feat: add streaming support to LLM API calls and configuration

.gitignore

@@ -8,6 +8,8 @@ wheels/
 .DS_Store
 .vscode
 .idea
+.codebuddy/
+
 
 .venv
 .env

config/base_config.yaml

@@ -16,6 +16,7 @@ construction:
     enable_fast_mode: true
     struct_weight: 0.3
     max_total_communities: 100
+  stream: false
 
 datasets:
   hotpot:

config/config_loader.py

@@ -35,6 +35,7 @@ class ConstructionConfig:
     datasets_no_chunk: list = None
     chunk_size: int = 1000
     overlap: int = 200
+    stream: bool = False
 
     def __post_init__(self):
         if self.datasets_no_chunk is None:

models/constructor/kt_gen.py

@@ -30,6 +30,7 @@ def __init__(self, dataset_name, schema_path=None, mode=None, config=None):
         self.llm_client = call_llm_api.LLMCompletionCall()
         self.all_chunks = {}
         self.mode = mode or config.construction.mode
+        self.stream = config.construction.stream
 
     def load_schema(self, schema_path) -> Dict[str, Any]:
         try:
@@ -112,7 +113,7 @@ def save_chunks_to_file(self):
         logger.info(f"Chunk data saved to {chunk_file} ({len(all_data)} chunks)")
 
     def extract_with_llm(self, prompt: str):
-        response = self.llm_client.call_api(prompt)
+        response = self.llm_client.call_api(prompt, stream=self.stream)
         parsed_dict = json_repair.loads(response)
         parsed_json = json.dumps(parsed_dict, ensure_ascii=False)
         return parsed_json 

utils/call_llm_api.py

@@ -29,27 +29,41 @@ def __init__(self):
         else:
             self.client = OpenAI(base_url=self.llm_base_url, api_key = self.llm_api_key)
 
-    def call_api(self, content: str) -> str:
+    def call_api(self, content: str, temperature: float = 0.3, stream: bool = False) -> str:
         """
         Call API to generate text with retry mechanism.
-        
+
         Args:
             content: Prompt content
-
+            temperature: Sampling temperature (0.0-2.0)
+            stream: Whether to use streaming mode
+
         Returns:
             Generated text response
         """
-            
+
         try:
             completion = self.client.chat.completions.create(
                 model=self.llm_model,
                 messages=[{"role": "user", "content": content}],
-                temperature=0.3
+                temperature=temperature,
+                stream=stream
             )
-            raw = completion.choices[0].message.content or ""
+
+            if stream:
+                # Handle streaming response
+                content_parts = []
+                for chunk in completion:
+                    if chunk.choices[0].delta.content is not None:
+                        content_parts.append(chunk.choices[0].delta.content)
+                raw = "".join(content_parts)
+            else:
+                # Handle non-streaming response
+                raw = completion.choices[0].message.content or ""
+
             clean_completion = self._clean_llm_content(raw)
             return clean_completion
-            
+
         except Exception as e:
             logger.error(f"LLM api calling failed. Error: {e}")
             raise e