cortex-plugin-rag-pipeline

Pluggable RAG: ingest PDFs/websites/codebases into vector store, query with hybrid search.

Installation

cortex plugin install marketplace:cortex-plugin-rag-pipeline
cortex plugin install github:CortexPrism/cortex-plugin-rag-pipeline
cortex plugin install ./manifest.json

Tools

rag_ingest

Ingest content into the RAG pipeline.

Parameters:

• source (string, required) — URL, file path, or raw text
• source_type (string, required) — url, file, text
• name (string, optional) — Label for this document
• chunk_size (number, optional, default 1000) — Chunk size
• chunk_overlap (number, optional, default 200) — Overlap between chunks

rag_query

Query the ingested knowledge.

Parameters:

• query (string, required) — Search query
• max_results (number, optional, default 5) — Maximum results
• search_type (string, optional, default "hybrid") — hybrid, vector, keyword
• filter_source (string, optional) — Filter by source ID

rag_list_sources

List all ingested sources.

rag_remove_source

Remove an ingested source.

Parameters:

• source_id (string, required) — Source ID to remove

rag_stats

Get RAG pipeline statistics (total sources, chunks, tokens).

Configuration

{
  "plugins": {
    "cortex-plugin-rag-pipeline": {
      "enabled": true,
      "config": {
        "vectorStore": "memory",
        "qdrantUrl": "",
        "pineconeApiKey": "",
        "defaultChunkSize": 1000,
        "defaultMaxResults": 5
      }
    }
  }
}

Capabilities

• tools — Provides tool implementations
• network:fetch — Fetches content from URLs
• fs:read — Reads files from the filesystem

License

MIT — See LICENSE file