"""OpenAI-based CAD verification using o3-mini model with structured outputs."""
import base64
import logging
import os
from pathlib import Path
from .generate_png_views import PNGPaths, VerificationResult
import openai
from pydantic import BaseModel
logger = logging.getLogger(__name__)
class StructuredVerificationResult(BaseModel):
"""Simple structured output for CAD verification."""
result: str # "PASS" or "FAIL"
analysis: str # Detailed reasoning
def encode_image_to_base64(image_path: Path) -> str:
"""Convert PNG image to base64 string for OpenAI API."""
with open(image_path, "rb") as image_file:
return base64.b64encode(image_file.read()).decode("utf-8")
def verify_cad_with_vllm(png_files: PNGPaths, criteria: str) -> VerificationResult:
"""
Verify CAD model using OpenAI o3-mini with structured outputs.
Args:
png_files: Dictionary mapping view names to file paths
criteria: Verification criteria string
Returns:
Dictionary containing verification result and analysis
"""
client = openai.OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
# Prepare image content for the API
image_content = []
# Iterate through PNGPaths model attributes
for view, file_path in png_files.model_dump().items():
if Path(file_path).exists():
base64_image = encode_image_to_base64(Path(file_path))
image_content.append({"type": "text", "text": f"View: {view}"})
image_content.append(
{
"type": "image_url",
"image_url": {"url": f"data:image/png;base64,{base64_image}"},
}
)
# Create the prompt
prompt = f"""
Analyze these 3D CAD model images and verify if they meet the following criteria:
Criteria: {criteria}
The images show different views of the same 3D model. Please:
1. Examine each view carefully
2. Check if the model matches the specified criteria
3. Provide detailed analysis of what is correct and what is incorrect
4. Give a final PASS or FAIL result
Be thorough in explaining your reasoning, including specific measurements, shapes, features, and any discrepancies you notice.
"""
# Prepare messages
messages = [
{"role": "user", "content": [{"type": "text", "text": prompt}, *image_content]}
]
# Call OpenAI API with structured output
response = client.beta.chat.completions.parse(
model="o4-mini", messages=messages, response_format=StructuredVerificationResult
)
# Extract the structured result
verification_result = response.choices[0].message.parsed
return VerificationResult(
status=verification_result.result,
reasoning=verification_result.analysis,
criteria=criteria,
)
