Commit Graph
4 Commits
Author SHA1 Message Date
adminandClaude Fable 5 f0a1711ebc fix(aqua): batch_runner path sanitize, loader batch embed, alias expansion
- batch_runner: sanitize paper_id to disk dir name (was skipping all
  papers whose paper_id contains :/()); keeps explicit result path
  compat; verified 5/5 no-skip on paren-DOI papers
- qdrant_loader: batch embed via Ollama array input (32/batch, 14x:
  2.7 -> 36.9 texts/s, 24.5k points ~10min vs 2.5h); per-batch fail
  falls back to single; checkpoint format unchanged
- entity_alias_aqua: +Streptococcus agalactiae (GBS), RAS
  (forbidden_expansions), Danio rerio; merged vernacular/Chinese
  aliases for O. niloticus / O. mykiss / S. salar / L. vannamei;
  g02 gap entities now normalize correctly
- note for P3: ingest manifest MUST carry paper sidecar (pilot
  manifest_pilot.json does not - would emit paper_id="result" dirt)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-09-13 01:06:38 +08:00
adminandClaude Fable 5 688edba169 docs(aqua): backfill measured budget numbers from 530-paper pilot
- measured: MD 1.6 calls/paper, PDF 28.2 calls/paper (assumed 8)
- full-scale LLM cost: ¥3110 (was ¥454 assumption-based)
- wall-clock: MD ~1.8d @4w; PDF ~4.5d @8w LLM-bound

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-09-12 22:20:06 +08:00
adminandClaude Fable 5 730692a1f6 feat(aqua): KB build tooling, aqua configs, QA artifacts (P1)
- phase2b_converter: result.json -> neo4j.cypher + qdrant_points.jsonl
  (payload keys replicate direct_flow_retriever contract; kb_version=
  aqua_srt_v1_20260930_normalized, source_system=Aqua_canonical)
- qdrant_loader: bge-m3 embed + batched upsert + checkpoint resume
- registry_builder: paper blocks -> registry jsonl (plant format)
- batch_runner: manifest -> convert -> neo4j -> qdrant, all checkpointed
- start_aqua_kb.sh: user-space trial instances (Neo4j 19889 / Qdrant 27435)
- smoke 11/11 green incl. resume drills
- aqua configs: entity_alias_aqua.yaml (22 entries), relation_taxonomy_aqua.yaml
  (45 groups); manifest points to aqua configs
- QA artifacts: golden_aqua.jsonl (41 Qs zh/en), qc_standards.md,
  budget_template.md + cost_model.py (measured pilot tokens)
- pilot: 500 MD + 30 PDF extracted, 0 failures; QC stats computed

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-09-12 22:19:40 +08:00
admin afa7cce938 feat(aqua): unlock normalized KB identity via env with plant fallback
- normalized_release EXPECTED_RELEASE defaults to NORMALIZED_RELEASE_ID (plant value)
- from_env falls back to constant on empty env; backend_guard uses env-or-constant
- start.sh sources env/active_library.env after prod.env (qa + legacy-rag)
- build_all.sh --library <name|dir> renders library identity (no name whitelist)
- add kb-libraries skeleton (plant/aqua identity cards) and tooling/aqua_kb
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-09-12 19:10:08 +08:00