deps: bump dikw-core from 0.6.0 to 0.6.5

[dependabot]

Bumps dikw-core from 0.6.0 to 0.6.5.

Release notes

Sourced from dikw-core's releases.

dikw-core v0.6.5

0.6.5 — Default scaffold ships Gitee embed + rerank; eval cache keys retrieval config and surfaces absolute relevance scores

Added

• Eval rows surface absolute relevance scores for OOD calibration. (#249) Each retrieval-eval per-query and negative row now carries top1_score (the top hit's score) and top1_vec_cosine — the reranker/fusion-independent raw top-1 vector cosine, captured by an eval-internal HybridSearcher.top_vector_cosine probe (the production search() path and its ranking are untouched). Fusion scores (RRF is rank-based; combsum/combmnz are per-leg min-max normalized) can't carry an absolute magnitude, so expect_none / out-of-distribution robustness was previously immeasurable from rank order alone; the absolute cosine makes it observable (covered query high, OOD query low). The vector probe is skipped for pure-bm25 ablations so they stay embedding-free. A score-based OOD metric is deferred — this release only surfaces the signals.

Changed

• Default scaffold ships Gitee embed + rerank; unified rerank/embed degrade-logging. dikw init (config.default_config) now defaults the embedder to Gitee bge-m3 (dim 1024) and ships a Gitee bge-reranker-v2-m3 reranker, both keyed by one GITEE_API_KEY, so a fresh base reranks out of the box (OpenAI has no /rerank endpoint, so the prior OpenAI embedding default couldn't pair a matching reranker). The LLM default stays Anthropic. Read-path resilience is now uniformly observable: a transient query-embedding failure degrades the hybrid query to FTS-only (vec leg dropped) instead of 500-ing — hybrid mode only, single-leg vector/bm25 ablations re-raise for eval purity; transient rerank / embed-batch-skip degrades now log at ERROR (a configured leg that failed, was WARNING); an enabled-but-unconfigured reranker and a write path that defers embedding (no embedder wired / version drift) each log a WARNING so the silently-off leg is visible. Permanent provider errors (401/403/404, bad key/model) still fail fast on both the read path (→ 500) and the write path (ingest aborts) — the fail-fast-on-misconfig invariant is unchanged.

Fixed

• Eval snapshot cache keys the ingest-time tokenizer and reads query-time retrieval config live. (#250) The eval corpus-snapshot cache key omitted RetrievalConfig, so under the default --cache read_write changing any retrieval knob (rrf_k / weights / fusion / rerank_enabled / graph_*) and re-running silently hit the stale snapshot and reused the previous config — no error, wrong numbers, exactly on the retrieval-ablation workflow. The cache key now includes the only ingest-time retrieval field, cjk_tokenizer (a change forces re-ingest); every search-time knob is read from the live config on each _run_queries, so ablations sharing one cache_root are now both fast and correct. A defensive guard re-raises if a cache hit's baked tokenizer ever disagrees with the live one.

... (truncated)

Changelog

Sourced from dikw-core's changelog.

0.6.5 — Default scaffold ships Gitee embed + rerank; eval cache keys retrieval config and surfaces absolute relevance scores

Added

• Eval rows surface absolute relevance scores for OOD calibration. (#249) Each retrieval-eval per-query and negative row now carries top1_score (the top hit's score) and top1_vec_cosine — the reranker/fusion-independent raw top-1 vector cosine, captured by an eval-internal HybridSearcher.top_vector_cosine probe (the production search() path and its ranking are untouched). Fusion scores (RRF is rank-based; combsum/combmnz are per-leg min-max normalized) can't carry an absolute magnitude, so expect_none / out-of-distribution robustness was previously immeasurable from rank order alone; the absolute cosine makes it observable (covered query high, OOD query low). The vector probe is skipped for pure-bm25 ablations so they stay embedding-free. A score-based OOD metric is deferred — this release only surfaces the signals.

Changed

• Default scaffold ships Gitee embed + rerank; unified rerank/embed degrade-logging. dikw init (config.default_config) now defaults the embedder to Gitee bge-m3 (dim 1024) and ships a Gitee bge-reranker-v2-m3 reranker, both keyed by one GITEE_API_KEY, so a fresh base reranks out of the box (OpenAI has no /rerank endpoint, so the prior OpenAI embedding default couldn't pair a matching reranker). The LLM default stays Anthropic. Read-path resilience is now uniformly observable: a transient query-embedding failure degrades the hybrid query to FTS-only (vec leg dropped) instead of 500-ing — hybrid mode only, single-leg vector/bm25 ablations re-raise for eval purity; transient rerank / embed-batch-skip degrades now log at ERROR (a configured leg that failed, was WARNING); an enabled-but-unconfigured reranker and a write path that defers embedding (no embedder wired / version drift) each log a WARNING so the silently-off leg is visible. Permanent provider errors (401/403/404, bad key/model) still fail fast on both the read path (→ 500) and the write path (ingest aborts) — the fail-fast-on-misconfig invariant is unchanged.

