fix: batch-3 adversarial review findings

fix: address 5 review issues in dashboard frontend

fix: stream stage events live, thread source_chunks, fix LangChain wrapper

fix: address remaining review issues (3/5)

feat: orchestrator.run_stream emits per-stage SSE events

fix: handle max_iterations=0 in run_stream to avoid UnboundLocalError

fix: eliminate redundant stream_complete() call in run_stream

fix: propagate LLM cost from run_stream to /ask/stream metrics

fix: /ask/stream parity with /ask, eliminate double LLM call

feat: add SQLite conversation sessions with session_id

feat: add streaming responses via SSE endpoint

fix: retrieval metrics use ranked sources, LLM judge wired, report complete

fix: cross-request state bleed, metadata accuracy, empty question validation

feat: Day 6 — serving layer with FastAPI app, routes, middleware

fix: wire top_k/strategy through orchestrator to SearchTool, add RAG integration test

feat: Day 5 — agent orchestrator with tool-use loop