Inspiration:
ChatGPT took the world by storm, but it is very limited with its context window (4K) and you need to break the text down into pieces of text to make it fit. Anthropic's Claude just got released a few days ago and now we built a "ChatGPT" for it, where you can pass massive documents (100,000 tokens) and ask any questions about them.
What it does:
You can chat with very large documents of text by uploading them into ChatClaude, including very large books, PDFs, Excel, Word, and CSV documents.
How we built it:
We built a Chat memory system, with an integration of an advanced PDF OCR recognition AI model for recognition of formulas, LaTeX, images, and charts, a backend with Python, Render, and a front-end with Streamlit with a document management system.
Accomplishments:
We're proud of the very high-quality results of both the AI image model and Anthropic's results.
What we learned:
Generally speaking, context and data retrieval in context-window work better than database data retrieval (semantic search with vector embedding databases)
What's next for ChatClaude:
Release coming soon.
Built With
- anthropic
- claude
- python
- render
- streamlit
Log in or sign up for Devpost to join the conversation.