feat(graph): two-tier L3 alignment thresholds + reconcile co-mention hint — same-type 0.85, cross-type 0.92

feat(graph): L3 semantic entity alignment via bge embeddings

feat(graph): cross-chunk relation reconciliation — top-K co-occurrence LLM pair check

feat(graph): map-reduce KG extraction — per-chunk two-pass under key-pool concurrency

feat(groq): round-robin key pool + fallthrough on 429/413

fix(graph): drop extraction max_tokens 8192→5120 — Groq free-tier 8K TPM ceiling

feat(graph): entity alignment — normalize + fuzzy alias + relation predicates

fix(graph): entity-dense docs — bump max_tokens 4096→8192, tighten entity description spec

fix(graph): unstarve KG extraction — reasoning_effort=medium + bump tokens + domain-neutral prompt

fix(agents): unstarve gpt-oss extraction — reasoning_effort=low + max_tokens bump

deploy: retrieval v2 + verifier tightening

deploy: strict ASCII citation format

deploy: accept .docx and .pptx uploads

deploy: OCR timeout + KG auto-retry

deploy: chat model picker + gpt-oss-120b default

deploy: poll_delay=7.5s

deploy: hybrid retrieval + reranker

fix: worker embedding load retries on HF transient errors

chore: sync backend code (no worker-path changes)

feat: video ingest (cv2 + ffmpeg)

auth: first/last name model fields

qdrant: payload indexes via shared ensure_collection

phase 3

deploy phase 2

initial worker deploy