Convert to test server to isolate routing issue

main.py

@@ -1,73 +1,34 @@
+
 from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel
-# We now import BitsAndBytesConfig to specify our quantization settings
-from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
-import torch
 import re
 import json
 
 # ----------------------------
-# 1. Configuration
-# ----------------------------
-
-MODEL_NAME = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
-# The device will be automatically handled by device_map="auto"
-# but we can keep this for logging.
-DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
-
-# ----------------------------
-# 2. FastAPI App Initialization
+# 1. FastAPI App Initialization
 # ----------------------------
 
 app = FastAPI(
-    title="AI Code Review Service",
-    description="An API to get AI-powered code reviews for pull request diffs.",
+    title="AI Code Review Service (Test Mode)",
+    description="A test version of the API without a live AI model.",
     version="1.0.0",
 )
 
 # ----------------------------
-# 3. AI Model Loading
+# 2. Mock AI Model Loading (Simulated)
 # ----------------------------
 
-model = None
-tokenizer = None
-
-def load_model():
-    """Loads the model and tokenizer into memory."""
-    global model, tokenizer
-    if model is None:
-        print(f"Loading model: {MODEL_NAME} on device: {DEVICE}...")
-        tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True)
-
-        # FIX: Define the quantization configuration for 4-bit loading.
-        # We explicitly set bnb_4bit_quant_type to "nf4", which is required for CPU execution.
-        quantization_config = BitsAndBytesConfig(
-            load_in_4bit=True,
-            bnb_4bit_quant_type="nf4",
-            bnb_4bit_compute_dtype=torch.bfloat16,
-            bnb_4bit_use_double_quant=False,
-        )
-
-        # Load the model with the specified quantization config.
-        # We also use device_map="auto" to let transformers handle device placement.
-        model = AutoModelForCausalLM.from_pretrained(
-            MODEL_NAME,
-            trust_remote_code=True,
-            quantization_config=quantization_config,
-            device_map="auto", # This is crucial for bitsandbytes to work correctly
-        )
-        print("Model loaded successfully.")
-
 @app.on_event("startup")
 async def startup_event():
     """
-    On server startup, we trigger the model loading.
+    In this test version, we just print a message.
+    We are not loading any real model.
     """
-    print("Server starting up...")
-    load_model()
+    print("Server starting up in test mode.")
+    print("Model loading is disabled.")
 
 # ----------------------------
-# 4. API Request/Response Models
+# 3. API Request/Response Models
 # ----------------------------
 
 class ReviewRequest(BaseModel):
@@ -82,79 +43,35 @@ class ReviewResponse(BaseModel):
     comments: list[ReviewComment]
 
 # ----------------------------
-# 5. The AI Review Logic
+# 4. The API Endpoint (with Mocked Response)
 # ----------------------------
 
-def run_ai_inference(diff: str) -> str:
+@app.post("/review", response_model=ReviewResponse)
+async def get_code_review(request: ReviewRequest):
     """
-    Runs the AI model to get the review.
+    This endpoint now returns a hardcoded, successful response.
+    It does not call an AI model.
     """
-    if not model or not tokenizer:
-        raise RuntimeError("Model is not loaded.")
+    print("Received request for /review. Returning mocked response.")
+    if not request.diff:
+        raise HTTPException(status_code=400, detail="Diff content cannot be empty.")
 
-    messages = [
-        {
-            "role": "system",
-            "content": """You are an expert code reviewer. Your task is to analyze a pull request diff and provide constructive feedback.\nAnalyze the provided diff and identify potential issues, suggest improvements, or point out good practices.\n\nIMPORTANT: Respond with a JSON array of comment objects. Each object must have three fields: 'file_path', 'line_number', and 'comment_text'.\nThe 'file_path' should be the full path of the file being changed.\nThe 'line_number' must be an integer corresponding to the line number in the *new* version of the file where the comment applies.\nThe 'comment_text' should be your concise and clear review comment.\n\nExample response format:\n[\n    {\n        "file_path": "src/utils/helpers.py",\n        "line_number": 42,\n        "comment_text": "This function could be simplified by using a list comprehension."\n    }\n]\n\nDo not add any introductory text or explanations outside of the JSON array.\n"""
-        },
+    # Create a fake response to prove the endpoint is working.
+    mock_comments = [
         {
-            "role": "user",
-            "content": f"Here is the diff to review:\n\n```diff\n{diff}\n```"
+            "file_path": "src/mock/test.py",
+            "line_number": 10,
+            "comment_text": "This is a test comment from the mock server. If you see this, the API is working!"
         }
     ]
-
-    # Note: We don't need to manually move inputs to a device when using device_map="auto"
-    inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
-    
-    outputs = model.generate(inputs, max_new_tokens=1024, do_sample=False, top_k=50, top_p=0.95, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
     
-    response_text = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
-    return response_text.strip()
-
-def parse_ai_response(response_text: str) -> list[ReviewComment]:
-    """
-    Parses the raw text from the AI to extract the JSON array.
-    """
-    print(f"Raw AI Response:\n---\n{response_text}\n---")
-    
-    json_match = re.search(r'\[.*\]', response_text, re.DOTALL)
-    if not json_match:
-        print("Warning: Could not find a JSON array in the AI response.")
-        return []
-
-    json_string = json_match.group(0)
-    
-    try:
-        comments_data = json.loads(json_string)
-        validated_comments = [ReviewComment(**item) for item in comments_data]
-        return validated_comments
-    except (json.JSONDecodeError, TypeError, KeyError) as e:
-        print(f"Error parsing JSON from AI response: {e}")
-        print(f"Invalid JSON string: {json_string}")
-        return []
-
-# ----------------------------
-# 6. The API Endpoint
-# ----------------------------
-
-@app.post("/review", response_model=ReviewResponse)
-async def get_code_review(request: ReviewRequest):
-    if not request.diff:
-        raise HTTPException(status_code=400, detail="Diff content cannot be empty.")
-
-    try:
-        ai_response_text = run_ai_inference(request.diff)
-        parsed_comments = parse_ai_response(ai_response_text)
-        return ReviewResponse(comments=parsed_comments)
-
-    except Exception as e:
-        print(f"An unexpected error occurred: {e}")
-        raise HTTPException(status_code=500, detail="An internal error occurred while processing the review.")
+    return ReviewResponse(comments=[ReviewComment(**c) for c in mock_comments])
 
 # ----------------------------
-# 7. Health Check Endpoint
+# 5. Health Check Endpoint
 # ----------------------------
 
 @app.get("/health")
 async def health_check():
-    return {"status": "ok", "model_loaded": model is not None}
+    """A simple endpoint to confirm the server is running."""
+    return {"status": "ok", "model_loaded": False} # Model is not loaded in test mode