Agent Context

Agent Context is another amazing feature of Agno. context is a dictionary that contains a set of functions (or dependencies) that are resolved before the agent runs.

Context is a way to inject dependencies into the description and instructions of the agent.

You can use context to inject memories, dynamic few-shot examples, “retrieved” documents, etc.

agent_context.py

import json
from textwrap import dedent

import httpx
from agno.agent import Agent
from agno.models.openai import OpenAIChat


def get_top_hackernews_stories(num_stories: int = 5) -> str:
    """Fetch and return the top stories from HackerNews.

    Args:
        num_stories: Number of top stories to retrieve (default: 5)
    Returns:
        JSON string containing story details (title, url, score, etc.)
    """
    # Get top stories
    stories = [
        {
            k: v
            for k, v in httpx.get(
                f"https://hacker-news.firebaseio.com/v0/item/{id}.json"
            )
            .json()
            .items()
            if k != "kids"  # Exclude discussion threads
        }
        for id in httpx.get(
            "https://hacker-news.firebaseio.com/v0/topstories.json"
        ).json()[:num_stories]
    ]
    return json.dumps(stories, indent=4)


# Create a Context-Aware Agent that can access real-time HackerNews data
agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    # Each function in the context is evaluated when the agent is run,
    # think of it as dependency injection for Agents
    context={"top_hackernews_stories": get_top_hackernews_stories},
    # Alternatively, you can manually add the context to the instructions
    instructions=dedent("""\
        You are an insightful tech trend observer! 📰

        Here are the top stories on HackerNews:
        {top_hackernews_stories}\
    """),
    # add_state_in_messages will make the `top_hackernews_stories` variable
    # available in the instructions
    add_state_in_messages=True,
    markdown=True,
)

# Example usage
agent.print_response(
    "Summarize the top stories on HackerNews and identify any interesting trends.",
    stream=True,
)

Adding the entire context to the user message

Set add_context=True to add the entire context to the user message. This way you don’t have to manually add the context to the instructions.

agent_context_instructions.py

import json
from textwrap import dedent

import httpx
from agno.agent import Agent
from agno.models.openai import OpenAIChat


def get_top_hackernews_stories(num_stories: int = 5) -> str:
    """Fetch and return the top stories from HackerNews.

    Args:
        num_stories: Number of top stories to retrieve (default: 5)
    Returns:
        JSON string containing story details (title, url, score, etc.)
    """
    # Get top stories
    stories = [
        {
            k: v
            for k, v in httpx.get(
                f"https://hacker-news.firebaseio.com/v0/item/{id}.json"
            )
            .json()
            .items()
            if k != "kids"  # Exclude discussion threads
        }
        for id in httpx.get(
            "https://hacker-news.firebaseio.com/v0/topstories.json"
        ).json()[:num_stories]
    ]
    return json.dumps(stories, indent=4)


# Create a Context-Aware Agent that can access real-time HackerNews data
agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    # Each function in the context is resolved when the agent is run,
    # think of it as dependency injection for Agents
    context={"top_hackernews_stories": get_top_hackernews_stories},
    # We can add the entire context dictionary to the instructions
    add_context=True,
    markdown=True,
)

# Example usage
agent.print_response(
    "Summarize the top stories on HackerNews and identify any interesting trends.",
    stream=True,
)