The latest articles on Forem by Adithya Srivatsa (@hundred-trillion). https://forem.com/hundred-trillion https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3812909%2Fb439fa4a-7f06-4718-961b-9e4228a13bee.jpg https://forem.com/hundred-trillion en Adithya Srivatsa Sun, 08 Mar 2026 12:47:27 +0000 https://forem.com/hundred-trillion/l88-an-agentic-local-rag-knowledge-engine-looking-for-feedback-contributors-178g https://forem.com/hundred-trillion/l88-an-agentic-local-rag-knowledge-engine-looking-for-feedback-contributors-178g <p>Hi everyone 👋</p> <p>I recently built <strong>L88</strong>, an experimental <strong>agentic RAG (Retrieval-Augmented Generation) knowledge engine</strong> designed to run locally and explore structured AI reasoning pipelines.</p> <p>The project focuses on combining retrieval systems, LLM orchestration, and modern web architecture into a unified platform for querying and interacting with knowledge.</p> <p>GitHub Repository:<br> <a href="https://github.com/Hundred-Trillion/L88-Full" rel="noopener noreferrer">https://github.com/Hundred-Trillion/L88-Full</a></p> <h2> What the Project Does </h2> <p>L88 allows users to upload and interact with knowledge sources using a multi-step AI reasoning pipeline.</p> <p>Instead of a simple prompt → response system, the project uses a structured workflow where queries pass through multiple stages before generating an answer.</p> <p>This makes the system more flexible for experimenting with <strong>agentic AI workflows and retrieval-based reasoning.</strong></p> <h2> Core Idea </h2> <p>The goal of the project is to experiment with a <strong>multi-stage RAG pipeline</strong> built with modern tooling.</p> <p>The system orchestrates different reasoning steps such as routing, query analysis, retrieval, and generation through an organized workflow architecture.</p> <p>This allows developers to experiment with how AI systems can reason over knowledge rather than relying on a single LLM call.</p> <h2> Tech Stack </h2> <h3> Frontend </h3> <ul> <li>React</li> <li>Vite</li> <li>TailwindCSS</li> <li>Framer Motion</li> </ul> <h3> Backend </h3> <ul> <li>FastAPI</li> <li>SQLModel</li> <li>JWT Authentication</li> <li>SQLite</li> </ul> <h3> AI Infrastructure </h3> <ul> <li>LangGraph orchestration</li> <li>Ollama for local LLM execution</li> <li>FAISS vector search</li> <li>BM25 keyword retrieval</li> <li>BGE embeddings and reranking</li> </ul> <h2> What I'm Looking For </h2> <p>I’m sharing this project with the community to get <strong>feedback, ideas, and contributions</strong> from other developers interested in AI infrastructure and RAG systems.</p> <p>If you have time to explore the repository, I would really appreciate:</p> <p>• ⭐ Starring the project if you find it interesting<br> • 🐞 Reporting bugs or issues you discover<br> • 💡 Suggesting improvements or new features<br> • 🔧 Submitting pull requests or fixes<br> • 🧠 Sharing architectural ideas or optimizations</p> <p>Even small feedback helps improve the project.</p> <h2> Contributing </h2> <p>If you'd like to contribute, feel free to:</p> <p>• Open an issue<br> • Suggest improvements<br> • Submit a pull request</p> <p>GitHub Repository:<br> <a href="https://github.com/Hundred-Trillion/L88-Full" rel="noopener noreferrer">https://github.com/Hundred-Trillion/L88-Full</a></p> <h2> Thanks </h2> <p>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.</p> <p>Looking forward to hearing your thoughts!</p> agents ai rag showdev