docs: Stage 18 — free-tier stability config + live/blocked eval status

fix: disable HyDE + MQ for free-tier stability; restore app to live

feat: maintenance banner as full-screen overlay

feat: enable maintenance banner

fix: contextual retrieval latency — max_chunks gate + prompt trim + model update

fix: hide contextualizing stage message from upload UI

docs: Stage 17 metadata filtering — evolution, decisions, architecture

fix: deselect deleted doc from filter in Sidebar handleDelete

feat: sidebar doc filter chips scope retrieval to selected documents

feat: chat route accepts filter_docs for scoped retrieval

feat: add filter_docs to HybridRetriever and get_retriever_filtered helper

fix: revert out-of-scope model default change in contextualize_chunks

feat: add source_type metadata to all ingested chunks

fix: restore score > 0 threshold in filtered BM25 path + use original index for scoring

feat: add filter_sources param to BM25Index.search()

docs: add metadata filtering implementation plan

docs: add metadata filtering design spec (Stage 17)

docs: update architecture.md to reflect current shipped state (v1.3.0 stack)

docs: audit learning.md — remove 6 non-GenAI concepts, update stale data, add Concept 21 contextual retrieval

docs: add Concept 20 — semantic chunking recall vs precision tradeoff

docs: update evolution, decisions for Stage 16 citation highlighting

fix: citation toggle stopPropagation + absolute UPLOAD_DIR path

feat: wire citation popover — click [N] in answer to view source chunk

feat: add CitationPopover component for citation highlighting

fix: show full chunk content in SourceExpander (remove 200-char truncation)

fix: replace startswith with is_relative_to() for path traversal security

feat: add GET /api/files/{filename} route for citation PDF links

docs: add citation highlighting implementation plan

docs: add citation highlighting design spec

feat: add v1.2.0 (MQ) and v1.3.0 (CTX) eval runs to dashboard

feat: HyDE re-eval (recall +21pp), briefing JSON fix, eval run cleanup

docs: update evolution.md with streaming responses (Stage 14)

fix: ensure setLoading(false) runs in finally after streamChat

feat: stream tokens into chat UI via SSE

feat: add streamChat SSE client to api.js

fix: remove web_search field, ensure cleanup runs on all paths in chat route

feat: stream POST /api/chat via SSE StreamingResponse

fix: remove unused imports in test_streaming.py

feat: add stream_query_with_web async generator to chain.py

docs: streaming responses implementation plan

docs: streaming responses design spec

eval: v1.3.0 clean run (25 samples) — recall=0.627, relevancy=0.837; fix eval script max_tokens+print

fix: eval script — max_tokens 500→1024, max_workers 2→1, timeout 120→180 to reduce RAGAS failures

eval: v1.4.0 semantic chunking — recall -9.8pp, reverted to fixed splits

feat: semantic chunking via SemanticChunker, toggled by config.yaml semantic_enabled

fix: strip trailing commas from LLM JSON in briefing generation

docs: log Stage 13 async embed upload

fix: guard against undefined documents in upload poll callback

feat: async embed upload — return 202 immediately, poll status for stage labels

docs: async embed upload design spec