informatics-isi-edu / deriva-mcp-core / 24767922384

Builds Branch Commit Type Ran Committer Via Coverage
24767922384 main tests(rag): add TestRagUpdateDocsAsync; wire progress into rag_update_docs_async rag_update_docs_async was missing the task_id_ref pattern and progress_cb wiring that rag_ingest already had, so background tasks showed null progress. Added t... push mikedarcy github
90.01
24766679176 main rag_status: use full key for enriched dataset sources instead of "enriched" push mikedarcy github
89.5
24766267042 main rag_status: aggregate indexed_sources by source name to prevent truncation Per-file compound keys ("source:path") produced unbounded output for large indexes. Aggregate to per-source summaries (chunk_count, file_count, last_indexed_at) so t... push mikedarcy github
89.5
24764329717 main fix: emit pure JSON audit events and suppress mcp per-request log noise Drop the file-based audit log fallback -- stderr is the correct default for containerized and supervised deployments where the runtime collects stdout/stderr. Remove th... push mikedarcy github
89.51
24762166718 main rag: tighten web crawler loop detection and pre-filter before enqueueing Lower repeat threshold from >2 to >1 so paths like /search/search/ are caught immediately. Pre-filter loop URLs before adding to the queue so Chaise nav links cannot c... push mikedarcy github
89.52
24743619944 main add unit test for per-batch progress in rag_ingest_datasets background task push mikedarcy github
89.58
24743432144 main feat: report per-batch progress in rag_ingest_datasets background task `_run_dataset_enricher` now accepts a progress_callback and fires it after each batch of 10 rows, reporting "N/total rows (X%), Y chunks indexed". `rag_ingest_datasets` ... push mikedarcy github
89.58
24586160874 main fix: Align data.py doc_type tag with plugin API default ("catalog-data") push mikedarcy github
89.53
24547966870 main docs: workplan - Per-User Request Correlation (Design Note) push mikedarcy github
89.53
24432753873 main fix: register RFC 9728 PRM endpoint in anonymous + Credenza mode Without auth= on the FastMCP instance the /.well-known/oauth-protected-resource endpoint was never registered, causing OAuth clients to fall back to a path-stripped authorizat... push mikedarcy github
89.53
24416332954 main return structured stats from rag_ingest_datasets task result _run_dataset_enricher now returns {"rows_fetched": N, "failed": N, "chunks": N} instead of None, so the rag_ingest_datasets task result carries self-documenting per-source stats r... push mikedarcy github
89.49
24388177523 main feat: make dataset enricher auto-run opt-in at two levels Dataset indexers registered via ctx.rag_dataset_indexer() no longer run automatically on catalog connect by default. Enrichment is now opt-in: - Plugin level: auto_enrich=False by d... push mikedarcy github
88.74
24225766165 main Phase 7: generalized RAG sources, dataset enrichment, resources, and prompts RAG subsystem (rag/): - WebCrawler + WebSource: BFS async crawler for public websites; rate limiting, domain scoping, path prefix filter, content-hash dedup - ... push mikedarcy github
88.73
23965527999 main Fix access logging by preventing uvicorn from reconfiguring loggers Pass log_config=None to uvicorn.Config so it does not overwrite the uvicorn.access logger setup from _init_access_logging(). Without this, uvicorn's dictConfig call re-ad... push mikedarcy github
90.87
23964869121 main Add access log separation, suppress health probe audit noise Route uvicorn access logs to a dedicated file (RotatingFileHandler) and optional syslog on LOCAL2, separate from app logs. Disable uvicorn's built-in access log to keep container ... push mikedarcy github
90.87