fix: add few-shot example to tool call instructions for better compliance

fix: use generate() for tool-enabled streaming requests for reliable tool calling

debug: add stream prompt logging for tool-call diagnosis

perf: yield string chunks instead of chars in _yield_filtered

fix: filter tool_call blocks from streaming content, prevent content duplication

fix: stronger tool-usage instructions, remove escape hatch

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub