Instructor Compatibility

import instructor
import openai

# Configure the OpenAI client
client = openai.OpenAI(
    base_url="https://api.vlm.run/v1/openai",
    api_key="<VLMRUN_API_KEY>"
)

# Configure the Instructor client
inst_client = instructor.from_openai(
    client, mode=instructor.Mode.MD_JSON
)

from pydantic import BaseModel, Field

# Let's define a simplified Pydantic model to represent the outputs. Let's say we want to generate an image of big ben at a distance, and crop into the clock face.
class AgentResponse(BaseModel):
    image_url: str = Field(description="The pre-signed URL of the image generated.")
    clock_image_url: str = Field(description="The pre-signed URL of the image of the clock face cropped from the generated image.")
    crop_xywh: tuple[float, float, float, float] = Field(description="The (x, y, width, height) of the clock face cropped from the generated image.")

# Now we can use the Instructor client to create a chat completion
response = inst_client.chat.completions.create(
    model="vlmrun-orion-1:auto",
    max_retries=0,
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Generate an image of big ben at a distance. Crop into the clock face and provide a close up of the clock face."},
            ],
        }
    ],
    response_model=AgentResponse,
)
logger.debug(f"type={type(response)}, response={response}")

# Adapt the above example with an additional "json_schema" parameter
# to enforce guided decoding of the JSON in the schema.
response = inst_client.chat.completions.create(
    ... # same as above
    response_model=AgentResponse,
    response_format={"type": "json_schema", "json_schema": AgentResponse.model_json_schema()},
)