Docker

Tablassert publishes a pre-built Docker image to GitHub Container Registry (ghcr.io) for containerized usage without a local Python installation.

Image

The image is based on python:3.14-slim with the Tablassert CLI as the entrypoint. All dependencies are included in the base install.

docker pull ghcr.io/skyeav/tablassert:latest

Version-pinned tags match the git tag (e.g., ghcr.io/skyeav/tablassert:v7.2.2).

Quick Start

# Show CLI help (default CMD)
docker run --rm ghcr.io/skyeav/tablassert:latest

# Check version
docker run --rm ghcr.io/skyeav/tablassert:latest version

Building a Knowledge Graph

The primary CLI command is build-knowledge-graph, which reads a graph configuration YAML file and produces KGX-compliant NDJSON output. Defined in src/tablassert/cli.py:51.

docker run --rm \
  -v /path/to/config:/data \
  -v /path/to/datassert:/datassert \
  ghcr.io/skyeav/tablassert:latest \
  build-knowledge-graph /data/graph-config.yaml

Verifying Table Configuration

The verify-table-configuration-syntax command validates a table configuration YAML against the schema without running a full build. Defined in src/tablassert/cli.py:105.

docker run --rm \
  -v /path/to/config:/data \
  ghcr.io/skyeav/tablassert:latest \
  verify-table-configuration-syntax /data/table-config.yaml

Included Capabilities

All dependencies ship in the base install, so the Docker image includes:

• Quality control — The QC pipeline in src/tablassert/qc.py runs a three-stage audit: exact match, then fuzzy matching via rapidfuzz (threshold >= 20), then BioBERT sentence embeddings with cosine similarity (threshold >= 0.2). The ONNX model is cached in .onnxassert/ (line 26).
• Web downloads — src/tablassert/downloader.py uses Playwright to download remote files with retry logic.
• Legacy Excel — modernize_xls() in src/tablassert/downloader.py converts .xls files using pyexcel.

Persistent Data Directories

Mount these volumes to persist data across container runs:

Directory	Source	Purpose
.storassert/	src/tablassert/utils.py:17 — STORE	Intermediate Parquet storage for compiled subgraphs
.logassert/	src/tablassert/log.py	Loguru log files with 100 MB rotation
.onnxassert/	src/tablassert/qc.py:26 — MODEL	Cached ONNX/BioBERT model

Example:

docker run --rm \
  -v ./config:/data \
  -v ./datassert:/datassert \
  -v ./.storassert:/app/.storassert \
  -v ./.logassert:/app/.logassert \
  -v ./.onnxassert:/app/.onnxassert \
  -w /app \
  ghcr.io/skyeav/tablassert:latest \
  build-knowledge-graph /data/graph-config.yaml

Runtime Considerations

• Datassert path — The graph configuration YAML specifies the datassert path for the entity-resolution database. Ensure it is accessible inside the container.
• Multiprocessing — src/tablassert/cli.py:63 uses multiprocessing.Pool for parallel table loading and section extraction.
• DuckDB connections — An ExitStack at src/tablassert/cli.py:81 opens read-only connections to all 10 Datassert DuckDB shards concurrently.
• Entity resolution — The fullmap module (src/tablassert/fullmap.py) shards terms across 10 DuckDB shards (SHARDS = 10) using xxhash64.
• Text normalization — src/tablassert/nlp.py provides level_one (strip + lowercase) and level_two (regex-based cleanup).

CI/CD Integration

Images are built by .github/workflows/docker.yml, which triggers on tag pushes (after autotag and PyPI publish complete). Tags match the repository version tag (e.g., v7.2.2).