Spaces:
Running
Running
Commit History
24 days ago
-
fix(llm): revert groq to openai/gpt-oss-20b per user request
c53fce6 -
fix(llm): use nemotron-3-ultra-550b for nvidia and llama3-70b-8192 for groq
c494041 -
fix : model fix
fed8ab9 -
fix : model fix
29b315d
25 days ago
-
fix : model fix
3cfec19 -
fix(security): mask Gemini API key in logs, downgrade NIM/Groq model strings, add challenge license
b1b706a
26 days ago
-
fix(core): resolve caching, lazy load, and thread leak defects
- Prevent generic fallback response from caching and poisoning state. - Read LOCAL_7B_P95 from environment variables instead of a magic hardcoded number. - Refactor local tier to stream generation and abort executor thread if deadline is exceeded. - Remove over-broad words 'JSON' and 'system' from default taboo list. - Wrap local model loading in a lazy-load function to prevent HF space boot failure.86a56f9
27 days ago
-
refactor(core): decouple logic from entrypoint and harden local slm constraints
- Extract LLM provider tiering, prompting, and retry logic from app.py into dedicated scripts/ modules. - Implement strict LlamaGrammar enforcement for Qwen2.5-3B local tier to eliminate JSON malformations. - Introduce validation+repair loop catching taboo constraints, word-count limits, and CTA proliferation. - Introduce dynamic budget checks for cloud-tier retries to prevent local starvation. - Fix FEW_SHOT_EXAMPLES substring-matching issue to correctly map all data triggers. - Restore TRIGGER PRIMACY in the local system prompt and configure local temperature (0.2) for optimal variance vs grounding. - Polish README and architecture documentation.c1c5ecb
30 days ago
-
feat: implement 4-tier LLM fail-fast cascade & benchmark-based local 3B selection
e67482e