Structured Outputs (json mode?)

General
1

Does Agno care that some models might have tons of text appended to the final json it returns? Ie - model response “Here is the data you requested: ```json…”

I assume agno will try to extract the json from wherever it is, even if it doesn’t start with ```json?

2

Hi @aberk, thank you for reaching out. Please refer to this FAQ, it should help you with your doubt

please let me know incase you have doubts

3

I read the docs… This doesn’t answer my question. I see the use_json_mode as a fallback, but again this doesn’t confirm what I asked above. Is it basically saying agno will try to resolve my response model as best it could? And if yes, can it handle the scenario I asked about? (A response with some other text information etc.?

4

Hi there!

Great question! Yes, Agno absolutely handles this scenario gracefully. The framework is designed to extract JSON from model responses even when they contain additional explanatory text.

Here’s a practical example of what Agno can handle:

Model Response:

Here is the movie script data you requested:

```json
{
  "setting": "New York City rooftop",
  "ending": "The hero saves the city and gets the girl",
  "genre": "action",
  "name": "Skyline Justice",
  "characters": ["Jack Steel", "Sarah Connor", "Dr. Evil"],
  "storyline": "A tough cop must stop a criminal mastermind from destroying the city using stolen technology."
}

Hope this helps with your project! Let me know if you need any adjustments to the script.


**Your Code:**
```python
class MovieScript(BaseModel):
    setting: str = Field(..., description="Provide a nice setting for a blockbuster movie.")
    ending: str = Field(..., description="Ending of the movie.")
    genre: str = Field(..., description="Genre of the movie.")
    name: str = Field(..., description="Give a name to this movie")
    characters: List[str] = Field(..., description="Name of characters for this movie.")
    storyline: str = Field(..., description="3 sentence storyline for the movie.")

agent = Agent(
    model=Claude(id="anthropic.claude-3-5-sonnet-20240620-v1:0"),
    response_model=MovieScript,
)

response = agent.run("Create a movie about New York")
print(type(response.content))  # <class 'MovieScript'>
print(response.content.name)   # "Skyline Justice"

Agno automatically:

  1. Detects and extracts JSON from code blocks (json)
  2. Handles various JSON embedding patterns
  3. Validates against your Pydantic model
  4. Returns a properly structured object

The system has multiple fallback strategies, so it works reliably across different models and response formats. You don’t need any special configuration - just set your response_model and Agno handles the rest.

Hope this clarifies things! Feel free to reach out if you have any other questions.

Best regards!