Commit History

7 days ago

  1. LLM brain: Groq llama-3.1-8b-instant @ api.groq.com/openai/v1

    Verified signature
    79842ae
  2. Correct stale Space claims; document slim-image + bucket deployment for HF Spaces

    Verified signature
    19a9e34
  3. Pin to single GPU (CUDA_VISIBLE_DEVICES=0): multi-GPU hosts OOM the boot-time model load; stack is single-GPU by design

    Verified signature
    5eadf1b
  4. Hardware findings: 2x-a10g multi-GPU host OOMs without the single-GPU pin (now baked in); a10g-large marked unverified

    Verified signature
    373f9c7
  5. Correct Space hardware guidance (l40s/a10g-large, a10g-small OOM warning); add demo Space link and RAM requirement

    Verified signature
    39cea70
  6. Cap build parallelism to avoid OOM on memory-limited builders (caught by Space build)

    Verified signature
    68ec264
  7. Fix build: libavswscale-dev does not exist in Ubuntu 22.04, correct package is libswscale-dev (caught by Space build)

    Verified signature
    26f0416
  8. README: GPU smoke verification results + fixed-bug inventory

    Verified signature
    71ff2b0
  9. smoke: follow 307 redirect, 20-min readiness window, readable logs

    Verified signature
    7e3368e
  10. smoke: patch upstream cosyvoice warmup call (stream=True keyword)

    Verified signature
    e2ac997
  11. Dockerfile: patch upstream cosyvoice warmup call at build time

    Verified signature
    710ddfd
  12. config: 48k stand-in voice path

    Verified signature
    d45674d
  13. smoke: resample stand-in voice to 48k before boot

    Verified signature
    90019d0
  14. config: 48k stand-in voice path

    Verified signature
    8a978c0
  15. Dockerfile: resample stand-in voice to 48k at build time

    Verified signature
    9c4f40a

8 days ago

  1. config: fix stand-in voice ref path to submodule asset

    Verified signature
    0dc6729
  2. config: fix stand-in voice ref path to submodule asset

    Verified signature
    bbe91cb
  3. smoke: kill full process tree on all exit paths

    Verified signature
    913e528
  4. smoke: pyarrow<16 pin + pre-boot CosyVoice import check

    Verified signature
    0fd0220
  5. Dockerfile: pyarrow<16 pin (numpy1-compatible)

    Verified signature
    a1d2054
  6. Dockerfile: --no-sync on post-install steps

    Verified signature
    aae61eb
  7. smoke: --no-sync on post-install uv runs (upgrade no longer reverted)

    Verified signature
    0c535cb
  8. Dockerfile: typing_extensions>=4.14 fix from GPU smoke test

    Verified signature
    3402560
  9. smoke: typing_extensions fix + correct readiness grep + traceback fast-fail

    Verified signature
    68a06cb
  10. smoke: bootstrap pip via ensurepip (base image lacks pip)

    Verified signature
    bbd43f6
  11. smoke: install python3.11 via deadsnakes to match Dockerfile

    Verified signature
    f842b68
  12. docs: face/voice collection spec + consent/legal

    Verified signature
    e532dba
  13. script: GPU end-to-end boot verification

    Verified signature
    c4ce248
  14. Dockerfile: pinned OpenAvatarChat + MuseTalk + local CosyVoice2, weights baked in

    Verified signature
    230f145
  15. compose: self-host on any CUDA 12.8 host

    Verified signature
    3784704
  16. config: MuseTalk production pipeline

    Verified signature
    5714994
  17. README: pipeline, deploy options, personalization guide

    Verified signature
    7b02b71
  18. config: LiteAvatar lightweight fallback

    Verified signature
    e5217b4
  19. docs: Zoom/Meet/Teams/LiveKit bridge routes + latency budget

    Verified signature
    88da9d7
  20. initial commit

    Verified signature
    90db54d