mirror of
https://github.com/Imbad0202/academic-research-skills.git
synced 2026-09-14 13:51:17 +08:00
Edward Cheng-I Wu
c5284f629b
The data-layer slice of #182 v3.11 (scope anchored in #299). Token-rendering and policy layers (finalizer matrix, formatter allowlist, citation_existence gate) are deferred to a later batch. Delta 1 (arXiv resolver): - scripts/arxiv_client.py: ArxivClient mirroring the crossref/openalex clients but parsing Atom XML, keyed by arXiv ID, with fixed ~3s pacing. arxiv_id_lookup + title_search + ArxivUnavailable degradation contract. - deep-research/references/arxiv_api_protocol.md: the API protocol doc. - literature_corpus_entry.schema.json: optional contamination_signals.arxiv_unmatched boolean + manual-entry not-rule extended to forbid it (anyOf now 4 lookups). - contamination_signals.py: resolve_arxiv_unmatched + build_signals_object arxiv extension (signal computation only). Delta 2 (persistent verification cache): - scripts/verification_cache.py: SQLite WAL cache keyed by (citation_key, resolver_name, query_form), 90-day TTL, get/put/invalidate. - cache= param threaded through all four resolver wrappers via a shared _cached_verdict helper; cache=None is byte-equivalent to prior behavior. query_form keys the whole attempt (id+title) to avoid caching a title-hit verdict under a bare-id key. Degradation is never cached; a malformed cached payload is treated as a miss. Manual entries never touch the cache. - scripts/ars_cache_invalidate.py + commands/ars-cache-invalidate.md. Tests: arxiv client (incl. version-suffix / URL fall-through / empty-title / XML degradation edge cases), cache round-trip + TTL + WAL, resolver cache integration (hit/miss/degradation/manual/malformed-payload), schema validation + not-rule. Full suite: 1981 passed, 3 skipped. Existing client/migration tests unchanged (byte-equivalence proof). [skip-closes-check] This is a data-layer slice; it does NOT complete or close #182 — the gate/policy deliverables remain. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
41 lines
1.3 KiB
Python
41 lines
1.3 KiB
Python
#!/usr/bin/env python3
|
|
"""/ars-cache-invalidate CLI — drop cached verification entries for a citation.
|
|
|
|
Backs the `/ars-cache-invalidate <citation_key>` slash command (Delta 2,
|
|
deliverable 6). Removes every cached resolver entry (all resolvers, all query
|
|
forms) for the named citation_key so the next pipeline run re-verifies it live.
|
|
Idempotent: invalidating a citation with no cached rows succeeds as a no-op.
|
|
|
|
Spec: docs/design/2026-05-21-v3.10-182-promote-citation-gate-spec.md §2 Delta 2.
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import argparse
|
|
import sys
|
|
|
|
try:
|
|
from verification_cache import VerificationCache
|
|
except ImportError:
|
|
from scripts.verification_cache import VerificationCache
|
|
|
|
|
|
def main(argv: list[str] | None = None) -> int:
|
|
parser = argparse.ArgumentParser(
|
|
prog="ars-cache-invalidate",
|
|
description="Drop cached verification entries for a citation_key.",
|
|
)
|
|
parser.add_argument(
|
|
"citation_key",
|
|
help="The citation_key whose cached resolver entries to invalidate.",
|
|
)
|
|
args = parser.parse_args(argv)
|
|
|
|
cache = VerificationCache()
|
|
cache.invalidate(args.citation_key)
|
|
print(f"[ars-cache-invalidate] cleared cache for '{args.citation_key}'")
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
sys.exit(main())
|