Documentation & practical guides
Learn the tools.
Read the evidence.
Start with the CLI Handbook for a complete command-line workflow, or choose a focused guide for your model-file question. Each path explains the evidence and its limits.
11 chapters / Command-line documentation
CLI Handbook
Install, audit, inspect tensors, export evidence and diagrams, compare models, automate checks, and connect local AI tools.
Read the handbook →Complete CLI command & option reference · generated from the executable, with command search and downloadable help.
Choose the question
- Review snapshots, OMOP connections and before/after evaluation evidence
- Evidence IR: common representations and evidence boundaries
- Compatibility explorer: input schemas, common IR and output mappings
- Connect OMOP data, cohorts and model lineage
- Inspect ONNX quantization without running the model
- Inspect mixed GGUF tensor encodings and storage
- Compare two deployment artifacts before release
- Review CUDA and Jetson Orin optimizations with MLflow
Candidate review with external evaluations
The optional DEEPBOM Review preview packages baseline/candidate evidence, fixed criteria, and hash-bound evaluation records. Use the CLI walkthrough and the MLflow or Olive adapter instructions. Imported measurements remain external evidence; the review tool does not execute them.
Common boundary
Static inspection does not establish measured latency, task accuracy, clinical validity, safety, regulatory compliance, or the execution provider and kernel selected at runtime.
Numerical evidence
Weight and activation evidence: optional distributions, explicit execution captures and shared Model IR references.