Add error handling for LLM response parsing

[davidpoblador]

Summary

The LLM module directly accesses dictionary keys without error handling, which will crash if the response format differs from expected.

Current code

src/blogtuner/ai/llm.py:48-50:

response = model.prompt(prompt, schema=MarkdDownContent)
markdown_content = json.loads(response.text())["markdown"]
return markdown_content

Issues

1. No try/except around json.loads() - will crash on invalid JSON
2. Direct ["markdown"] access - will crash if key is missing
3. No validation that response matches expected schema

Proposed change

try:
    response = model.prompt(prompt, schema=MarkdDownContent)
    data = json.loads(response.text())
    markdown_content = data.get("markdown")
    if markdown_content is None:
        raise ValueError("LLM response missing 'markdown' field")
    return markdown_content
except json.JSONDecodeError as e:
    raise ValueError(f"Invalid JSON response from LLM: {e}") from e

Priority

High

File

src/blogtuner/ai/llm.py