def analyze_image()

in 4-mmrag_tooluse/mmrag_bh.py [0:0]

• 39 lines of code
• 1 McCabe index (conditional complexity)

def analyze_image(base64_image: str, quarter_info: str) -> Dict:
    system_prompt = f"""
    Analyze the image below and determine if it contains graphs or tabular data. 

    - If the image contains a table:
    - Shorten the table title to one of ["Free_Cash_Flow_Reconciliation", "Free_Cash_Flow_Less_Principal_Repayments", "Free_Cash_Flow_Less_Equipment_Finance Leases"].
    - Transcribe the table's title under the "content_output" key.
    - Set "image_category" to "table".

    - If the image contains graphs:
    - Set "image_category" to "graphs".
    - Provide a detailed analysis/summary of the graphs, including:
        - **Descriptions** of what each graph represents.
        - **Key data points** presented as bullet points or numbered lists.
        - **Insights or takeaways or trends** derived from the graphs.

    The quarter information is: {quarter_info}. Please use that as the value for the JSON key "quarter_info".
    """
    response = client.chat.completions.create(
        model=Config.GPT_MODEL,
        response_format={
            "type": "json_schema",
            "json_schema": {
                "name": "image_analysis",
                "schema": {
                    "type": "object",
                    "properties": {
                        "image_category": {"type": "string"},
                        "content_output": {"type": "string"},
                        "quarter_info": {"type": "string"}
                    },
                    "required": ["image_category", "content_output", "quarter_info"],
                    "additionalProperties": False
                },
                "strict": True
            }
        },
        messages=[
            {
                "role": "system",
                "content": system_prompt
            },
            {
                "role": "user",
                "content": [
                    {"type": "image_url", "image_url": {
                        "url": f"data:image/png;base64,{base64_image}", "detail": "high"}}
                ]
            }
        ],
        temperature=0.0,
    )
    response_string = response.choices[0].message.content
    response_dict = json.loads(response_string)
    return response_dict