Closes via PR #2805).
This was referenced Oct 21, 2026
ranxianglei
added a commit
that referenced
this pull request
Oct 21, 2026
Third instance of the same miss as the anthropic round-trip harness: the #3989 view-latch induction judge calls solveSmt2 with no solver-availability gate, so every solverless CI shard fails with 'no solver available' instead of skipping per KDD #14. Same probe + skip pattern as the google harness; the solver-free behavior-matrix test keeps running everywhere.
ranxianglei
pushed a commit
that referenced
this pull request
Oct 21, 2026
…dev-machine paths + missing no-solver skips Incidental CI-red fix found while triaging #4050's run: master 56a0f5a (this branch's base) fails its own test shards on pre-existing verif-harness defects unrelated to the #4050 diff — verified against master CI run 39528617351: same ENOENT on the anthropic round-trip tests (ubuntu-22 shard 2) and same cvc5 ModuleNotFoundError on view-latch (ubuntu-22 shard 3). 1. tests/verif-wire-anthropic-roundtrip.test.ts + tests/verif-wire-google-rebuild-stability.test.ts carried absolute dev-machine paths (/home/dog/tmp/bili-verif2/kernel/src/wire/*.ts) into translateFunction({ file }) -> ENOENT on every runner. Now resolved repo-relative via new URL(..., import.meta.url), the pattern the other verif harnesses already use. 2. tests/verif-wire-anthropic-roundtrip.test.ts + tests/verif-view-latch-discipline.test.ts called solveSmt2 without the no-solver skip guard the other nine verif test files carry (KDD #14: solver-less environments auto-skip, never red). Added the module-level probe + { skip: !haveSolver } option exactly as tests/verif-wire-google-rebuild-stability.test.ts does it. Verified both directions locally: with a solver present all 8 tests run and pass; with CVC5_HOME/Z3_BIN pointed away (simulating the ubuntu-22 runners) the 5 solver tests SKIP and the 3 real-code baselines still pass.
This was referenced Oct 21, 2026
ranxianglei
pushed a commit
that referenced
this pull request
Oct 21, 2026
… (CI triage #4062) Found while triaging the #4062 CI red: the async test body kept executing after t.skip() because the gate line had no return. On every solverless CI shard it then threw deeper (no-solver error on Linux, doubled-drive-letter ENOENT on Windows), producing a 'not ok ... # SKIP' TAP line that is NOT counted in the failure total but reads like a failure and misled this triage. The return makes the skip honest per the KDD #14 auto-skip contract; behavior with a solver present is unchanged (branch never taken).
cskwork
pushed a commit
to cskwork/billion-context
that referenced
this pull request
Oct 21, 2026
…lei#14) Adapts the proxy to acp-kernel 0.0.45's absorb pipeline: - src/absorb.ts: absorb settings resolution (absorbEnabled, applyAbsorbConfig with percent parsing) and executeAbsorb — the wire result text carries the model's summary verbatim inside <absorbed-summary> (the absorb tool_use is rewritten to text, so this is the only durable record). - Tool injection for all three protocols (anthropic/openai/responses, streaming loops included), plugin manifest + MCP tool list now advertise absorb; plugin tool API can execute it. - Kernel 0.0.31 -> 0.0.45-pr.160.14 (exact pin; flips to 0.0.45 when kernel release PR ranxianglei#160 merges): - openaiToCore now hoists the leading system/developer prefix out of the fold space: prepareOpenai re-injects systemText (and restores a developer head role) or the host system prompt is silently lost. - KEEP_LAST_ORPHANED=2 changed hide-consumed semantics: plugin-mode wire echoes of executed compress calls are orphans that stay visible; hidePluginEchoCompressCalls drops the consumed tail pair. - Tests: roundtrip hoist contract, manifest/MCP 5-tool lists, 6 new absorb unit tests. 518/518 pass.
This was referenced Oct 21, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Release v0.1.4
Cache-health release. Three prefix-cache killers fixed + durable logging.
Includes (since v0.1.3)
condenseOldToolResults(prefix-cache killer) — stubbed old tool results every turn, rewriting stable history bytesisTitleGenflip (OpenAI/chat path) —messages.length<=2made turn-1 system differ from turn-2+, breaking cache every sessioncache_controlthrough the core round-trip + system block (M2+M3 from audit)~/.local/state/billion-context/bili.logby default (file + stderr, 10MB rotate)[acp-usage]cache logging on the OpenAI chat path — previously only the Responses/Codex path logged cache hit %Verified cache health
Pre-flight
Release mechanism
Merging this PR (commit
release v0.1.4on*_release-v*branch) triggersrelease.yml→ npm publish--tag latest+ git tagv0.1.4.Tracking
reapOrphanBlocks→ acp-kernel move: refactor: reapOrphanBlocks should live in acp-kernel, not the proxy #12 (not in this release)