All documentation
Concepts
Understand plugins, presets, and runtime configuration.
Plugins & Presets
Profiles install compatible sets of optional integrations. Runtime presets choose parser, chunker, and reranker defaults per request. The source and vector-store plugin API is experimental while its compatibility and release gates mature.
| pip extra | Docker tag | Use |
|---|---|---|
| profile-slim | :slim | MarkItDown, fast chunking, FlashRank |
| profile-balanced | :balanced | Default — Docling, semantic chunking, hybrid RAG |
| profile-quality | :quality | OCR parsers, semantic chunks, quality rerankers |
| profile-agents | :agents | Balanced + AutoGen agentic RAG |
| profile-mcp | :mcp | Balanced + hosted MCP server dependency |
| profile-gpu | — | MinerU / PaddleOCR on quality base; pip profile only |
| profile-eval | — | RAGAS evaluation; pip profile only |
- GET /profiles reports installed profiles and runtime presets; GET /plugins reports installed integrations and source/vector capabilities.
- The optional Qdrant vector adapter is installed with `docpipe-sdk[qdrant]` and selected through the `vector_store` provider/options envelope.
- The optional S3 source is installed with `docpipe-sdk[s3]`; it resolves `s3://bucket/key` under operator-configured bucket, prefix, endpoint, credential, and size policies.
- S3 integration CI uses SeaweedFS as an S3-compatible service; MinIO-specific interoperability has not been verified by the project.
- Configure S3 restrictions in operator-owned `DOCPIPE_SOURCE_PLUGIN_OPTIONS`; requests cannot replace the allowlists, endpoints, credentials, timeouts, or temporary-file directory.
- Plugin API v1 remains experimental while its compatibility and release gates mature; use the dedicated Qdrant and S3 guides for adapter limits and operator responsibilities.
- Operator guardrails include parser/plugin allowlists and a default runtime preset; request options cannot replace S3 security policy or raw credentials.
S3 operator policy example
export DOCPIPE_SOURCE_PLUGIN_OPTIONS='{"s3":{"allowed_buckets":["reports"],"allowed_prefixes":["published/"],"temporary_root":"/var/lib/docpipe/source-artifacts","access_key_id_ref":{"kind":"environment","name":"DOCPIPE_S3_ACCESS_KEY"},"secret_access_key_ref":{"kind":"environment","name":"DOCPIPE_S3_SECRET_KEY"}}}'
# Create temporary_root with private permissions (0700) before startup.
# Export credential values separately through your secret manager.CLI discovery
docpipe plugins list
docpipe profiles list
docpipe resolve invoice.pdf --goal ingest