feat(api): expose executor result in NL2SQLResponse and return full result to UI

fix(ui): remove all mock/Spider fallbacks and route queries to real backend only

refactor(ui): clean up Repair tab, sync outputs with backend and add examples

feat(pipeline): enable SQL-only repair + log-only repair for early stages with full traces/metrics

feat(pipeline): enable SQL-only repair + log-only repair for early stages with full traces/metrics

feat(metrics): initialize all counters with zero and extend Prometheus rules for full Grafana coverage

fix(state): enforce TypedDict types and cleanup stale DB removal logic

feat(metrics): enable repair tracking and Prometheus lifecycle for observability

fix(smoke): align smoke_run and smoke_metrics for CI stability and disambiguated queries

fix: resolve SQLite path mismatch and ambiguity false positives

fix(adapter): add fallback lookup for uploaded DBs in /tmp/nl2sql_dbs

fix(tests): disable API key auth in conftest to fix 401 errors

fix(tests): disable API key auth in conftest to fix 401 errors

feat(observability): add refined Prometheus rules and Grafana dashboard polish

fix(sqlite): unify adapter path resolution and load from env for stable pipeline execution

feat(metrics): instrument in-memory NL→SQL cache with hit/miss counters

feat(metrics): add repair_attempts_total and wire into repair loop outcomes

chore(logging): add informative logs to SQLiteAdapter init and execute

refactor(main): remove legacy _load_db_map logic after state module migration

feat(api): add /health endpoint

feat(auth): guard core endpoints with X-API-Key (dev-friendly)

feat(cache): add lightweight in-memory TTL cache for NL→SQL responses

feat(pipeline): auto-derive schema_preview when db_id is provided

feat(api): add /schema endpoint for database introspection

fix(router): align _select_adapter with upload_db via shared DB_MAP + TTL cleanup

fix(docker): minimal image + explicit ENTRYPOINT to run start.py on PORT 7860

fix(docker): disable base ENTRYPOINT and run start.py (enable Gradio UI)

fix(docker): override base ENTRYPOINT to run start.py (enable Gradio UI)

fix(docker): set PORT env and ensure Gradio launches on correct port

fix(demo): ensure Gradio UI launches on PORT alongside FastAPI

chore(debug): list /app contents before CMD

fix(runtime): use start.py entrypoint to keep FastAPI & Gradio alive on Hugging Face

fix(runtime): use unified start.py to keep FastAPI & Gradio alive on HF

fix(docker): run FastAPI (app.main) + Gradio (demo.app) for HF deployment

fix(docker): align Dockerfile with HF runtime (expose 7860 & run FastAPI + Gradio)

feat(docker): run uvicorn and demo app concurrently

fix(docker): hybrid startup for FastAPI + Gradio with HF healthcheck

feat(docker): run FastAPI backend and Gradio UI together with healthcheck

feat(docker): convert to multi-stage build with healthcheck and HF optimization

fix(hf-space): force build from Dockerfile (disable HF python_version override)

fix(hf-space): forward Gradio logs and ensure server binds to 0.0.0.0:7860

fix(hf-space): forward Gradio logs and ensure server binds to 0.0.0.0:7860

fix(demo): ensure Gradio binds to 0.0.0.0:7860 via env + Docker config

fix(demo): bind gradio to 0.0.0.0:7860 for HF Spaces

feat(docker): add full-stack setup with FastAPI + Gradio and healthcheck

chore(docker): update .dockerignore to reduce build size

style(makefile): tidy and reorder sections for clarity

chore(hf): move Space config to README

fix(hf): quote python_version as "3.10" to avoid pulling python:3.1

docs(hf): add Space metadata block for Hugging Face repo card