feat: add llm_endpoint, llm_api_key, llm_model to QueryRequest

feat: BYO-LLM pipeline — gap detection + Gemini cleaner + their LLM answers

fix: catch all pipeline errors, add cost to response

fix: robust token tracking, retry, Jina fallback, no more 500s

fix: add cost field to QueryResponse

fix: 503 retry + token tracking + cost calc per response

upgrade: gemini-3.1-flash-lite-preview + jina X-Engine=direct timeout=30s

perf: gemma-3-27b-it → gemini-2.5-flash-lite (3-5x faster, 1000 RPD free)

fix: jina web search with Bearer auth header

fix: full fresh acra.py with Jina web search, no duckduckgo

web search: switch to jina s.jina.ai — no rate limit issues

web search: jina ai — no key needed, works from cloud

fix: real web search via duckduckgo-search, use_web skips local entirely

add duckduckgo-search for real web search

fix: use_web=True bypasses L0 early return so web search always runs

fix: use_web=True skips local entirely — pure web results

fix: use_web=True filters local docs <0.8 similarity to prevent noise

fix: auth PGRST116 — replace .single() with .execute() + check data

perf: batch compress — N Gemma calls → 1 call, L2 17s→5s L3 43s→12s

fix: L0 always retrieves then lets Gemma decide context vs knowledge

fix: L0 checks docs first (similarity>0.75) before model fallback

fix: QueryResponse includes complexity + retrieval_source + use_web

fix: add complexity + retrieval_source + use_web to QueryResponse

fix: supabase 2.15 + drop pinned httpx (let pip resolve)

fix: resolve httpx conflict — supabase 2.10 + httpx>=0.28.1

fix: acra.py — switch to google-genai + gemini-embedding-001

fix: requirements.txt — switch to google-genai + gemini-embedding-001

fix: upgrade google-generativeai to 0.8.3 for v1 API

deploy: README.md

deploy: Dockerfile

deploy: requirements.txt

deploy: db.py

deploy: credits.py

deploy: auth.py

deploy: acra.py

deploy: classifier_inference.py

deploy: app.py

initial commit