fix: added add_generation_prompt=True without which the LLM does not know when to start a response

Author: guicho271828

mellea/backends/huggingface.py

@@ -342,6 +342,7 @@ def _generate_from_context_standard(
             input_ids = self._tokenizer.apply_chat_template(  # type: ignore
                 ctx_as_conversation,
                 tools=convert_tools_to_json(tools),  # type: ignore
+                add_generation_prompt=True,
                 return_tensors="pt",
                 **self._make_backend_specific_and_remove(model_options),
             ).to(self._device)  # type: ignore