Show HN: Enhancing GPT with domain knowledge using vector databases

12 points

3 years ago

Autoregressive language models such as GPT and LLaMA have a tendency to hallucinate when asked about out-of-domain questions. An easy way to minimize these hallucinations is to supply LLMs with external sources of domain-specific data using a vector database.

To this end, we built OSSChat, a GPT- and Zilliz-powered project that you can use to ask questions about, summarize, and analyze open source projects by indexing the documentation and automatically searching for relevant documents when asking a question (these documents are then injected into the LLM as prompt).

While we decided to focus on open source projects - Pytorch, Transformers, and Milvus - for now, OSSChat is easily extensible to a variety of other types of data. We are committed to adding support for more projects in the future, so don't hesitate to let us know your favorite project!

site: https://osschat.io discord: https://discord.com/invite/NPGhucKNcz blog: https://zilliz.com/blog/ChatGPT-VectorDB-Prompt-as-code

Enjoy the project!

1 comment

1 comment