Grounded Solana intelligence for Clawd agents

This service is a compact RAG API for the Solana Clawd NVIDIA blueprint stack. It retrieves project context from a FAISS index, optionally reranks and generates with NVIDIA models, and returns answers that are tied back to the local knowledge store.

Store readytrue

Indexed chunks150

Generation modeNVIDIA/NIM

What the RAG looks like

1. SourceSolana docs, Clawd skills, blueprint notes, model-kit material.

2. ChunkMarkdown, JSONL, and PDFs are normalized into retrieval-sized passages.

3. EmbedNVIDIA embeddings when configured, hash fallback for local/dev mode.

4. RetrieveFAISS returns the nearest project-context chunks for the question.

5. AnswerNVIDIA/NIM generation when keyed, otherwise retrieval-only context with source snippets.

Try the API

curl -sS https://solana-clawd-rag.fly.dev/query \
  -H "Content-Type: application/json" \
  -d '{"question":"What does the Solana Clawd NVIDIA RAG pipeline do?","top_k":5}'

For users

Ask questions about the Solana Clawd training stack, NVIDIA blueprints, model-kit flow, local Mac setup, and RAG/deployment contracts.

For agents

Use POST /query as a grounded context endpoint before producing model, trading, DAO, or deployment instructions.