Spaces:
Sleeping
Sleeping
Commit History
6 days ago
-
Upload patch_beatapi_routing.py
Verified signature5d002f3
-
Delete patch_error_message.py, patch_frontend_image_gen.py, patch_image_gen.py, patch_none_guard.py, patch_openrouter_resilience.py
Verified signaturee617ff4 -
Upload Dockerfile
Verified signaturea4ac4c2 -
Upload patch_agent_backend.py
Verified signature1bdccc4 -
Upload patch_ming_models.py
Verified signature82d81e8 -
Upload patch_frontend_ming_models.py
Verified signature11757fc -
Upload BUILD_TRIGGER
Verified signatured6a11d7
-
Upload Dockerfile
Verified signature6199151 -
Upload patch_image_gen.py
Verified signaturecb8a603 -
Upload patch_image_gen.py
Verified signatureaed30d1 -
Upload patch_frontend_image_gen.py
Verified signature73290fa -
Upload patch_frontend_image_gen.py
Verified signatureba387fa -
Upload patch_none_guard.py
Verified signaturecec709f -
Upload patch_none_guard.py
Verified signature6e89546 -
Activate patch_none_guard.py to fix 'NoneType' object has no attribute 'usage'
Verified signature37d9be7 -
Fix 'NoneType' object has no attribute 'usage': add terminal LLMResult after the retry for-loop in both LLM helpers + defensive None/usage guards in run_agent
Verified signaturec6bf198 -
Add image generation frontend: ming-image model in picker + route to /api/image/generate with rendered result
Verified signatureb32e3e8 -
Fix silent auto-approve + add provider-aware auth error message
Verified signatureabfe09a -
Add provider-aware auth error message: OpenRouter models must not be told to set HF_TOKEN
Verified signature360e614 -
Fix silent auto-approve failure: anchor correctly, add runtime opt-out, drop dead mid-task block
Verified signaturec74247c -
Upload patch_none_guard.py
Verified signaturee06d7b5 -
Trigger rebuild: add typesafe/jev-router, fix auto-approve patch, add OpenRouter resilience (tool-less retry + 429 backoff)
Verified signaturea2cd061 -
Fix ordering bug in v2: anchor after auto-approval attrs are initialized so set_auto_approval_policy is not overwritten
Verified signature5b7db21 -
Add patch_openrouter_resilience.py: tool-less model retry (jev-router) + 429 backoff (qwen3.8-27b); add typesafe/jev-router
Verified signatureef66a81 -
Handle tool-less models (jev-router) and OpenRouter shared-pool 429 with bounded retry
Verified signature81a456d -
Add typesafe/jev-router to the model picker (3 chat models total)
Verified signature86c9a1d -
<content>"""Patch backend: register the new chat models using the Laguna S 2.1 model-path convention.
Verified signatureEvery model id is prefixed with `openai/`, which `_resolve_llm_params` strips before calling OpenRouter (and `patch_openrouter_fix.py` forces `custom_llm_provider="openrouter"` so LiteLLM uses the right client). Verified against the live OpenRouter catalog (GET /api/v1/models, 464 models, fetched repeatedly with identical results): - openai/dots-studio/dots-3-note-preview:free PRESENT, chat-capable, 512k ctx, free, supports tools. - openai/qwen/qwen3.8-27b:free PRESENT, chat-capable, 262k ctx, free, supports tools. NOTE: OpenRouter sometimes answers 429 "temporarily rate-limited upstream" for this id. That is a shared-pool provider limit, not a config problem -- see patch_jev_router_tools.py for the retry/fallback handling. - openai/typesafe/jev-router PRESENT, chat-capable, 1M ctx, multimodal in (text/image/audio/video/file). supported_parameters is EMPTY, so it accepts no `tools`/`tool_choice`. The agent loop always sends tools,…e30ec39 -
Fix auto-approve: wrong anchor (multi-line __init__ signature) AND wrong API (auto_approval dict vs auto_approval_enabled flag)
Verified signaturee940c93 -
Trigger rebuild after dropping ming-image-0.1-design and span-01-lite
Verified signature7650a29 -
Add only the 2 chat-capable models to the picker; exclude ming-image-0.1-design (image-gen) and span-01-lite (not on OpenRouter)
Verified signature75b487e -
Fix NameError: remove stray import_anchor_holder line
Verified signature11f3b44 -
Register only the 2 chat-capable models; drop ming-image-0.1-design (image-gen, chat/completions 404) and span-01-lite (absent from OpenRouter catalog)
Verified signaturea1aefbf -
Trigger rebuild: dots-3-note-preview, qwen3.8-27b, span-01-lite, ming-image-0.1-design
Verified signature4fdfe61 -
Fix idempotency bug: guard _available_models insert, assert no duplicate entries
Verified signature1c4ea53 -
Dockerfile: run new patches for dots-3-note / qwen3.8-27b / span-01-lite / ming-image-design
Verified signature84cceba -
Add frontend patch for 4 new models (dots-3-note, qwen3.8-27b, span-01-lite, ming-image-design)
Verified signaturea682471 -
Add backend patch for 4 new models (dots-3-note, qwen3.8-27b, span-01-lite, ming-image-design)
Verified signature75f4728
-
Upload Dockerfile
Verified signature39c03a8 -
Upload Dockerfile
Verified signature3517d32 -
Upload patch_restore_full_app.py
Verified signature1d6872e -
Upload BUILD_TRIGGER
Verified signature8ddc36d -
Upload patch_frontend.py
Verified signature7155db3 -
Upload README.md
Verified signature1ba0abe -
Upload patch_models.py
Verified signature12981e9 -
Upload BUILD_TRIGGER
Verified signature3f601ad -
Upload README.md
Verified signaturef1d8dde -
Upload Dockerfile
Verified signature82cf28b -
Upload patch_space_bunny.py
Verified signature8dbf890 -
Upload patch_models.py
Verified signaturec7a9183
-
Fix OpenRouter models: add custom_llm_provider='openrouter' so litellm stops failing on non-provider org prefixes (upstage/nvidia/liquid)
9800705