ChainForge/GUIDE.md

Installation Guide

The simplest and safest way to install the latest public build of ChainForge is to:

1. Create a new directory and cd into it
2. (Optional, but recommended!) Create a virtual environment. On Mac, you can do

   python -m venv venv
   source venv/bin/activate
   

3. Install chainforge via pip:

   pip install chainforge
   

4. (If you wish to use non-OpenAI models): Make sure you have installed the relevant packages. For more info, see below. You can also come back to do this later.
5. Run:

   chainforge serve
   

6. Open localhost:8000 on a recent version of Google Chrome.

Note

ChainForge alpha is tested on Google Chrome. It currently does not work in earlier versions of Safari. We recommend you open it in Chrome.

Install model APIs

Though you can run Chainforge, you can't do anything with it without the ability to call an LLM. Currently we support OpenAI models GPT3.5 and GPT4, Anthropic model Claudev1, Google PaLM model text-bison-001, and (locally run) Dalai-served Alpaca.7b at port 4000.

To use a specific model, you need to do two things:

1. Install the relevant package to your Python environment (for all non-OpenAI models)
2. Set the relevant API key (for all non-Dalai models)

1. Install packages (Anthropic, Google PaLM, and Dalai-hosted models)

To use Anthropic and Google PaLM models, you need to install the relevant Python package in your Python environment before you can run those models:

• For Anthropic, do pip install anthropic.
• For Google PaLM, do pip install google-generativeai. (Note that PaLM officially supports Python 3.9+, but there's a minor type error that's easily fixed to make it work in Python 3.8.8.)
• For Dalai, install dalai and follow the instructions to download alpaca.7b. When everything is setup, run:

  npx dalai serve 4000
  

2. Set API keys (OpenAI, Anthropic, Google PaLM)

If you're just messing around, we recommend you input the API keys manually via the Settings button in the top-right corner. If you'd prefer to not be bothered every time you load ChainForge, you can set them as environment keys:

• For OpenAI models, you can set an environment variable with your OpenAI key: https://help.openai.com/en/articles/5112595-best-practices-for-api-key-safety . For Mac, for instance, follow:

  echo "export OPENAI_API_KEY='yourkey'" >> ~/.zshrc
  source ~/.zshrc
  echo $OPENAI_API_KEY
  

  Then, reopen your terminal.
• To set Anthropic's API key on Mac, do the same as above but with ANTHROPIC_API_KEY replaced for OpenAI_API_KEY.
• To set Google PaLM's API key on Mac, do the same as above but with PALM_API_KEY replaced for PALM_API_KEY.

---

For developers

Below is a guide to running the alpha version of ChainForge directly, for people who want to modify, develop or extend it. Note that these steps may change in the future.

Install requirements

Before you can run ChainForge, you need to install dependencies. cd into chainforge and run

pip install -r requirements.txt

to install requirements. (Ideally, you will run this in a virtualenv.)

To install Node.js requirements, first make sure you have Node.js installed. Then cd into chainforge/react-server and run:

npm install

Serving ChainForge manually

To serve ChainForge manually, you have two options:

1. Run everything from a single Python script, which requires building the React app to static files, or
2. Serve the React front-end separately from the Flask back-end and take advantage of React hot reloading.

We recommend the former option for end-users, and the latter for developers.

Option 1: Build React app as static files (end-users)

cd into react-server directory and run:

npm run build

Wait a moment while it builds the React app to static files.

Option 2: Serve React front-end with hot reloading (developers)

cd into react-server directory and run the following to serve the React front-end:

npm run start

Serving the backend

Regardless of which option you chose, cd into the root ChainForge directory and run:

python -m chainforge.app serve

Note

You can add the --dummy-responses flag in case you're worried about making calls to OpenAI. This will spoof all LLM responses as random strings, and is great for testing the interface without accidentally spending $$.

This script spins up two servers, the main one on port 8000 and a SocketIO server on port 8001 (used for streaming progress updates).

If you built the React app statically, go to localhost:8000 in a web browser to view the app (ideally in Google Chrome). If you served the React app with hot reloading with npm run start, go to the server address you ran it on (usually localhost:3000).

Problems?

Open an Issue.

Contributing to ChainForge

If you want to contribute, welcome! Please fork this repository and submit a Pull Request with your changes.

If you have access to the main repository, we request that you add a branch dev/<your_first_name> and develop changes from there. When you are ready to push changes, say to address an open Issue, make a Pull Request on the experimental repository and assign the main developer (Ian Arawjo) to it.