PromethAI

Open-source framework that gives you AI Agents that help you navigate decision-making, get personalized goals and execute them

Share promethAI Repository

---

We took all the work we did with PromethAI into our new product, cognee -> check it out here

What is it

PromethAI is a Python-based AGI project that recommends choices based on a user's goals and preferences and can modify its recommendations based on user feedback.

Our focus is currently on food, but the system is extendible to any area.

💡 Features

• Optimized for Autonomous Agents
• Personalized for each user
• Introduces decision trees to help user navigate and decide on a solution
• Runs asynchronusly
• For App builds, check out this repo promethAI-GUI
• Supports automating tasks and executing decisions
• Multiple Vector DBs supported trough Langchain
• Low latency
• Easy to use
• Easy to deploy

💻 Demo

🛣 Architecture

🛣 Roadmap

⚙️ Setting up

1. Download the repo using git clone https://github.com/topoteretes/PromethAI-Backend-Backend.git in your terminal or directly from github page in zip format.
2. Navigate to the directory using cd PromethAI-Backend and create a copy of .env.template and name it .env.
3. Enter your unique OpenAI API Key, Google key, Custom search engine ID without any quotes or spaces in .env file. Follow the links below to get your keys:

Keys	Accessing the keys
OpenAI API Key	Sign up and create an API key at OpenAI Developer
Pinecone API Key	Sign up and create an API key at Pinecone.io
Google API key	Create a project in the Google Cloud Console and enable the API you need (for example: Google Custom Search JSON API). Then, create an API key in the "Credentials" section.
Custom search engine ID	Visit Google Programmable Search Engine to create a custom search engine for your application and obtain the search engine ID.

4. Ensure that Docker and Docker Compose are installed in your system, if not, Install it from here.
5. Once you have Docker Desktop running, run command : docker-compose up promethai --build in promethai directory. Open your browser and go to localhost:3000 to see promethAI running.

Resources

Papers like "Generative Agents: Interactive Simulacra of Human Behavior"

Quick start

Make sure to add your credentions in the .env file.Launch the app with:

docker-compose build promethai && docker-compose up promethai

How it Works

Here is what happens everytime the AI is queried by the user:

1. AI vectorizes the query and stores it in a Pinecone Vector Database
2. AI looks inside its memory and finds memories and past queries that are relevant to the current query
3. AI thinks about what action to take
4. AI stores the thought from Step 3
5. Based on the thought from Step 3 and relevant memories from Step 2, AI generates an output
6. AI stores the current query and its answer in its Pinecone vector database memory

How to use

docker-compose build  promethai

6. Access the API by doing CURL requests, example:

curl -X POST "http://0.0.0.0:8000/data-request" -H "Content-Type: application/json" --data-raw 

Example of available endpoint

The available endpoint:

POST request to '/recipe-request' endpoint that takes a JSON payload containing 'user_id', 'session_id', 'factors' keys, and returns a JSON response with a 'response' key.

All endpoints receive a payload in JSON format and return a response in JSON format.

Example of curl requests

curl --location --request POST 'http://0.0.0.0:8000/recipe-request' \
--header 'Content-Type: application/json' \
--data-raw '{
  "payload": {
    "user_id": "659",
    "session_id": "459",
    "model_speed":"slow",
    "prompt":"I would like a healthy chicken meal over 125$"
    
  }
}'

🔰 Notice

PromethAI is a work in progress, delivered to you without any guarantees, whether explicit or implied. By choosing to use this application, you consent to take on any associated risks, including data loss, system failure, or any other complications that may arise.

The creators and contributors of PromethAI disclaim any responsibility or liability for any potential losses, damages, or any other adverse effects resulting from your use of this software. The onus is solely on you for any decisions or actions you take based on the information given by PromethAI.

Please be aware that usage of the GPT-4 language model could incur significant costs due to its token consumption. By using this software, you acknowledge and agree to monitor your own token usage and manage the associated costs. We strongly suggest routinely checking your OpenAI API usage and implementing necessary limits or alerts to avoid unexpected fees.

Given its experimental nature, PromethAI may generate content or perform actions that do not align with real-world business norms or legal obligations. It falls on you to ensure that any actions or decisions based on this software’s output adhere to all relevant laws, regulations, and ethical standards. The creators and contributors of this project will not be held accountable for any fallout from using this software.

By utilizing PromethAI, you agree to protect, defend, and absolve the creators, contributors, and any affiliated parties from any claims, damages, losses, liabilities, costs, and expenses (including reasonable attorneys' fees) that arise from your use of this software or your violation of these terms.

📝 License

MIT License

Credits:

Teenage AGI -> https://github.com/seanpixel/Teenage-AGI Baby AGI -> https://github.com/yoheinakajima/babyagi