RagBucket packages semantic vectors, FAISS indexes, chunks, retrieval memory, and runtime metadata into a single portable `.rag` artifact. Most RAG systems today are tightly coupled to vector DBs and retrieval pipelines. RagBucket makes retrieval memory portable and reusable across projects, environments, and providers. Build once. Query anywhere. Supports: • OpenAI • Cohere • Gemini • Voyage AI • Groq • Anthropic • Local SentenceTransformers

Mantle — The ultimate founder launchpad: Cap tables, equity & 409As

The ultimate founder launchpad: Cap tables, equity & 409As

Hey everyone 👋 I built RagBucket after repeatedly facing the same issue while building RAG systems: the retrieval memory was always trapped inside vector databases, embedding pipelines, and infrastructure setups. ML models are portable: `.pt` `.onnx` `.gguf` But RAG systems usually are not. So RagBucket introduces portable `.rag` artifacts that package: • semantic vectors • FAISS indexes • chunks • retrieval configs • runtime metadata into a reusable file that can be loaded anywhere. One use case I’m especially excited about: building reusable domain-specific retrieval artifacts like: medical.rag finance.rag legal.rag engineering.rag and loading them into different applications without rebuilding embeddings/indexes every time. Would genuinely love feedback from the community 🙌

Mantle — The ultimate founder launchpad: Cap tables, equity & 409As

The ultimate founder launchpad: Cap tables, equity & 409As

RagBucket

Build once. Query anywhere with portable RAG artifacts

Build once. Query anywhere with portable RAG artifacts