Update api.py

api.py

@@ -5,22 +5,22 @@ import os
 import json
 import time
 import uuid
+import logging
 
 api = Blueprint("api", __name__)
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
-# Load config from llm_config.json
+# Load config
 with open("llm_config.json", "r") as f:
     config = json.load(f).get("openai", {})
 
-# Model details
 REPO_ID = "mradermacher/distilabeled-Hermes-2.5-Mistral-7B-GGUF"
 MODEL_FILENAME = "distilabeled-Hermes-2.5-Mistral-7B.Q2_K.gguf"
 HF_TOKEN = os.environ.get("HF_API_TOKEN")
 CACHE_DIR = "/app/.cache/huggingface"
-
 os.makedirs(CACHE_DIR, exist_ok=True)
 
-# Download model if not already cached
 MODEL_PATH = hf_hub_download(
     repo_id=REPO_ID,
     filename=MODEL_FILENAME,
@@ -28,18 +28,16 @@ MODEL_PATH = hf_hub_download(
     token=HF_TOKEN
 )
 
-# Load model
 llm = Llama(
     model_path=MODEL_PATH,
     n_ctx=2048,
     n_threads=4,
-    n_gpu_layers=0  # Adjust >0 for GPU acceleration if needed
+    n_gpu_layers=0
 )
 
 @api.route("/v1/chat/completions", methods=["POST"])
 def chat_completions():
-    print("Received request at /v1/chat/completions")  # <-- Log each call
-
+    logger.info("Received request at /v1/chat/completions")
     try:
         data = request.get_json(force=True)
         messages = data.get("messages", [])
@@ -48,19 +46,20 @@ def chat_completions():
         top_p = float(data.get("top_p", config.get("top_p", 0.95)))
         stop = data.get("stop", config.get("stop", ["User:", "Assistant:"]))
 
-        # Build prompt text from chat messages
         prompt_lines = []
         for msg in messages:
             role = msg.get("role", "").lower()
             content = msg.get("content", "").strip()
+            if not content:
+                continue
             if role == "user":
                 prompt_lines.append(f"User: {content}")
             elif role == "assistant":
                 prompt_lines.append(f"Assistant: {content}")
+
         prompt_lines.append("Assistant:")
         prompt = "\n".join(prompt_lines)
 
-        # Generate text from the model
         response = llm(
             prompt,
             max_tokens=max_tokens,
@@ -69,10 +68,11 @@ def chat_completions():
             stop=stop
         )
 
-        # Extract generated text safely
-        result_text = response.get("choices", [{}])[0].get("text", "").strip()
+        result_text = ""
+        choices = response.get("choices")
+        if choices and isinstance(choices, list):
+            result_text = choices[0].get("text", "").strip()
 
-        # Return response in OpenAI chat completion format
         return jsonify({
             "id": f"chatcmpl-{uuid.uuid4().hex}",
             "object": "chat.completion",
@@ -89,5 +89,5 @@ def chat_completions():
         })
 
     except Exception as e:
-        # Return error message as JSON response
+        logger.error(f"Error in chat_completions: {e}", exc_info=True)
         return jsonify({"error": str(e)}), 500