neuralnode 2.0.3__tar.gz → 2.0.4__tar.gz

{neuralnode-2.0.3 → neuralnode-2.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: neuralnode
-Version: 2.0.3
+Version: 2.0.4
 Summary: Comprehensive AI Framework with 50+ LLM Providers, Advanced Agents, Chains, Memory, RAG, and 100+ Tools
 Project-URL: Homepage, https://assem.cloud/
 Project-URL: Documentation, https://neuralnode.readthedocs.io

{neuralnode-2.0.3 → neuralnode-2.0.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "neuralnode"
-version = "2.0.3"
+version = "2.0.4"
 description = "Comprehensive AI Framework with 50+ LLM Providers, Advanced Agents, Chains, Memory, RAG, and 100+ Tools"
 readme = "README.md"
 requires-python = ">=3.9"

{neuralnode-2.0.3 → neuralnode-2.0.4}/src/neuralnode/__init__.py

@@ -42,7 +42,7 @@ Quick Start::
     text = sr.listen()
 """
 
-__version__ = "2.0.3"
+__version__ = "2.0.4"
 __author__ = "NeuralNode Contributors"
 
 # ── Core types ────────────────────────────────────────────────────────────────

{neuralnode-2.0.3 → neuralnode-2.0.4}/src/neuralnode/providers/horus.py

@@ -237,10 +237,10 @@ class HorusProvider(BaseLLMProvider):
         force_download: bool = False,
         resume_download: bool = False,
         max_new_tokens: int = 512,
-        temperature: float = 0.7,
-        top_p: float = 0.9,
+        temperature: float = 0.3,
+        top_p: float = 0.85,
         top_k: int = 50,
-        repetition_penalty: float = 1.1,
+        repetition_penalty: float = 1.2,
         do_sample: bool = True,
         num_beams: int = 1,
         early_stopping: bool = False,
@@ -433,13 +433,18 @@ class HorusProvider(BaseLLMProvider):
         repo_id, filename = self._split_repo_and_filename(self.model_id)
         model_path = filename
         if repo_id and HF_HUB_AVAILABLE:
-            model_path = hf_hub_download(
-                repo_id=repo_id,
-                filename=filename,
-                cache_dir=self.cache_dir,
-                local_files_only=self.local_files_only,
-                token=self.token,
-            )
+            with warnings.catch_warnings():
+                warnings.filterwarnings(
+                    "ignore",
+                    message=r".*You are sending unauthenticated requests to the HF Hub.*",
+                )
+                model_path = hf_hub_download(
+                    repo_id=repo_id,
+                    filename=filename,
+                    cache_dir=self.cache_dir,
+                    local_files_only=self.local_files_only,
+                    token=self.token,
+                )
 
         llama_kwargs: Dict[str, Any] = {
             "model_path": model_path,
@@ -788,20 +793,34 @@ class HorusProvider(BaseLLMProvider):
             output = self.model(
                 prompt,
                 max_tokens=int(settings.get("max_new_tokens", 512)),
-                temperature=settings.get("temperature", 0.7),
-                top_p=settings.get("top_p", 0.9),
+                temperature=settings.get("temperature", 0.3),
+                top_p=settings.get("top_p", 0.85),
                 top_k=settings.get("top_k", 50),
-                repeat_penalty=settings.get("repetition_penalty", 1.1),
-                stop=["\nUser:", "\nSystem:", "\nTool:"],
+                repeat_penalty=settings.get("repetition_penalty", 1.2),
+                stop=["<|end|>", "<|user|>", "\n<|user|>", "<|assistant|>", "\n<|assistant|>"],
                 echo=False,
             )
-        return output["choices"][0]["text"].strip()
+        return self._clean_generated_text(output["choices"][0]["text"])
+
+    def _clean_generated_text(self, text: str) -> str:
+        """Clean generation artifacts from Horus outputs."""
+        cleaned = text or ""
+        cleaned = cleaned.replace("<|assistant|>", "").replace("<|user|>", "").replace("<|end|>", "")
+        cleaned = re.sub(r"(?im)^\s*assistant\s*[:：]?\s*", "", cleaned)
+        cleaned = re.sub(r"\n{3,}", "\n\n", cleaned)
+        cleaned = cleaned.strip()
+        # Cut obvious accidental next-turn continuation if present.
+        for marker in ("\nUser:", "\nSystem:", "\nTool:"):
+            idx = cleaned.find(marker)
+            if idx > 0:
+                cleaned = cleaned[:idx].strip()
+        return cleaned
 
     def chat(
         self,
         messages: List[Dict[str, Any]],
         model: Optional[str] = None,
-        temperature: float = 0.7,
+        temperature: Optional[float] = None,
         max_tokens: Optional[int] = None,
         top_p: Optional[float] = None,
         frequency_penalty: Optional[float] = None,
@@ -833,7 +852,7 @@ class HorusProvider(BaseLLMProvider):
         
         prompt = self._render_prompt(normalized)
         generation_kwargs = {
-            "temperature": temperature,
+            "temperature": temperature if temperature is not None else self.generation_config["temperature"],
             "max_new_tokens": max_tokens or self.generation_config["max_new_tokens"],
             "top_p": top_p if top_p is not None else self.generation_config["top_p"],
             **kwargs,