Building from source

Building from source

You need Go 1.26+, Node 22+, and PostgreSQL 16+ to run it. Python 3.10+ for the OCR engines.

git clone https://github.com/anand34577/docveta && cd docveta
cd web && npm ci && npm run build && cd ..     # builds the web app into internal/webui/dist
go build -tags nodynamic -o docveta ./cmd/docveta                  # one binary with the web app inside
./docveta                                        # setup page on http://localhost:8080

The build is pure Go (CGO_ENABLED=0 works). -tags nodynamic keeps the HEIC/AVIF decoders from linking to the system loader (glibc), so the binary also runs on Alpine and in minimal containers. Cross-compiling is one command:

GOOS=windows GOARCH=amd64 go build -tags nodynamic -o docveta.exe ./cmd/docveta
GOOS=linux   GOARCH=arm64 go build -tags nodynamic -o docveta     ./cmd/docveta
GOOS=darwin  GOARCH=arm64 go build -tags nodynamic -o docveta     ./cmd/docveta

Development

# backend with readable logs (http://localhost:8080)
DOCVETA_DEV=true go run ./cmd/docveta serve
# frontend with hot reload (http://localhost:5173, proxies the API to :8080)
cd web && npm run dev

Tests:

make test                       # Go, TypeScript, worker tests
DOCVETA_TEST_DATABASE_URL=postgres://user:pass@localhost:5432/docveta_test go test -race ./internal/app -run Integration

The integration test uses a throwaway schema and validates every API response against internal/api/openapi.yaml.

The OCR engine

cd workers/onnx
pip install "../sdk-python[heic,ppocr]" onnxruntime          # onnxruntime-directml on Windows
pip install "paddlepaddle==3.2.0" "paddle2onnx==2.1.0" onnx pyyaml packaging && python convert.py --out models   # once
python worker.py --list-devices
python worker.py --self-test
pip install pyinstaller && pyinstaller docveta-ocr.spec        # standalone build in dist/docveta-ocr

Windows installer

With Inno Setup 6:

iscc /DAppVersion=0.2.0 /DArch=x64 /DSourceDir=C:\path\to\folder-with-docveta.exe deploy\windows\docveta.iss

The folder needs docveta.exe, LICENSE, README.txt and optionally ocr\.

Releases

Pushing a tag v* runs .github/workflows/release.yml: it builds Docveta for 9 platforms, the OCR engine for Windows/Linux/macOS, the Windows installers and Docker images, and publishes a GitHub release. Actions → Release → Run workflow makes a draft release for testing.

More: CONTRIBUTING.md, worker protocol.