Skip to content

tools and capabilities

Renderflow exposes the runtime provider registry through the same structured model used by planning and diagnostics.

List providers

renderflow tools list
renderflow tools list --format json
renderflow tools list --format yaml

Use --transforms <FILE> to include providers inferred from a transform YAML file, including arbitrary command providers that are not part of the built-in catalog.

Inspect one provider

renderflow tools inspect tool.ffmpeg
renderflow tools inspect tool.pandoc --format json

Inspection includes discovery strategy, live availability state, selected executable, installed version evidence, determinism/locality/fidelity metadata, capability IDs, fallbacks, and diagnostics.

List provider variants

renderflow tools variants tool.upscayl-ncnn
renderflow tools variants tool.upscayl-ncnn --models-dir /path/to/models
renderflow tools variants tool.upscayl-ncnn --models-dir /path/to/models --format json

The generic variant surface exposes stable provider-specific model/variant identities. When runtime material is supplied, structured output can include checksums and discovery diagnostics without changing the canonical built-in catalog.

List capabilities

renderflow capabilities
renderflow capabilities --format json

Capability IDs and provider IDs are stable machine-readable identifiers. Human-readable CLI output is rendered from the same data returned by JSON/YAML modes.

Use --matrix to emit the generated artifact capability conformance matrix:

renderflow capabilities --matrix
renderflow capabilities --matrix --format json

The matrix reports implemented, experimental, unavailable, and planned support together with executor, provider, validator, fixture, platform, determinism, and loss-profile evidence.

Toolchain fingerprints

Graph planning fingerprints only providers selected by the final DAG. The fingerprint includes the selected provider IDs, compatible installed versions, relevant executable identity, provider capability metadata, selected variant/model evidence when present, and the target OS/architecture. It does not hash the entire host environment.

Graph execution uses that fingerprint in artifact-cache compatibility and writes the selected toolchain evidence into Renderflow state for reproducibility/provenance consumers.