How to use Pydantic with OpenRouter API

This page lists a series of tutorials for using Pydantic with different LLMs on OpenRouter

Three things to note:

1 While OpenRouter does provide a way to specify the schema as one of the API parameters, I have not had much luck getting it to work with the LLMs I have tried. It could have been an error on my part, but there are two more points to consider.

2 Not all LLMs support structured outputs (via OpenRouter’s API), but you can hard code the JSON schema into the prompt itself to get around this problem

3 It is much easier to customize and “engineer” the prompt using my system compared to the API parameter exposed by OpenRouter

Basic Setup

This is the structure of the prompt

prompt = f"""
{system_instruction} 

Report: {input_text}

JSON Schema:
{json_schema_text}
"""

Notice that there are three parts to the prompt:

• The system instruction
• The input text
• The response schema we expect

This is the system instruction

Here is the input text

This is the schema I am using

This basic setup applies to all the tutorials listed below

While you should be able to follow along with all the tutorials by reading the code I have published (and getting a bit of help from some LLM if necessary), you can also download the entire Python project from the course below (open the project in PyCharm inside a venv and pip install the requirements)

Tutorials