fix: keep test builds and versions in sync#8
Open
123kaze wants to merge 1 commit into
Open
Conversation
Goldentrii
added a commit
that referenced
this pull request
Jun 20, 2026
…early prior Unify recall confidence onto a single 0..1 axis, drill into the lossless archive when the model tier is unsure, and push a correction-derived prior before the agent reasons. - confidence.ts: calibratedConfidence(score, scale) + CONFIDENCE_FLOOR (high .66 / medium .4 / low .2). Deletes the 3 divergent private scoreLabel fns (smart-recall, supabase recall-backend, local-vector-backend) and routes every call site through it with its TRUE native scale (cosine | rrf-local | rrf-supabase). calibrated:number now set at scoring time on every item. - drill-down.ts: fetchVerbatim(project, key) — local-only, journal (date-regex validated) + palace (sanitizeSlug + path-escape assert), capped 1200 chars, never throws into recall. - smart-recall.ts: stamp verbatimKey on journal/palace items; add drilldown kill-switch + bridged[] result. Bridge gate reads the STORED calibrated (scoring time), NOT the hot-window/feedback-boosted final score (Risk #8); caps <=2 items; skips high-confidence and graph-walk items. - prior-builder.ts: buildPriors(prompt, corrections, blindSpots) — pure, exported, reuses check-action tokenize/overlap (now exported), >=2 overlap gate, caps 2. hook-ambient emits priors ABOVE the fact list, best-effort. - memory-query: when the high/medium filter empties, attach the bridged verbatim under `fallback` with a verify-before-relying caution instead of a bare "nothing found"; MCP recall.ts + memory-query.ts render it. Tests: confidence, drill-down, prior-builder, memory-query (.test.mjs, node:test, import from dist). Full suite green (378 tests, 0 fail). Local-only: no version bump, no publish, no push. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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
1 participant
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.
背景
当前仓库在干净 checkout 后直接运行测试会失败，因为测试文件会从
dist/目录加载构建产物，但测试脚本本身不会先执行构建。此外，部分 workspace 的版本信息存在漂移：包版本已是3.4.19，但 core 的VERSION常量、内部依赖和 lockfile 仍停留在旧版本。变更内容
npm test在运行测试前自动执行必要的 build。agent-recall-core的内部依赖统一更新到3.4.19。packages/core/src/types.ts中导出的VERSION对齐到3.4.19。package-lock.json中的 workspace 版本信息。价值
这能保证新环境或 CI 中直接运行测试时不会因为缺少
dist/构建产物而失败，同时避免版本漂移导致一致性测试失败或发布信息不准确。验证
已通过：
npm testnpm run lint