OCP Pipeline

The OCP (Open Common Play) pipeline plugin turns media utterances into playback actions in OVOS. Utterances such as "play metallica", "put on some movie", "pause", or "next" pass through this plugin as an intent pipeline stage. On a match, the plugin searches registered media skills over the message bus and drives playback through the OCP API.

The plugin exposes two entry points under the opm.pipeline group:

ovos-ocp-pipeline-plugin (OCPPipelineMatcher), the main pipeline. It matches play, open, media_stop, next, prev, pause, resume, save_game, and load_game intents at high, medium, or low confidence.
ovos-ocp-pipeline-plugin-legacy (MycroftCPSLegacyPipeline), a bridge for older Mycroft CommonPlay skills, using the play:query / play:query.response / play:start handshake.

OCPPipelineMatcher also classifies media type, such as music, movie, or podcast. It extracts named entities with ahocorasick-ner, using per-language vocabulary files and optional user-supplied entity CSVs (the entity_csvs config option). It tracks player state per session through an OCPPlayerProxy, kept in sync with ovos.common_play.status and track.state bus events.

Architecture

Media classification

Media-type classification, the rich provider-ready descriptive Signals, and the content filter are all delegated to the standalone ovos-media-classifier. The pipeline builds the classifier's two context inputs from its own state — the per-session now-playing PlayerStatus and the skill-registered entities (ner_list) — so relative control follow-ups ("next", "pause", "something else") and entity routing work, and it forwards the classifier's lossless mediavocab.Signals (medium / playback_type / content_genres / content_form / programme_format / variant_kind / accessibility / picture_format) to the MediaProviders alongside the legacy media_type.

With no extra configuration the lean keyword (.voc) backend is used — the zero-ML-dependency floor (all heavy backends OFF, online OFF, adult content blocked). The classifier is configured under the pipeline config (or a nested media_classifier block); every key is optional.

Backend selection

Key	Default	Description
`media_classifier_plugin`	unset	name of an external `opm.media.classifier` entry-point plugin to load
`media_classifier_onnx_model`	unset	path to an opt-in ONNX trained bundle (requires the `[onnx]` extra)
`media_classifier_embedding_router`	unset	path to a learned embedding-router bundle (requires the `[onnx]` extra)
`media_classifier_embedding_router_hybrid`	`true`	run the router as a keyword+router hybrid (`false` = router standalone)

On any failure (missing extra, bad bundle, unknown plugin) the classifier falls back to the keyword backend, so the zero-ML default is always preserved.

Gazetteer / entity library (embedding-router backend only)

Key	Default	Description
`media_classifier_gazetteer`	`true`	inject the bundled offline gazetteer of common real titles so bare titles route without a network call
`media_classifier_gazetteer_size`	classifier default	cap on titles per media type
`media_classifier_entity_library`	unset	`{label: [titles]}` of the user's own media library, injected at runtime (no retraining)

Online metadata layer (embedding-router hybrid only)

Key	Default	Description
`media_classifier_online_metadatarr`	`false`	opt into the online metadatarr last-resort layer (adds latency)
`media_classifier_online_timeout`	`4.0`	per-request timeout, seconds
`media_classifier_online_min_confidence`	`0.5`	minimum confidence to accept an online answer

Content filter (applied at routing)

Blocked content (adult by default) is never routed to providers.

Key	Default	Description
`allow_adult_content`	`false`	top-level convenience flag; `true` lifts the default adult block
`media_content_filter.enabled`	`true`	master switch for the filter
`media_content_filter.blocked_genres`	`["adult"]`	genres to block
`media_content_filter.blocked_media_types`	`[]`	media types to block

Install

pip install ovos-ocp-pipeline-plugin

Usage

OVOS core loads this plugin automatically once installed, through the opm.pipeline entry-point group. Enable it in mycroft.conf under the intents section:

{
  "intents": {
    "ovos-ocp-pipeline-plugin": {
      "entity_csvs": []
    }
  }
}

The plugin reads the pipeline's intents config block, then falls back to a legacy OCP config block for backward compatibility.

Related projects

OpenVoiceOS/ovos-plugin-manager defines the ConfidenceMatcherPipeline base class and the opm.pipeline entry-point group this plugin implements.
OpenVoiceOS/ovos-workshop provides OVOSAbstractApplication, the bus-app base class this plugin also subclasses.
OpenVoiceOS/ovos-bus-client provides the OCP message-bus API (OCPInterface) used to search and control playback.

License

Apache-2.0

Release files for ovos-ocp-pipeline-plugin 1.4.0a1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ovos-ocp-pipeline-plugin 1.4.0a1
File	Size	Uploaded
ovos_ocp_pipeline_plugin-1.4.0a1.tar.gz	193.9 kB	Sep 2, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for ovos-ocp-pipeline-plugin 1.4.0a1
File	Interpreter	ABI	Platform	Reset
ovos_ocp_pipeline_plugin-1.4.0a1-py3-none-any.whl 216.6 kB Sep 2, 2026	Python 3	none	any	Details

Total release size: 410.5 kB

Release files / ovos_ocp_pipeline_plugin-1.4.0a1.tar.gz

Download URL	ovos_ocp_pipeline_plugin-1.4.0a1.tar.gz
Size	193.9 kB
Tags	Source
SHA-256 checksum How to use checksums	`e6dbb3cba4bf8983064d2db5df36c543b831f320ca57a4a831f96ec45c14dd22`
BLAKE2b-256 checksum How to use checksums	`6e29dabf0c9f08bd4e0a93c84b0dbb29f41663f2b3948f83abf6fad8468e0e2d`
Upload date	Sep 2, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Release files / ovos_ocp_pipeline_plugin-1.4.0a1-py3-none-any.whl

Download URL	ovos_ocp_pipeline_plugin-1.4.0a1-py3-none-any.whl
Size	216.6 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`d500b7f78fe558367953709a0eddcaa6a81762ba4b00877136de0326eac9e8e0`
BLAKE2b-256 checksum How to use checksums	`c01cba898534ac9f41f81712dfef771b2f2cf2b0ec7d59b4d78583e59ff04205`
Upload date	Sep 2, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.13.14`