Document ingestion for AI builders
Prepare your knowledge for retrieval.
Choose from 12 source types, 6 vector stores, and 4 embedding providers. Configure the services you need for each project.
Connect sources or bring your files
Browse supported connected sources, select documents, upload files, or crawl a website. Connection methods and monitoring support vary by source.
Sync documents and files from selected folders.
- OAuth connection
- Folder browsing
Folder monitoring supported
Setup guide for Google DrivePrepare selected issues for search and retrieval.
- OAuth connection
- Project browsing and JQL search
Select and sync files from Dropbox folders.
- OAuth connection
- File and folder browsing
Folder monitoring supported
Setup guide for DropboxLoad accessible workspace pages and database content.
- OAuth connection
- Page selection
Load files from S3 or compatible object storage.
- Access key connection
- Bucket and prefix browsing
Folder monitoring supported
Setup guide for Amazon S3Load files from storage buckets, rather than database tables.
- Your Supabase project URL and API key
- Bucket and folder browsing
Folder monitoring supported
Setup guide for Supabase StorageSync selected repository files, issues, and discussions.
- OAuth connection
- Repository browsing
Load documents from storage containers.
- Connection string
- Container and directory browsing
Folder monitoring supported
Setup guide for Azure Blob StorageLoad documents from cloud storage buckets.
- Service account connection
- Bucket and directory browsing
Folder monitoring supported
Setup guide for Google Cloud StorageCrawl accessible pages on a selected website.
- URL-based setup
- Depth control and SSRF protection
Upload local documents for processing.
- File selection or drag and drop
- Uploaded files use billable storage
Store vectors in your configured database
Provide a reachable endpoint and the required credentials. Self-hosted services must be reachable from the hosted app; a server on your laptop is not automatically accessible.
Store vectors in a configured Supabase PostgreSQL project.
- Project URL and service key required
- External projects may require one-time SQL setup
Connect a public PostgreSQL database with pgvector over TLS.
- Encrypted project credential
- Tenant-scoped table and HNSW initialization
Use a managed vector index with project namespaces.
- API key required
- Index initialization
Use a globally distributed Vectorize index for semantic retrieval.
- Scoped API token
- Project namespaces and metadata isolation
Use a Qdrant collection on a server reachable by the app.
- Configure an API key for the sync workflow
- Collection initialization
Use Milvus or a compatible Zilliz Cloud endpoint.
- Configure a token for the sync workflow
- Collection initialization
Choose an embedding provider
Provider charges are separate. Changing models or vector dimensions can require reprocessing documents and configuring a compatible destination.
Choose an embedding model for your project.
- text-embedding-3-small / large
- Your provider API key
Choose English or multilingual embeddings.
- embed-english-v3.0 / embed-multilingual-v3.0
- Your provider API key
Configure a supported Gemini embedding model.
- gemini-embedding-001 and other configured models
- Your provider API key
Use embeddings from your own reachable Ollama service.
- Configurable model and dimensions
- Server must be reachable from the hosted app