Skip to content

README.md

Latest commit

 

History

History
149 lines (118 loc) · 6.6 KB

README.md

File metadata and controls

149 lines (118 loc) · 6.6 KB
Raw

Documentation Index

Documentation for OpenProcessor: the inference API and the curation subsystem.

Start here

Document What it covers
README.md Overview, feature list, quick start, first project, the cars and wheels example
INSTALLATION.md The one-line installer, every flag, the openprocessor CLI, VLM selection, install from source
VISION_AND_GOALS.md What the project is for, the v0.1.0 scope and the standards the code is held to
SECURITY.md No authentication, LAN exposure, VLM URL policy, reporting
CONTRIBUTING.md Dev setup, tests, contracts, doc checks, commit conventions
CLAUDE.md Orientation for AI coding agents working in the repo

Curation and labeling

Document What it covers
CURATION.md User guide: projects, ingest, region profiles, prompt packs, VLM endpoints, import, combine, export, training
design/curation_api_contract.md Route table and wire models for /curation
design/curation_design_rationale.md Why the subsystem is built the way it is
ARCHITECTURE.md Services, project isolation, data model, the multi-box region cascade, config store, workers
opensearch_schema_design.md Global and per-project index schemas, region_boxes, clustering
../contracts/README.md Generated OpenAPI and TypeScript contracts

The generated schema, contracts/openapi/curation.json, is the source of truth for routes and models.

Inference and models

Document What it covers
OCR.md PP-OCRv5 detection and recognition setup
FACE_RECOGNITION_IMPLEMENTATION.md SCRFD and ArcFace
../export/README.md Exporting models to TensorRT, PE-Core encoders, dual-head detectors
MIGRATION_TRITON_26.md Re-exporting engines for Triton 26.06 and TensorRT 11
PERFORMANCE.md FastAPI and gRPC tuning, profiling
Technical/TRITON_BEST_PRACTICES.md Triton batching, instance groups, tuning
security/triton_cve_hardening.md Triton image CVE posture
../benchmarks/README.md The triton_bench tool

Scripts and tooling

Document What it covers
../scripts/README.md Setup, CLI, release, dataset, example and curation scripts
../ATTRIBUTION.md Third-party code and licenses
../CHANGELOG.md Release history

API surface

All routes are on one port (4603 by default).

Prefix Description
/detect, /faces, /embed, /search, /ingest, /ocr, /analyze, /clusters, /query, /models, /health Inference and visual search; also under /v1
/curation/projects Project registry: list, create, combine
/curation/projects/{project}/... Everything project scoped: classes, crops, regions, clusters, review, scores, VLM, prompt packs, region profiles, settings, keymap, datasets, reprocess, ingest, export, training
/curation/vlm/... Deployment-wide VLM endpoint registry, catalog and local model selection
/curation/events Global event stream

Repository layout

OpenProcessor/
├── README.md, INSTALLATION.md, SECURITY.md, CONTRIBUTING.md, CLAUDE.md
├── openprocessor              # management CLI (bash)
├── setup-openprocessor.sh     # one-line installer
├── docker-compose.yml         # deploy-safe stack
├── docker-compose.dev.yml     # checkout overlay: local builds, hot reload
├── docker-compose.gpu-arbiter.yml  # opt-in overlay
├── env.template               # every setting
├── Makefile
│
├── src/
│   ├── main.py                # FastAPI app
│   ├── routers/               # core routers; routers/curation/ for /curation
│   ├── services/              # curation/, config_store/, projects/, labeling/,
│   │                          #   detection/, training/ and the core services
│   ├── clients/               # Triton, OpenSearch, OCC helpers, PE encoder
│   ├── config/                # CurationConfig, RegionFields, projects, retired env
│   └── schemas/               # Pydantic models
│
├── scripts/                   # setup, lib/, release/, codegen/, datasets/, docs/,
│   └── curation/              #   examples/; curation/ has workers and tools
├── models/                    # Triton model repository
├── export/                    # model export scripts
├── docker/                    # segmenter, trainer, evaluator, test harness
├── examples/                  # region profiles, prompt packs, bake-off, VLM catalog
├── contracts/                 # generated API contracts
├── benchmarks/                # triton_bench (Go)
├── docs/, docs-site/          # documentation, Docusaurus site
├── monitoring/                # Prometheus, Grafana, Loki configuration
└── tests/                     # offline suite; tests/live needs the harness

Common tasks

make up                    # start the core stack
make curation-up           # start the curation workers
curl http://localhost:4603/health
make test                  # offline pytest suite
make contracts             # regenerate contracts/ after an API change
.venv/bin/python scripts/docs/check_docs_vs_code.py   # check docs against code

Model exports:

make export-models         # YOLO TensorRT
make export-mobileclip     # MobileCLIP encoders
make setup-face-pipeline   # SCRFD + ArcFace
make setup-ocr             # PP-OCRv5
make export-pe             # PE-Core encoders (curation)

Check model status:

make models-list           # or: ./openprocessor models

External resources