Respite-API / app.py

Commit History

  1. fix(gate): accept x-ael-key alongside x-respite-key; drop dead key read⁠

    pi
    b3596ff

  1. security: give the searx bridge its own gate key⁠

    64e2bf0
  2. security: key-gate the searx bridge⁠

    d2bc83a

  1. chore: remove the temporary env diagnostic

    Verified signature
    8251b07
  2. chore: temporary names-only env diagnostic

    Verified signature
    977bce0
  3. fix: read the shared secret per request; add a names-only env diagnostic

    Verified signature
    3fee5c9
  4. chore: rebuild so the container picks up RESPIRE_SPACE_KEY

    Verified signature
    bb7d60b
  5. security: require a server-held shared secret for the compute routes

    Verified signature
    13dcd95
  6. security: overwrite Gradio's reflected CORS on /respite/*

    Verified signature
    f1ffefe

  1. Add SearXNG userspace runtime with DDG + Google engines⁠

    Buffy
    42f6abb
  2. Add Node runtime capability probe (userspace, no root)⁠

    Buffy
    ba9ed6e

  1. Expose live audio generation API route

    Verified signature
    58e1c4d
  2. Pass Qwen prompt through async job

    Verified signature
    d3842e2
  3. Fix Qwen multimodal prompt content

    Verified signature
    024ac84
  4. Diagnose Qwen generation termination

    Verified signature
    8f2c140

  1. Enable Qwen reasoning by default

    Verified signature
    ca39a8e

  1. Add CPU Qwen3.8 inference job

    Verified signature
    e8b6e3f
  2. Normalize TinyLlama chat template inputs

    Verified signature
    9a9fa33
  3. Fix CPU TinyLlama PyTorch initialization

    Verified signature
    4380259
  4. Accept JSON body for TinyLlama inference

    Verified signature
    fc0acf8
  5. Add CPU TinyLlama inference to Gradio Space

    Verified signature
    95167c1
  6. Test RAM loaded GGUF without pipe deadlock

    Verified signature
    3956ce2
  7. Rebuild llama.cpp with stable CPU backend

    Verified signature
    f3c6ec5
  8. Limit CPU thread pools to cgroup CPUs

    Verified signature
    1b0a401
  9. Prevent llama output pipe deadlock

    Verified signature
    a8a716e
  10. Probe llama CPU process state

    Verified signature
    8ef48d6
  11. Run loader diagnosis as background job

    Verified signature
    3fdbf36
  12. Add CPU model loader diagnostics

    Verified signature
    a5de9d9
  13. Run CPU benchmarks asynchronously

    Verified signature
    318e7cb
  14. Diagnose single-thread CPU model loading

    Verified signature
    50b3226
  15. Add tiny GGUF CPU control test

    Verified signature
    74b57c5
  16. Use RAM loaded CPU model benchmark

    Verified signature
    6a6f8c0
  17. Separate model loading from CPU warmup

    Verified signature
    005dc00
  18. Add CPU model smoke diagnostics

    Verified signature
    a903372
  19. Serialize CPU benchmark requests

    Verified signature
    75c0e18
  20. Keep local model benchmarks CPU only

    Verified signature
    677b59e
  21. Select llama benchmark executable

    Verified signature
    0329f82
  22. Fix llama benchmark binary cache

    Verified signature
    6e550c2
  23. Use llama bench for bounded CPU measurements

    Verified signature
    4c94260
  24. Fix CPU benchmark diagnostics

    Verified signature
    ce1fcf1

  1. Restore stable Space server launch

    Verified signature
    bc0692a
  2. Add share=True to bypass broken proxy

    Verified signature
    15d9481
  3. Restore working create_app monkey-patch with staticmethod

    Verified signature
    afb0c33
  4. Background thread inject routes into running Gradio server

    Verified signature
    c7f0d89
  5. Fix: inject routes via background thread after launch

    Verified signature
    5e3fbbf
  6. Fix: patch App.__init__ to inject routes into the actual served app

    Verified signature
    d0e8976
  7. Fix: monkey-patch without staticmethod wrapper, spaces before torch, ssr_mode=False

    Verified signature
    e244295
  8. Fix route injection: try demo.app directly, fallback to monkey-patch

    Verified signature
    e3186fc
  9. Add model filter param to benchmark endpoint

    Verified signature
    a76c80d
  10. Fix: build llama.cpp with static linking, find binary after build

    Verified signature
    2ab0e84