cjm-capability-primitives

The dep-light primitives library for the cjm capability ecosystem — the shared cross-task result DTOs (data nouns) that tool capabilities emit and task adapters, workflow cores, and the composition layer all consume.

Modules

• cjm_capability_primitives.forced_alignment — Standardized word-level forced-alignment DTOs — the data noun forced-alignment tool capabilities emit and task adapters / workflow cores consume, wire-registered so results cross the worker boundary typed.
• cjm_capability_primitives.media_processing — Standardized result DTOs for the media-processing task — the data nouns media-processing tool capabilities (ffmpeg today) emit and the multi-method task adapter / workflow cores consume, wire-registered so results cross the worker boundary typed.
• cjm_capability_primitives.monitoring — Standardized telemetry DTOs for the system-monitor capability — the data nouns a monitor tool capability emits (host CPU/RAM + aggregated GPU stats, and per-process GPU usage) and the substrate scheduler consumes for resource-derived admission + GPU subtree attribution.
• cjm_capability_primitives.sentence_segmentation — Standardized result DTO for the sentence-segmentation task — the data noun sentence-segmentation tool capabilities emit and task adapters / workflow cores consume, wire-registered so results cross the worker boundary typed.
• cjm_capability_primitives.source_separation — Standardized result DTO for the source-separation (audio-preprocessing) task — the data noun source-separation tool capabilities emit and task adapters / workflow cores consume, wire-registered so results cross the worker boundary typed.
• cjm_capability_primitives.speaker_diarization — Standardized result DTO for the speaker-diarization task — the data noun diarization tool capabilities emit and task adapters / workflow cores consume, wire-registered so results cross the worker boundary typed. Turns are ANONYMOUS clusters (SPEAKER_00, ...) — mapping clusters to named identities is the correction TUI's assignment lane (human judgment), never this DTO's job (DEC 44afb2df).
• cjm_capability_primitives.transcription — Standardized result DTO for the transcription task — the data noun tool capabilities emit and task adapters / workflow cores consume, wire-registered so results cross the worker boundary typed.
• cjm_capability_primitives.vad — Standardized result DTO for the voice-activity-detection task — the data noun VAD tool capabilities emit and task adapters / workflow cores consume, wire-registered so results cross the worker boundary typed.

API

cjm_capability_primitives.forced_alignment

• ForcedAlignItem class — A single word-level alignment result.
• ForcedAlignResult class — Standardized output for all forced alignment capabilities.

cjm_capability_primitives.media_processing

• MediaArtifactResult class — A single produced audio artifact (the convert / extract_audio output).
• MediaMetadata class — Probed metadata for a media file (the get_info result) — inline data, no artifact.
• MediaSegment class — One produced segment file from a segment_audio batch cut.
• MediaSegmentationResult class — A BATCH of produced segment files (the segment_audio output).

cjm_capability_primitives.monitoring

• MonitorToolProtocol class — The native surface a system-monitor tool capability exposes.
• ProcessStats class — Per-process GPU usage, reported by a monitor's list_processes (GPU subtree attribution).
• SystemStats class — Standardized snapshot of host + GPU resources (the scheduler's admission input).

cjm_capability_primitives.sentence_segmentation

• SentenceSegmentationResult class — Standardized output for sentence-segmentation capabilities.
• SentenceSpan class — One sentence's character span over the ORIGINAL (punctuated) text —

cjm_capability_primitives.source_separation

• SourceSeparationResult class — Standardized output for source-separation (audio-preprocessing) capabilities.

cjm_capability_primitives.speaker_diarization

• SpeakerDiarizationResult class — Standardized output for speaker-diarization capabilities.
• SpeakerTurn class — One diarized speech turn: a source-coordinate time span bound to an
• load_turns_artifact function — Load a source's turns artifact; None = no diarization predictions yet.
• turns_artifact_path function — Source-keyed artifact path: <root>/diarization/<algo>-<hexdigest>.json.
• write_turns_artifact function — Persist the source-keyed turns artifact (latest-wins per source).

cjm_capability_primitives.transcription

• TranscriptionResult class — Standardized output for all transcription plugins.

cjm_capability_primitives.vad

• TimeRange class — A temporal segment within an audio source (the VAD speech/silence span).
• VADResult class — Standardized output for voice-activity-detection capabilities.

Dependencies

Depends on: cjm-substrate Used by: cjm-capability-demucs, cjm-capability-ffmpeg, cjm-capability-monitor-nvidia, cjm-capability-pyannote, cjm-capability-pysbd, cjm-capability-qwen3-forced-aligner, cjm-capability-silero-vad, cjm-capability-voxtral-hf, cjm-capability-whisper, cjm-sentence-segmentation-adapter-interface, cjm-speaker-diarization-adapter-interface, cjm-transcript-correction-tui, cjm-transcript-decomp-core, cjm-transcription-adapter-interface, cjm-transcription-core