Supercog - your connected co-pilot - is here to help!

Overview

Welcome to the Commmunity home for Supercog.ai.

Go try out the product now, for free: https://app.supercog.ai

You can help find docs at the Wiki or visit the forum for help.

Join our Discord server.

Supercog Developer Edition

This repo will help you get setup to run the Supercog system on your local machine. This is especially useful if you need to:

• Develop new agent tools
• Run tools which require localhost access
• Want to run agents locally with full data controls

Requirements

You will need these core pieces:

1. Docker
2. OpenAI API key

Optional:

3. Google Oauth client to enable Google Auth
4. AWS configuration for Simple Mail Service. Provide keys that are enabled for SES if you want to use the built in "Send Email" tool.

Installation

First, copy env.example to env.base.

Edit env.base to set your OPENAI_API_KEY.

Run setup:

./setup.sh

Run from Docker:

docker compose up

And now open the Dashboard:

http://localhost:3000/register/

Now create an account with an email address to get started.

Using Supercog

When you first login you will see a set of example agents that you can try out. Just click on the agent and start typing into the Prompt box. You can try adjusting the Agent Instructions to see how you can change the agent's behavior.

Now try opening the special Supercog agent. This pre-built agent has instructions for dynamically adding and working with tools. Try a conversation like this:

    you) What tools are available?_

    Here are the tools that are available for use:

    Admin
    REST API (Authorized)
    Send Email (built-in)
    ...

    you) Let's add the Github tool
    ...
      enabling the Github tool
    ...

when you add the Github tool, you will be prompted to enter an access token. You can make one from Github->Settings (look for Developer Settings). Once you have connected Github then you can access your repos, issues, etc...

Here are some other things to try with tools:

• Try adding the Image Recognition tool (just type "add image recognition") then upload an image and ask Supercog to explain what's in it.

• Try using "text to speech" to create audio output. We use this in the "Morning News Reporter" demo agent.

• Use the Database tool to connect to a database you have any, then try text to SQL by asking questions about your data in plain English. Bonus points: add the Charting tool and try creating charts from your data.

Creating new tools

To create new tools, just add tool class definitions in Python into the directory ./supercog/tools. The DummyTool example tool, which provides a single calc_fibonacci function, is placed there as an example.

Checkout the Tool Builder Agent default agent which appears on the home page when you login.

Using LLMs besides those from OpenAI

Using non-free service tools

A number of standard tools require API keys in order to operate. You will need to register with the indicated service to get the API key, then set an Env Variable on the Settings page to store the value.

Examples of tools which require an API key include:

• Tavily search tool
• ScaleSERP Web Search
• Zyte Web screenshots

Most of these services have a free tier you can get started with.

Accessing uploaded files

We use MinIO to store any files you upload locally. (See their Github and Docker Hub). These files will persist in the docker volumne community_minio-data. To view the files and images in MinIO, use their console by going to:

http://localhost:9003

All images will be stored in the images-bucket and all other files will be stored in the files-bucket.