LoginSign Up
← Back to all sparks
R

ragnar

AI-ASSISTANTS
Velocity0.0

RAG in R

github.com/tidyverse/ragnar ↗

ragnar turned its RAG store into an MCP server, so coding agents can search it directly.

rragmcpembeddingsduckdbllm-tooling
◆Current state
ragnar builds retrieval-augmented generation stores in R on DuckDB, handling document chunking, embedding, and hybrid vector plus BM25 retrieval, and registering itself as a tool for ellmer chats. Version 0.3.0 adds mcp_serve_store(), which exposes a store over MCP to local clients such as Codex CLI and Claude Code, alongside Azure AI Foundry and Snowflake Cortex embedding providers. Store version 2, introduced in 0.2.0, brought chunk deoverlapping on retrieval and automatic heading augmentation.
◆Where it's heading
The package keeps widening who can reach a store and how many ways they can query it. Retrieval accepts vectors of queries, the ellmer tool withholds chunks it has already returned so an agent can dig deeper across calls, and now the store is reachable from outside R entirely. Embedding providers are added steadily — LM Studio, then Azure and Snowflake — which keeps the store portable across whoever supplies the vectors. Breaking changes are accepted readily at this stage, including a renamed default tool prefix and a flipped default in ragnar_find_links().
◆Prediction
More MCP surface is the natural next step now that serving exists, since the retrieval tool already has the multi-query and no-repeat behavior that agent-driven search depends on.

◆Recent moves

  1. 6mo ago

    mcp_serve_store() exposes a RagnarStore over MCP

    ⚡ SPARK

    mcp_serve_store() takes the store outside R, letting local MCP clients query it directly. The same release adds Azure and Snowflake embedding providers, vectorized queries, a parallel ingest path via mirai, and an embedding visualizer.

    View source ↗
  2. 0y ago

    Retrieval tool withholds already-returned chunks for deeper search

    The registered retrieval tool stops returning chunks it has already served, so an LLM can search a store progressively across repeated calls. Also adds sitemap.xml parsing with optional HEAD-request link validation and an LM Studio embedding provider.

    View source ↗
  3. 1y ago

    Store version 2 adds chunk deoverlapping and heading augmentation

    Introduces store version 2, adding chunk deoverlapping on retrieval and automatic augmentation with headings, on a changed internal schema with version 1 kept compatible. Also adds retrieval filters, MotherDuck remote stores, and BM25 tuning parameters.

    View source ↗