OpenGPTs

This is an open source effort to create a similar experience to OpenAI's GPTs. It builds upon LangChain, LangServe and LangSmith. OpenGPTs gives you more control, allowing you to configure:

The LLM you use (choose between the 60+ that LangChain offers)
The prompts you use (use LangSmith to debug those)
The tools you give it (choose from LangChain's 100+ tools, or easily write your own)
The vector database you use (choose from LangChain's 60+ vector database integrations)
The retrieval algorithm you use
The chat history database you use

Check out a simple hosted version here

Quickstart

Start the backend

Install requirements

cd backend
pip install -r requirements.txt

Set up persistence layer

The backed by default uses Redis for saving agent configurations and chat message history. In order to you use this, you need to a REDIS_URL variable.

export REDIS_URL=...

Set up vector database

The backend by default also uses Redis as a vector database, although you can easily switch this out to use any of the 50+ vector databases in LangChain. If you are using Redis as a vectorstore, the above environment variable should work (assuming you've enabled redissearch)

Set up language models

By default, this uses OpenAI, but there are also options for Azure OpenAI and Anthropic. If you are using those, you may need to set different environment variables.

export OPENAI_API_KEY="sk-..."

Other language models can be used, and in order to use them you will need to set more environment variables. See the section below on LLMs for how to configure Azure OpenAI, Anthropic, and Amazon Bedrock.

Set up tools By default this uses a lot of tools. Some of these require additional environment variables. You do not need to use any of these tools, and the environment variables are not required to spin up the app (they are only required if that tool is called).

For a full list of environment variables to enable, see the Tools section below.

Set up monitoring

Set up LangSmith. This is optional, but it will help with debugging, logging, monitoring. Sign up at the link above and then set the relevant environment variables

export LANGCHAIN_TRACING_V2="true"
export LANGCHAIN_API_KEY=...

Start the backend server

langchain serve --port=8100

2. Start the frontend

cd frontend
yarn
yarn dev

Navigate to http://localhost:5173/ and enjoy!

Features

As much as possible, we are striving for feature parity with OpenAI.

Sandbox - Provides an environment to import, test, and modify existing chatbots.
- The chatbots used are all in code, so are easily editable
Custom Actions - Define additional functionality for your chatbot using OpenAPI specifications
- Supported by adding tools
Knowledge Files - attach additional files that your chatbot can reference
- Coming soon
Tools - Provides basic tools for web browsing, image creation, etc.
- Basic DuckDuckGo and PythonREPL tools enabled by default
- Image creation coming soon
Analytics - View and analyze chatbot usage data
- Use LangSmith for this
Drafts - Save and share drafts of chatbots you're creating
- Supports saving of configurations
Publishing - publicly distribute your completed chatbot
- Can do by deploying via LangServe
Sharing - Set up and manage chatbot sharing
- Can do by deploying via LangServe
Marketplace - Search and deploy chatbots created by other users
- Coming soon

Repo Structure

frontend: Code for the frontend
backend: Code for the backend
- app: LangServe code (for exposing APIs)
- packages: Core logic
  - agent-executor: Runtime for the agent
  - gizmo-agent: Configuration for the agent

Customization

The big appeal of OpenGPTs as compared to using OpenAI directly is that it is more customizable. Specifically, you can choose which language models to use as well as more easily add custom tools. You can also use the underlying APIs directly and build a custom UI yourself should you choose.

LLMs

You can choose between different LLMs to use. This takes advantage of LangChain's many integrations. It is important to note that depending on which LLM you use, you may need to change how you are prompting it.

We have expose four agent types by default:

"GPT 3.5 Turbo"
"GPT 4"
"Azure OpenAI"
"Claude 2"

We will work to add more when we have confidence they can work well.

If you want to add your own LLM or agent configuration, or want to edit the existing ones, you can find them in backend/packages/gizmo-agent/gizmo_agent/agent_types

Claude 2

If using Claude 2, you will need to set the following environment variable:

export ANTHROPIC_API_KEY=sk-...

Azure OpenAI

If using Azure OpenAI, you will need to set the following environment variables:

export AZURE_OPENAI_API_BASE=...
export AZURE_OPENAI_API_VERSION=...
export AZURE_OPENAI_API_KEY=...
export AZURE_OPENAI_DEPLOYMENT_NAME=...

Amazon Bedrock

If using Amazon Bedrock, you either have valid credentials in ~/.aws/credentials or set the following environment variables:

export AWS_ACCESS_KEY_ID=...
export AWS_SECRET_ACCESS_KEY=...

Tools

One of the big benefits of having this be open source is that you can more easily add tools (directly in Python).

In practice, most teams we see define their own tools. This is easy to do within LangChain. See this guide for details on how to best do this.

If you want to use some preconfigured tools, these include:

DuckDuckGo Search

Search the web with DuckDuckGo. Does not require any API keys.

Tavily Search

Uses the Tavily search engine. Requires setting an environment variable:

export TAVILY_API_KEY=tvly-...

Tavily Search (Answer Only)

Uses the Tavily search engine. This returns only the answer, no supporting evidence. Good when you need a short response (small context windows). Requires setting an environment variable:

export TAVILY_API_KEY=tvly-...

You.com Search

Uses You.com search, optimized responses for LLMs. Requires setting an environment variable:

export YDC_API_KEY=...

SEC Filings (Kay.ai)

Searches through SEC filings using Kay.ai. Requires setting an environment variable:

export KAY_API_KEY=...

Press Releases (Kay.ai)

Searches through press releases using Kay.ai. Requires setting an environment variable:

export KAY_API_KEY=...

Arxiv

Searches Arxiv. Does not require any API keys.

PubMed

Searches PubMed. Does not require any API keys.

Wikipedia

Searches Wikipedia. Does not require any API keys.

Deployment

1. Build the frontend

cd frontend
yarn
yarn build

2. Deploy to Google Cloud Run

You can deploy to GCP Cloud Run using the following command:

First create a .env.gcp.yaml file with the contents from .env.gcp.yaml.example and fill in the values. Then run:

gcloud run deploy opengpts --source . --port 8001 --env-vars-file .env.gcp.yaml --allow-unauthenticated --region us-central1 --min-instances 1

thatguykelly/opengpts