Inspiration

RAG technology sparked an idea: a Discord bot that could not only answer questions but also attribute knowledge to its original authors. This concept aims to recognize community contributions while efficiently addressing repeated queries.

What it does

Tiger Social enhances Discord communities by:

  • Efficiently answering repeated questions
  • Utilizing the community's collective knowledge
  • Providing proper attribution for sourced information

How we built it

Our development process involved:

  1. Creating a Discord Python Bot
  2. Integrating OpenAI's LLM
  3. Implementing Jina AI for embeddings
  4. Utilizing TiDB for vector database storage
  5. Using raw SQL for data and vector storage

Challenges we ran into

  • Adapting to Jina AI embeddings when prototyping with Claude
  • Navigating TiDB's beta vector database capabilities
  • Resolving issues with the Llama-Index TiDB plugin

Accomplishments that we're proud of

  • Successfully integrating TiDB and OpenAI function calling
  • Implementing vector embeddings for the first time
  • Completing the project despite technical hurdles

What we learned

  • Gained experience with vector embeddings
  • Explored TiDB's features, including HTAP capabilities
  • Enhanced problem-solving skills through troubleshooting

What's next for Tiger Social

Future plans include:

  • Implementing support for replies and threads
  • Developing a dashboard for question analytics
  • Expanding attribution features to include analytics

Built With

Share this project:

Updates