monogate-graph 0.1.0

Internal prototype evidence graph toolkit for Monogate artifact workflows.

monogate-graph

Internal prototype evidence graph toolkit for Monogate artifact workflows.

monogate-graph reads small JSON graph files and emits dependency order, cycle warnings, blocked-node summaries, critical-path estimates, action queues, and DOT exports.

The Python import module remains eml_graph for internal compatibility.

This package is not published. It does not certify evidence, prove release readiness, or make public claims.

CLI

monogate-graph validate examples/simple_release_gate.json
monogate-graph summary examples/simple_release_gate.json
monogate-graph topo examples/simple_release_gate.json
monogate-graph blocked examples/simple_release_gate.json
monogate-graph action-queue examples/simple_release_gate.json
monogate-graph export-dot examples/simple_release_gate.json
monogate-graph export-dot examples/kernel_observatory_export.json --out /tmp/observatory.dot

# Compatibility alias:
eml-graph validate examples/simple_release_gate.json

Kernel Observatory Integration

Kernel Observatory can export its cards into the eml-graph JSON shape:

python tools/kernel_observatory/observatory.py --build
python tools/kernel_observatory/observatory.py \
  --export-eml-graph /tmp/monogate_kernel_observatory_v0/observatory_graph.json

PYTHONPATH=src python -m eml_graph.cli validate \
  /tmp/monogate_kernel_observatory_v0/observatory_graph.json
PYTHONPATH=src python -m eml_graph.cli action-queue \
  /tmp/monogate_kernel_observatory_v0/observatory_graph.json

The example examples/kernel_observatory_export.json is a small fixture for the same export shape. It is intentionally much smaller than the live Observatory graph.

Graph Shape

{
  "graph_id": "example",
  "nodes": [
    {
      "id": "tests",
      "label": "Tests",
      "status": "pass",
      "weight": 1,
      "evidence_paths": ["tests/result.json"],
      "risk_flags": []
    }
  ],
  "edges": [
    {"source": "tests", "target": "release", "type": "depends_on"}
  ]
}

Edge direction means target depends on source, so topological order lists prerequisites first.

Boundaries

• no PyPI publish in this sprint;
• no deployment;
• no token or secret handling;
• graph output is planning/evidence structure, not certification.