L88 – An Agentic Local RAG Knowledge Engine (Looking for Feedback & Contributors)

Hi everyone 👋

I recently built L88, an experimental agentic RAG (Retrieval-Augmented Generation) knowledge engine designed to run locally and explore structured AI reasoning pipelines.

The project focuses on combining retrieval systems, LLM orchestration, and modern web architecture into a unified platform for querying and interacting with knowledge.

GitHub Repository:
https://github.com/Hundred-Trillion/L88-Full

---

What the Project Does

L88 allows users to upload and interact with knowledge sources using a multi-step AI reasoning pipeline.

Instead of a simple prompt → response system, the project uses a structured workflow where queries pass through multiple stages before generating an answer.

This makes the system more flexible for experimenting with agentic AI workflows and retrieval-based reasoning.

---

Core Idea

The goal of the project is to experiment with a multi-stage RAG pipeline built with modern tooling.

The system orchestrates different reasoning steps such as routing, query analysis, retrieval, and generation through an organized workflow architecture.

This allows developers to experiment with how AI systems can reason over knowledge rather than relying on a single LLM call.

---

Tech Stack

Frontend

• React
• Vite
• TailwindCSS
• Framer Motion

Backend

• FastAPI
• SQLModel
• JWT Authentication
• SQLite

AI Infrastructure

• LangGraph orchestration
• Ollama for local LLM execution
• FAISS vector search
• BM25 keyword retrieval
• BGE embeddings and reranking

---

What I'm Looking For

I’m sharing this project with the community to get feedback, ideas, and contributions from other developers interested in AI infrastructure and RAG systems.

If you have time to explore the repository, I would really appreciate:

• ⭐ Starring the project if you find it interesting
• 🐞 Reporting bugs or issues you discover
• 💡 Suggesting improvements or new features
• 🔧 Submitting pull requests or fixes
• 🧠 Sharing architectural ideas or optimizations

Even small feedback helps improve the project.

---

Contributing

If you'd like to contribute, feel free to:

• Open an issue
• Suggest improvements
• Submit a pull request

GitHub Repository:
https://github.com/Hundred-Trillion/L88-Full

---

Thanks

Thanks to anyone who takes the time to check out the project and share feedback. Open-source collaboration is what makes projects like this improve and evolve.

Looking forward to hearing your thoughts!