search

jas3333 / gptchat_pinecone

MIT License
14 stars 1 forks source link

readme

App Description:

This app is an interface to OpenAI's gpt-3.5-turbo API with its own set of features.

Breaking Changes:

Made changes to the .env requirements. QUERY_URL and UPSERT_URL are no longer needed. The app now only needs PINECONE_URL=yourindexurl

Features include:

  1. Mulitple personas to choose from. More to come later.
  2. A very large memory for the AI to dig through. My previous version had a limit of around 10 previous conversations, while this version has a potential of a million+ thanks to Pinecone's vector database and OpenAI's embedding model.
  3. Added a search and inject section to manually search vector DB and inject the data into the prompt.
  4. Added textarea to upload text or notes. Format the text in markdown.

Roadmap:

  1. Add the ability to import documents into the databases increasing the knowledge of the bots.
  2. Add ability for the user to use custom personas from the frontend.

If you have a feature you would like to see implemented, open up a discussion.

To install:

git clone the repo:

git clone https://github.com/jas3333/gptchat_pinecone
cd gptchat_pinecone
npm install

To run:

Things You Will Need:

  1. An API key from Pinecone, you can setup a free account here: Pinecone DB Website

Index Setup:

- Create a new Index
- Use a P1
- Dimensions: 1536

Make a note of the index location. You should see something like indexname-2342342a.svc.us-east1-gcp.pinecone.io. You'll need it for the environments in the .env

  1. An OpenAI API key to access their LLMs. This app uses gpt-3.5-turbo and Ada-002 for embedding. OpenAI Their website is horrible. You'll just have to click around until you can find the signup link....

  2. A MongoDB setup.

Setting Up:

Next you'll need to create a .env file, then add the following:

 PORT=5000 # This will change the port the express server uses.
 OPEN_AI_KEY=your-openai-key
 PINECONE_API=your-pinecone-key
 PINECONE_URL=https://indexname-2342342a.svc.us-east1-gcp.pinecone.io # Insert your index location here
 MONGODB=mongodb://localhost:27017/memories # Put in your mongodb connection string here

Running the App:

Once the .env is setup it's time to run the app. In the terminal type node server.js then open a browser and navigate to localhost:5000 or whatever port you set in the .env

Docker:

I created a docker image, the repo can be found here Dockerhub Repo with instructions and a docker-compose example.