Backend Code - Step 5: Blog Topic Generation

# Step 5: Blog Topic Generation & Cluster Analysis

@app.post("/keywords/generate-topics")
async def generate_topics(
    file: UploadFile = File(...),
    min_cluster_size: int = Form(4),
    max_workers: int = Form(10),
    model_name: str = Form("gpt-4o-mini"),
) -> TopicGenerationResponse:
    """
    Generate blog topics and analyze clusters from denoised keywords.
    
    Process:
    1. Load denoised CSV from Step 4
    2. Filter out noise rows (is_noise == "no")
    3. Group by ParentCluster and FinalArticleCluster
    4. Send each cluster to GPT for analysis
    5. GPT returns:
       - Cluster label
       - Good for blog (Yes/No)
       - Reason
       - Needs review flag
       - 3 suggested titles
       - 3 article descriptions
       - Additional noise keywords
    6. Remove identified noise keywords
    7. Calculate cluster metrics (search volume, keyword count)
    8. Remove clusters with < min_cluster_size keywords
    9. Sort results (problematic clusters to bottom)
    10. Save output and removed keywords
    """
    ...

# GPT Prompt:
# - Analyzes clusters for SMB invoicing/finance content potential
# - Generates blog titles and descriptions
# - Identifies noise keywords
# - Flags inconsistent clusters for review
# - Returns structured JSON with all fields