fix: resolve workflow linking on resume, update proposals auto-commit, and validation baseline rules

feat: add quick demo login, fix archived knowledge filter and workflow completion state

perf: optimize login transition and dashboard loading performance

fix: eliminate N+1 queries on knowledge list/search endpoints

fix: auto-commit used workspace_id as committed_by (FK violation)

fix: activity filter shows all event types per doc; knowledge filter persists on back

fix: concurrency lock, full stage instrumentation, submission docs

fix forgot password

fix signup

mcp update

feat: use email OTP for password recovery

Remove demo mode from signup OTP verification flow

feat: proposal archive/restore governance, knowledge derived status, documents live polling, yellow archive UI

fix: harden knowledge governance and reconciliation

fix: complete knowledge lifecycle and activity pagination

feat: improve knowledge extraction and wire knowledge labels

fix: restore knowledge extraction and token tracking

feat: multi-provider LLM (Ollama local + HuggingFace deployed + Groq), fix think-tag parsing, 15K batch

fix: 5K char batch — Groq free tier strict payload limit

restore: gpt-oss-120b model, 30K batch size, proper system prompt

debug: reduce max_tokens to 2048, add request size logging

fix: minimal prompt + 3K batch for Groq limits, optimistic delete in UI

fix: switch to gpt-oss-20b (faster, larger limits), reduce batch to 8K

fix: reduce batch size to 20K chars for Groq free tier TPM limits

docs: comprehensive README, fix delete button (204), retry on any status

fix: resolve LLM context limits and tool-calling errors using Qwen JSON batching

feat: batch knowledge extraction — processes all sections sequentially, no data loss

refactor: knowledge extraction uses native JSON mode, no tool calling, 50K char buffer

fix: reduce prompt size aggressively (6K cap, 4K retry, 3K fallback) for Groq limits

feat: auto-detect document type from file extension

fix: cap knowledge extraction prompt to 12K chars, reduce retry to 3 sections

feat: switch email service to Resend HTTP API (no SMTP port needed)

fix: use SMTP_SSL port 465 with timeout to prevent hanging on HF

feat: real email OTP via SMTP, forgot password email, Settings on mobile nav

fix: remove unsupported JsonPlusSerializer kwarg for deployed langgraph version

feat: serve frontend from FastAPI in Docker, auto same-origin in prod

restore: revert deployment hacks, back to clean local-first state with Docker for HF Pro

feat: HF Pro Docker deployment - full stack in one container

fix: lazy-load torch/sentence-transformers to avoid OOM on 512MB

fix: add app/storage module (was excluded by gitignore)

chore: switch deployment target to Render, skip model preload

fix: remove docling (websockets conflict with gradio), bump to gradio 5

feat: HuggingFace Gradio Space deployment (bypasses Docker paywall)

chore: deployment configs - Dockerfile (HF Spaces), vercel.json, pinned deps

fix: use started_at instead of created_at on WorkflowRun

feat: retry failed docs, graceful shutdown, optimistic approve, timeline fix, dashboard latest workflow

fix: auto-approve handles empty decision + debug logging in complete node

feat: auto-resume orphaned workflows on server restart + cancel button

feat: topic_relevance validation rule using hybrid retrieval

feat: live stage progression during workflow execution