perf(ingest): KI-125 — bump MPS embedding batch 64→128 for bulk re-ingest

fix(admin): KI-122 — clear LLM availability + in-use indicators

refactor(session): KI-118 — rip out session_id concept, name-based profiles only

fix(rag): KI-117 — boot-time cleanup of dangling profile chunks

fix(rag+admin): KI-112 hardening + KI-116 banner + admin copy

fix(indic): KI-110 — repair Hindi pipeline raising on every turn

fix(safety+closer): KI-108 + KI-109 — apply CoT strip to all reply paths + broaden closer regex

docs: KI-101..KI-107 — audit + CLAUDE.md + ADR sync

feat(closer): KI-105 — wire ranked recommendation handoff on closer intent

fix(rag): KI-107 — graceful handling of Chroma get(ids=...) for missing ids

fix(safety): KI-106 — graceful timeout/exception handling on /api/chat

fix(safety): KI-104 — strip chain-of-thought / instruction-echo leakage from replies

fix(fact-find): KI-103 — break no_trailer fallback loop after 2 failed attempts

fix(privacy): KI-102 — scope profile-RAG retrieval to session_id only

fix(orchestrator): KI-101 — drop function-local 'import logging' to fix UnboundLocalError

fix(latency): KI-100 — outer wait_for(25s) around drive_fact_find

fix(latency): KI-099 — bound every per-turn LLM call

fix(admin): KI-096 + KI-097 — public /api/version + password-only admin

fix(safety): KI-095 — 5 defensive guards against profile-field clearing

fix(orchestrator): KI-094 — extractor cannot clear a filled profile field

fix(fact-find): KI-092 — lenient name parser when slot explicitly awaiting name

fix(orchestrator): KI-091 — skip profile_extractor + faithfulness judge on fact-find turns (drop dependent LLM calls on saturated chains)

docs: backend/README — fix stale ADR-027 link on needs_finder + add fact_find_brain row + mark paraphraser deleted

fix(fact-find): KI-090 — lenient FF-block parser; accept JSON tail without <FF> tags

feat(llm-chain): KI-088 — cap NIM outbound concurrency at 2 + serialise probe + drop inner retry

docs: ADR-032 LLM chain architecture reference + KI-080..KI-087 doc sync

feat(llm-chain): KI-087 — NIM-first election preference; non-NIM only when NIM pool empty

feat(llm-chain): KI-085 — proactive credit tracking; election gates by quota not just liveness

feat(llm-chain): KI-084 — cut probe token consumption 50× + 429-demote for 1h + per-phase httpx timeouts

docs: KI-080 — sync CLAUDE.md + ENTERPRISE_AUDIT + new ADR-031 + 11 stale KI-025/50-50 surface sweep

feat(llm-chain): KI-080 — sticky primary election; 1-2 LLM calls per turn instead of 5-6

feat(llm-chain): KI-079 — Groq earlier in FAST_BRAIN_CHAIN + escalate to BRAIN_CHAIN on timeout before canonical fallback

feat(llm-chain): KI-078 — tighten per-link timeout + narrow exception catch + stamp fallback reason for telemetry

feat(profile): KI-077 — Name field + show current captured profile in Build-your-profile panel

fix(fact-find): KI-075 — bump brain timeout 12s → 25s so NIM cold-start doesn't kill the call

fix(fact-find): KI-074 — greedy multi-slot capture in fallback + stricter name parse + strip ** markdown leak

docs: KI-071 — reconcile docs with code reality (Qwen 80B / Nemotron 30B / Mistral 675B as current primaries)

fix(fact-find): KI-072 — canonical fallback now captures user's answer + doc sync (CLAUDE.md / ADR-030 / ENTERPRISE_AUDIT)

feat(fact-find): KI-070 — single-LLM-call fact-find replaces 3-layer template stitching

fix(profile + ux): KI-068 humanize readback / strip markdown / async log; KI-069 fix name-capture false-positive

fix(fact-find): KI-067 — recognize "first policy" / "new to insurance" as zero existing cover

fix(voice): KI-066 — TTS normalizes ₹/L/Cr shorthand to spoken language

feat(profile): KI-063 — shown / selected / rejected policy tracking on profile

feat(profile + voice): KI-059 / KI-060 / KI-061 / KI-062 — opening-turn name capture, looser VAD silence, personalized welcome-back, unique persona ID

feat(ux): KI-056 — natural-language acknowledgers + spouse capture + Q3 paren consistency

refactor + feat: KI-050b/053/054 — rename completeness + eval speed + tax questions

refactor: KI-050 — complete data/ → 40-data/ rename across all Python refs

feat(admin): KI-048 — /api/admin/profiles + /api/admin/performance

refactor: KI-047 — bucket reorg (Option A safe subset)

fix: KI-046 — explicit refusal on adversarial/fanciful out-of-corpus questions