Configuration builder

Pick a component for each stage to get a .ragrc.json for the rag-pipeline CLI. The file is the same one rag-pipeline init would write for that selection. The test suite checks this for every combination, and validates each one against .ragrc.schema.json.

Leave an option blank to use the connector's default, shown as the placeholder. API keys are read from environment variables and never go in the file.

Embedder

localTF-IDF, offline, no API keyopenaiOPENAI_API_KEYollamalocal Ollama servercohereCOHERE_API_KEY

Retriever

file-storeJSON index on diskmemoryin-process only

Index file (optional)

Reranker

nonescoringBM25 re-scoring, offline

Language model

nonequery prints the passagesopenaiOPENAI_API_KEYanthropicANTHROPIC_API_KEYollamalocal Ollama server

Valid .ragrc.json

.ragrc.json

{
  "$schema": "https://raw.githubusercontent.com/DevilsDev/rag-pipeline-utils/main/.ragrc.schema.json",
  "namespace": "default",
  "metadata": {
    "name": "rag-pipeline",
    "createdAt": ""
  },
  "pipeline": [
    {
      "stage": "loader",
      "name": "auto"
    },
    {
      "stage": "embedder",
      "name": "local"
    },
    {
      "stage": "retriever",
      "name": "file-store",
      "config": {
        "path": ".rag/index.json"
      }
    }
  ]
}

Next steps

Same as rag-pipeline init --template local.

npx rag-pipeline doctor
npx rag-pipeline ingest ./docs
npx rag-pipeline query "What is the remote work policy?"

Using the file

• Save it as .ragrc.json in the directory you run the CLI from, or pass --config <path>.
• rag-pipeline doctor checks the file, your API keys and the index before you ingest.
• The "$schema" line gives editors such as VS Code autocompletion and validation. The CLI ignores it.
• Custom plugins (your own .js modules) are added with rag-pipeline init or by hand; see Configuration.