tram11/fantasy-football-advisor
render assistant markdown in chat ui (marked + dompurify)
perf: warm up embedding model at boot, cache nfl state, refresh every 72h
fix 413 payload overflow: cap tool result size and chat history window
add data freshness: persist last-refresh stamp, expose via api, show in ui header
add self-refreshing data: in-app scheduler reruns scrape+embed pipeline every 6h
bake sqlite db and chroma vector store into repo via git lfs for deployment
add docker image and hugging face spaces deployment config
add football-themed react chat ui with all 32 team logos and field background
add fastapi chat server: /chat with multi-turn history, /health, /nfl-state
expand eval battery with per-week, schedule, and Phase F questions
populate projections, link news to players, derive matchups, add Vegas lines
add SQL craft tips for variance, recent form, and matchup queries
add per-week stats, schedule, snap counts, season-form chunks
add eval harness with question battery, checks, runner, and cli
harden agent loop: gpt-oss model, rate-limit backoff, tool spam guardrail
add llm agent loop with groq tool-calling and tests
add tool layer for llm agent: 5 async tools (sql, search, players, news, nfl state) plus registry and tests
fix scraper/db/embedding bugs and add pytest test suite
implement client to connect with groq for llm integration
add embedding pipeline scripts and fix macOS compatibility
feat: Add ChromaDB vector store and embedder setup
update project status in readme
add chunker to convert player data into embeddable text chunks
fix full_name not included in model_dump and handle NA injury status
add data pipeline to orchestrate scrapers and database
add ESPN scraper for news and player data
add .env.example template
update README with project overview and setup instructions
add scraper layer with base class and Sleeper API integration
add depth chart and status fields to player model
setup database models and SQLite
Initial commit