Project Status: Hobby Project (Work in Progress) This is currently a hobby project I'm working on and it is not yet finished. Here is the current state of things:
- RLM (Recursive Language Model) Memory: Working
- LoopLM Reasoning: Working
- WebMCP / Workflows: Most tools and workflows are working
- DGM (Darwin Gödel Machine): About half done
- DSPy Integration: Planned (Not started)
An experimental platform exploring the next generation of AI agents.
Cognitive Chat is a unified ecosystem integrating MS Agent Framework, LoopLM, Recursive Language Model (RLM) based memory, and Darwin Gödel Machine (DGM). The goal is to create a seamless bridge between advanced reasoning models and a usable chat application interface. We also plan to integrate DSPy in the future to programmatically compile and optimize our agent prompts.
- Triple-Stack Architecture: Deno (Backend) + Angular (Frontend) + Python (Inference), bundled within Docker for easier deployments.
- Advanced Reasoning: Deep integration with LoopLM and recursive language models.
- Self-Improving Agents: Powered by Darwin Gödel Machine for autonomous code generation and optimization (still in development).
- DSPy (Planned): A future integration to automatically optimize prompts and pipelines, reducing the need for manual prompt engineering.
- WebMCP Integration: Seamlessly implements the Model Context Protocol (MCP) over WebSockets. WebMCP allows the platform to connect local and remote AI models with your file system, terminal, and custom APIs without complex boilerplate.
- Long-term Memory: Explores Retrieval-Augmented Generation (RAG) using
PostgreSQL and
pgvector.
Recursive Language Models Scaling Latent Reasoning via Looped Language Models Darwin Gödel Machine
If you want to poke around and try it locally, you can get up and running using Docker:
- Docker Desktop
- If you want to use local models, check out the SETUP_GUIDE.md for instructions on integrating with LM Studio!
# 1. Clone the repo
git clone https://github.com/TykoDev/cognitive-llm-chat.git && cd cognitive-llm-chat
# 2. Setup your environment variables
cp .env.example .env
# 3. Spin up the containers
docker compose --profile full up -d --buildYou can then view the user interface at http://localhost:8000.
If you'd like to read more about how the project is structured or how to run tests, check out the Documentation Center:
Since this is a hobby project, any help or suggestions are welcome! Please read the Contributing Guidelines and Code of Conduct if you're interested.
This project is licensed under the MIT License - see the LICENSE file for details.