# HNMovies

> Cinema and standalone documentaries recommended across 49.3 million Hacker News stories and comments (2007–present), ranked by genuine community endorsements with meme filtering, obscurity discovery, and verbatim HN discussion threads.

## Links
- [HNMovies Screening Room](https://hnmovies.org/): Live interactive catalog and film dossiers (updated weekly)
- [Hugging Face Dataset](https://huggingface.co/datasets/gnumanth/hnmovies): Parquet + JSONL dataset (`gnumanth/hnmovies`, updated weekly)
- [Search API](https://hnmovies.org/api/movies?limit=20): Public JSON query API (`q`, `vibe`, `type`, `max_votes`, `sort`)
- [OpenAPI 3.1 Spec](https://hnmovies.org/openapi.json): Machine-readable API schema
- [MCP Server Card](https://hnmovies.org/.well-known/mcp/server-card.json): SEP-1649 Model Context Protocol metadata
- [A2A Agent Card](https://hnmovies.org/.well-known/agent-card.json): Agent-to-Agent capability manifest
- [Agent Skills Index](https://hnmovies.org/.well-known/agent-skills/index.json): RFC v0.2.0 agent skill manifest
- [Author](https://h3manth.com): Hemanth HM

## Features
- 2,281 curated feature films and standalone documentaries (strictly excluding multi-season TV series, episodes, and video games)
- Multi-stage entity disambiguation resolving collisions like Solaris, Threads, Serenity, Primer, and Brazil
- Meme-filtered endorsement scoring plus 20-year Lindy persistence curves
- Zero-signup local taste recommender powered by HN comment co-mention graphs
- Verbatim spoiler-shielded Hacker News quotes linked directly to original Y Combinator threads
- Level 5 Agent-Native support: Markdown content negotiation (`Accept: text/markdown`), MCP, A2A, WebMCP, RFC 9727 API catalog, and RFC 9421 Web Bot Auth JWKS

## Tech stack
- Vanilla HTML5, CSS3, and ES2022 JavaScript (zero runtime dependencies, zero build step)
- Cloudflare Workers + Static Assets edge runtime
- Python + PyArrow + NumPy weekly incremental ingestion pipeline