| Hi HN Memovee is an agentic movie database that lets you explore movies using natural language instead of filters or rigid queries. You can ask things like: - “Movies that take place in someone’s mind” - “Top mystery films on Netflix released in the last 5 years” - “Slow-burn sci-fi movies with strong world-building” Under the hood, this isn’t just an LLM wrapper. Memovee uses a structured movie database and an agent layer that translates natural-language intent into deterministic queries and aggregations, then reflects on the results before responding. The agent implementation used by Memovee is open source:
https://github.com/upmaru/memovee-tama This repository shows how intent parsing, query planning, execution, and result refinement are handled step-by-step, rather than relying on opaque prompt chains. The core engine itself is not open-sourced yet. Memovee is one concrete application built on top of a more general engine whose goal is to make it possible to build systems like this for any enterprise domain — not just movies. Engine project:
https://kritama.com This is still early and focused purely on movies (not TV yet). Coverage and regional availability vary, and there’s a lot left to improve — especially around reasoning depth, evals, and edge cases. Happy to answer questions about: - The agent architecture - How natural-language intent is mapped to structured data - Agentic vs deterministic tradeoffs - The path toward a reusable enterprise engine Looking forward to feedback — especially critical ones. |