Hi everyone, I’m pretty new to working with LLM APIs and could use some advice.
I’m working on a project where I need to pull specific information out of emails and support tickets, like dates, product names, and issue types, and then put that into a structured JSON format. I’m planning to build this in Python.
I’ve been looking at OpenAI, Anthropic, and maybe Google’s Gemini, but I’m a bit overwhelmed. For a task like structured data extraction, is one provider generally better than the others? I care about accuracy first, but cost is also a factor since we’ll be processing a decent volume.
I’ve seen people talk about using function calling or structured outputs for this. Does that vary a lot between providers? Any tips on which one might be easier to implement for a beginner would be super helpful. 😅
Also, if anyone has run into issues where the model “hallucinates” extra fields or ignores the format, I’d love to hear how you handled it.