Fixed

• Eval snapshot cache keys the ingest-time tokenizer and reads query-time retrieval config live. (#250) The eval corpus-snapshot cache key omitted RetrievalConfig, so under the default --cache read_write changing any retrieval knob (rrf_k / weights / fusion / rerank_enabled / graph_*) and re-running silently hit the stale snapshot and reused the previous config — no error, wrong numbers, exactly on the retrieval-ablation workflow. The cache key now includes the only ingest-time retrieval field, cjk_tokenizer (a change forces re-ingest); every search-time knob is read from the live config on each _run_queries, so ablations sharing one cache_root are now both fast and correct. A defensive guard re-raises if a cache hit's baked tokenizer ever disagrees with the live one.

... (truncated)

Commits

• f8c1045 chore(release): prepare 0.6.5 (#254)
• c15b4ee feat(eval): surface absolute relevance scores in eval rows for OOD calibratio...
• c4e6bb4 fix(eval): key cjk_tokenizer + read query-time retrieval config live (#250) (...
• 91b530a feat: unify rerank/embed degrade-logging; default scaffold to Gitee embed+rer...
• c9762e4 chore(docker): bump DIKW_VERSION to v0.6.4 (#248)
• 8b4279a chore(release): prepare 0.6.4 (#247)
• 8d61455 refactor(prompts): extract knowledge system prompts to .md, unify product nam...
• e3032e4 chore(docker): bump DIKW_VERSION to v0.6.3 (#245)
• aab4ebd chore(release): prepare 0.6.3 (#244)
• 247eb63 feat(delivery): add tools/loop_metrics.py — delivery-loop effectiveness metri...
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

[dependabot]

Labels

The following labels could not be found: deps. Please create it before Dependabot can add it to a pull request.

Please fix the above issues or remove invalid values from dependabot.yml.

[helebest]

Code Review — dikw-core 0.6.0 → 0.6.5

Consistency: PR title matches the diff. All three plugin packages (dikw-converter-epub, dikw-converter-example, dikw-converter-mineru) have their minimum constraint updated from >=0.6.0 to >=0.6.5, and the lock file reflects the new wheel/sdist hashes. ✅

Correctness: Changes are limited to pyproject.toml dependency specifiers and the corresponding lock file entries. The lock file wheel URL, sdist URL, and hashes are consistent with the stated 0.6.5 release. ✅

Conventions: Standard workspace dependency-bump pattern. ✅

Security: dikw-core 0.6.1–0.6.5 changes are additive: eval OOD calibration fields, Gitee embed/rerank scaffold defaults, eval cache key correctness fix. All changes land in server/engine territory (dikw_core.server.*, dikw_core.domains.*). Per the repo's layering invariants, plugins may not import from those modules — they depend only on the Converter Protocol and Path. The public plugin API surface is unchanged across this version range. No CVEs or security advisories noted. ✅

Effectiveness: Bumping the minimum constraint ensures plugins receive the bug fix in 0.6.5 (eval cache key, not a plugin concern but correct version hygiene). ✅

Test coverage: No plugin source code changed; no new tests required. ✅

Approving. Will merge once CI is green.

---

Generated by Claude Code

[helebest]

Code Review — automated

Summary: dikw-core lower-bound bump from >=0.6.0 → >=0.6.5 across all three plugin packages (dikw-converter-epub, dikw-converter-example, dikw-converter-mineru) plus the lock file.

Dimension	Result
Consistency	✅ PR description matches the diff exactly; all three pyproject.toml files and the lock file are updated coherently
Correctness	✅ The lower-bound bump correctly reflects the new minimum; the 0.6.1–0.6.5 changes cover eval scoring, default scaffold configuration, and eval cache-key fixes — none of which touch the Converter Protocol or Path types these plugins depend on
Conventions	✅ dikw-core appears under each package's [project] dependencies, as expected
Security	✅ Changelog shows no security advisories; the new functionality (eval probes, Gitee defaults, reranker config) is in server/engine territory that the layering invariants in CLAUDE.md explicitly prohibit these plugins from importing, so the new code is simply not reachable from plugin code
Effectiveness	✅ Achieves its stated purpose
Test coverage	✅ No plugin logic changed; compatibility with the new dikw-core version is validated by the test suite running against the updated lock file

Verdict: LGTM. Proceeding to CI verification.

---

Generated by Claude Code

[helebest]

CI Verification — All checks green ✅

All 4 required checks passed:

Check	Result
test (3.12)	✅ success
test (3.13)	✅ success
analyze	✅ success
CodeQL	✅ success

No regressions. Proceeding to squash merge.

---

Generated by Claude Code

[helebest]

Verification & CI Results — automated

Local test suite: All 134 tests pass (uv run pytest --all-packages).

CI checks: All 4 checks completed green.

Check	Result
test (3.12)	✅ success
test (3.13)	✅ success
analyze (CodeQL)	✅ success
CodeQL	✅ success

All steps passed — proceeding with squash merge.

---

Generated by Claude Code