Carbon Api

Unified file ingestion API for AI and RAG applications. 40+ connectors covering Google Drive, Notion, Dropbox, OneDrive, Confluence, GitHub, Slack, web scraping, and more. Automatically chunks, embeds, and syncs files into your vector store (Pinecone, Weaviate, Chroma, Qdrant) or Carbon's hosted vector index. Designed for agents and RAG pipelines needing live, synced knowledge bases. Confidence is docs-derived.

REST API: POST /auth/v1/access_token for OAuth connector auth; POST /uploadfile for direct file upload; POST /urls for web page ingestion; GET /embeddings for chunk retrieval; DELETE /deletefile; webhook on file sync completion; integrations (OAuth flows) via Carbon Connect React component or headless API; /list_chunks for paginated chunk listing with metadata.

API key per account (passed as Authorization: Bearer); customer_id scoping enables multi-tenant deployments (each end-user has isolated file space); OAuth tokens for third-party connectors stored and refreshed by Carbon; no raw file content exposed via REST API — only chunks and embeddings; data processed within Carbon's SOC 2 Type II compliant infrastructure.

File sync is asynchronous; webhook delivers file_ready or sync_error events; GET /check_embedding_availability for sync status polling; chunking strategy and embedding model (OpenAI, Cohere, custom) configurable per request; incremental sync detects and re-ingests only changed files; rate limits per plan with Retry-After; delete removes both source file metadata and embeddings.

Documentation covers connector OAuth setup, direct file upload, web scraping configuration, embedding model selection, vector store sync (Pinecone, Weaviate, Chroma, Qdrant), chunk metadata schema, and React Carbon Connect component quickstart. Designed for RAG/agent builders. Confidence is docs-derived.