fix(eval): wire hard_provider into the eval harness — routing was never exercised (#66)

fix(eval): eval-023/eval-029 referenced nonexistent card aliases and a wrong ruling (#65)

feat(prompt): few-shot example for chain placement-order vs LIFO resolution (v7) (#64)

feat(rag): flip hard_query_routing and keyword_family_extra defaults (4.2+4.3, 3.5) (#63)

chore(rag): log HyDE failures instead of swallowing them silently (#62)

fix(eval): eval-015/eval-038 referenced nonexistent card names (#61)

chore(cache): warn at startup when Redis cache is disabled (#60)

fix(rag): hard provider decoupled from the main provider — routing was dead in prod (#59)

feat(rag): hard-query routing — thinking model + full-rulebook stuffing (4.2+4.3, flag off) (#58)

feat(retrieval): keyword family completion beyond top_k (3.5, flag-gated) (#57)

feat(ingest): contextual-retrieval machinery + version-scoped chunk ids (#56)

fix(eval): strict rule-code lineage — recall no longer paper-hits by family (#55)

fix(ingest): match family-sentence rule titles, capture first bold only (#54)

feat(config): cap generation output tokens (max_output_tokens=1024) (#51)

feat(prompt): add few-shot chaining examples to system prompt (v6) (#50)

fix(cache): never cache degraded responses (#49)

feat(rag): fine rule-title sections + reserved keyword slot (hard-bucket retrieval) (#48)

feat(eval): replace eval-019/030 with annotated QOTD questions (#47)

feat(config): enable cross-encoder reranker by default (#46)

fix(config): replace retired gemini-2.0-flash default with gemini-flash-lite-latest (#45)

feat(eval): annotate gold rule references and add hard-bucket diagnosis (#44)

feat(rag): add multi-card reasoning scaffold threaded via extra_system (#43)

feat(rag): wire cross-encoder reranker behind enable_reranker (#42)

feat(rag): implement GeminiProvider.hyde() for dual-arm retrieval (#41)

feat(frontend): detect HF Space cold start and auto-retry once awake

docs(env): sync backend .env.example with actual production config

fix(observability): pin langfuse to v2.x

fix(eval): remove erroneous await on synchronous answer_question

fix(generation): retry Gemini 429s to match openai-compat path

fix(pipeline): guard empty Answer section with retry and fallback

Merge pull request #34 from GonzaloAsencio/feat/corpus-version-reresolve

refactor(pipeline): split answer_question into focused helpers

feat(api): re-resolve corpus_version on demand when empty at startup

refactor(db): extract resolve_corpus_version, share with startup

refactor(pipeline): pass corpus_version explicitly, stop mutating settings

fix(pipeline): whole-word match for exact-card confidence

perf(card-detect): cache compiled patterns per vocabulary

perf(retrieval): add trigram index on lower(section)

fix(api): log exception type, not message, for infra errors

feat(ratelimit): allow shared Redis storage for counters

feat(config): require proxy_shared_secret in production

fix(generation): handle empty Gemini response gracefully

fix(api): return 503 with Retry-After on pool exhaustion

feat(config): make DB pool size configurable

fix(db): use ThreadedConnectionPool and discard broken connections

fix(generation): cap LLM retry backoff and add jitter

refactor(api): run /query as sync handler in the threadpool

refactor(pipeline): make answer_question synchronous

refactor(cache): make get_cached/set_cached synchronous

feat(confidence): match exacto de carta = alta confianza