Skip to content

Knowledge Base RAG

FieldValue
Nameknowledge-base-rag
Categoryautomation
Complexityadvanced
Tagsrag, embeddings, pgvector, google-drive, document-processing, llm-generation, nats-events, s3-storage
Authorrandybias
Min Version0.1.0

Ingest documents from Google Drive, generate embeddings, store in pgvector, and answer questions with source citations. The ingest pipeline runs on a daily schedule; the query path is an independent entry point triggered manually. Uses OpenAI for embeddings and Claude for answer generation.

Ingest pipeline:
poll-drive → store-originals → extract-and-chunk → generate-embeddings → store-vectors
Query path (independent):
answer-query
NodePurpose
poll-drivePoll Google Drive folders for new or updated documents
store-originalsStore original documents in S3
extract-and-chunkExtract text and split into chunks
generate-embeddingsGenerate vector embeddings via OpenAI API
store-vectorsStore embeddings in pgvector
answer-queryRetrieve relevant chunks, generate answer with citations
  • manual
  • cron — daily at 6:00 AM (0 6 * * *)
ServiceTypeRequired
Google Drive APIExternalYes
OpenAI APIExternalYes (embeddings)
Anthropic APIExternalYes (answer generation)
Slack webhookExternalOptional
tentacular-postgresExoskeletonYes (with pgvector extension)
tentacular-rustfsExoskeletonYes
tentacular-natsExoskeletonOptional
KeyDefaultDescription
timeout300sPer-node timeout
retries1Retry count per node
drive_folder_ids(placeholder)Google Drive folder IDs to ingest
query(empty)Question to answer (for query path)
top_k5Number of nearest chunks to retrieve
  • google.access_token — Google API access token for Drive
  • openai.api_key — OpenAI API key for embedding generation
  • anthropic.api_key — Claude API key for answer generation
  • slack.webhook_url — Slack webhook for notifications (optional)
Terminal window
tntc scaffold init knowledge-base-rag
tntc scaffold init knowledge-base-rag my-custom-name
tntc scaffold info knowledge-base-rag

Scaffold source: quickstarts/knowledge-base-rag/