Skip to main content

hugpy-video

Execution and artifacts for Hugpy video work, extracted from the monolith's abstract_hugpy_dev.video_intel (+ managers/video, managers/video_gen, comms/studio_assist_log). Import name hugpy_video.

Status: extracted. Runtime code lives here; the package imports with the monolith blocked and with no optional stack (torch/diffusers/ffmpeg bindings) installed. Allowed ecosystem imports: hugpy_platform, hugpy_control, hugpy_engine, hugpy_media. Never hugpy_oracle, hugpy_fleet, hugpy_server (see py/partition.toml, PARTITION.md).

Layout

  • intel/ — media references + jail (media_store), sqlite media job bus (media_bus) and its control-plane bridge (job_bridge), frozen job specs and registry (job_schema), presets, GPU reservation engine (reservation/), runners (ffmpeg, diffusers plane, studio, identity relays, MLT), the studio spine (studio/).
  • chat_video/ — chat-side frame analysis over a hugpy_media vision runner.
  • video_gen/ — StudioVideoRunner / VideoGenRequest: the engine task runner for text-to-video / image-to-video.
  • plugin.py — hugpy_engine.tasks entry point (hugpy_video.plugin:register); import-light, runner resolved lazily.
  • hooks.py — PromptCoordinator protocol (oracle installs its implementation; no-op default). jobs.py — public bus API incl. register_job for upper-layer jobs (the oracle's video_performance).
  • state.py — injectable state roots (HUGPY_VIDEO_STATE_DIR, HUGPY_MEDIA_JOBS_DB, HUGPY_RESERVATIONS_DB, platform storage roots).
  • intel/plane.py — the one seam onto the engine's execute_prompt.
  • cli.py — hugpy-video --help | jobs list | jobs registry | selftest | state.
  • hugpy-video models audit [--json] inventories every declared model's runner gaps, weight pin and minimum VRAM without loading weights.
  • config.py — canonical HUGPY_API_KEY / HUGPY_BASE loader.

Extras: render (numpy/Pillow/requests), studio (torch/diffusers zoo), identity (abstract-identity client), test.

The identity extra installs the separate abstract-identity distribution. The central video runner relays mesh and char360 jobs to its service using IDENTITY_RENDER_URL and IDENTITY_RENDER_TOKEN; installing the extra alone does not start a GPU service. Studio model sweeps use POST /video/studio/tester; each attempt records its model, input image, sampler settings, frame count and negative prompt in the battery log.

Rules (inherited from char360 / hugpy-agent)

  • Heavy imports lazy — must import on a CPU-only box.
  • Version single-sourced from pyproject (importlib.metadata; no hardcoded __version__).
  • Credentials only via hugpy_video.config.load_config — canonical names HUGPY_API_KEY / HUGPY_BASE; legacy aliases accepted but recorded in Config.deprecations.
  • Build: python -m build; publish = copy dists (console dir needs the sha256 sidecar; the pypi index computes hashes itself, dir name must be the normalized dashed name hugpy-video).

Metadata

Release files for hugpy-video 0.2.12

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

Source distribution (sdist)

Source distribution for hugpy-video 0.2.12
File Size Uploaded
hugpy_video-0.2.12.tar.gz 823.6 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for hugpy-video 0.2.12
File Interpreter ABI Platform
hugpy_video-0.2.12-py3-none-any.whl Python 3 none any Details

Total release size: 1.5 MB

Release files / hugpy_video-0.2.12.tar.gz

Download URL hugpy_video-0.2.12.tar.gz
Size 823.6 kB
Tags Source
SHA-256 checksum
How to use checksums
2eb038a1f58f0b2364dcdd3eb2d9abd80b51bc3f34cefbf9f45e6de7f30fd2b6
BLAKE2b-256 checksum
How to use checksums
3fa5003e8b733d7d9f5a29b443c8031f877271d34d867c77f26cb37e46c2182b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.12.3

Release files / hugpy_video-0.2.12-py3-none-any.whl

Download URL hugpy_video-0.2.12-py3-none-any.whl
Size 714.8 kB
Tags Python 3
SHA-256 checksum
How to use checksums
14c6bb88566b8e8b7a7749a68f3650a75095707cedbce97ca15b2f1f37247a3f
BLAKE2b-256 checksum
How to use checksums
29c2ef55587084fc743b7e402579f3a7760b55285147c1066906ba84b3af615f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.12.3
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page