---
description: AGNO framework, and you master all the agent parameters
globs:
alwaysApply: false
---
# AGNO Agent Parameters Reference
> You are an expert in AGNO framework, and you master all the agent parameters. You focus on building intelligent agents that can think, analyze, and reason through complex problems step by step. In each agent remeber to set the debug_mode=true to let the developer have full logging.
This rule provides a comprehensive reference of all parameters available when creating AGNO agents. Parameters are organized by functional categories for easier reference.
## Core Agent Parameters
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `model` | `Model` | Required | The language model powering the agent |
| `name` | `str` | None | Name of the agent (useful for multi-agent systems) |
| `description` | `str` | None | Description of the agent's purpose and capabilities |
| `role` | `str` | None | Role the agent should assume |
| `instructions` | `List[str]` | [] | List of instructions for guiding agent behavior |
| `expected_output` | `str` | None | Format specification for the expected response |
| `markdown` | `bool` | False | Whether the agent should use markdown formatting |
| `session_id` | `str` | None | Unique identifier for the agent session (auto-generated if None) |
| `run_id` | `str` | None | Unique identifier for the run (auto-generated if None) |
## Tools and Functionality
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `tools` | `List[BaseTool]` | [] | List of tools the agent can use |
| `knowledge` | `Knowledge` | None | Knowledge base for the agent to search |
| `search_type` | `Literal["hybrid", "semantic", "keyword"]` | "hybrid" | How to search the knowledge base |
| `verbose` | `bool` | False | Display more detailed information during execution |
| `debug_mode` | `bool` | False | Enable debug information and logging |
| `show_tool_calls` | `bool` | False | Show tool calls in the response |
| `show_tool_calls_in_message` | `bool` | False | Include tool calls in the message sent to the model |
| `add_citations_to_tool_calls` | `bool` | False | Add citations to tool call responses |
| `cache_dir` | `str` | ".agno/cache" | Directory for caching data |
## Memory and State Management
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `storage` | `Storage` | None | Storage driver for saving agent sessions |
| `memory` | `Memory` | None | Memory driver for user memories and preferences |
| `enable_agentic_memory` | `bool` | False | Allow agent to manage memories |
| `enable_user_memories` | `bool` | False | Automatically create user memories |
| `add_history_to_messages` | `bool` | False | Add historical messages to context |
| `num_history_runs` | `int` | 0 | Number of history runs to add to messages |
| `num_history_responses` | `int` | 0 | Number of history responses to add |
| `read_chat_history` | `bool` | False | Add tool to let agent read chat history |
| `read_tool_call_history` | `bool` | False | Add tool to read previous tool calls |
## Response Configuration
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `response_model` | `Type[BaseModel]` | None | Pydantic model for structured output |
| `structured_outputs` | `bool` | False | Format responses as structured JSON |
| `json_mode` | `bool` | False | Force model to return valid JSON |
| `max_tokens` | `int` | None | Maximum tokens in the response |
| `temperature` | `float` | None | Controls randomness (0.0 to 2.0) |
| `top_p` | `float` | None | Controls diversity via nucleus sampling |
| `top_k` | `int` | None | Limits vocabulary to top K tokens |
| `frequency_penalty` | `float` | None | Penalizes repeated tokens |
| `presence_penalty` | `float` | None | Penalizes tokens already used |
## Knowledge Search Parameters
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `knowledge_search_params` | `dict` | {} | Customized knowledge search settings |
| `knowledge_search_k` | `int` | 5 | Number of results to retrieve |
| `knowledge_search_score_threshold` | `float` | 0.0 | Minimum similarity score to include |
| `knowledge_search_rerank` | `bool` | False | Apply re-ranking to results |
| `knowledge_max_tokens` | `int` | None | Maximum tokens for knowledge content |
## Runtime Behavior
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `max_retries` | `int` | 0 | Maximum retries on API failure |
| `retry_sleep` | `float` | 1.0 | Seconds to wait between retries |
| `add_datetime_to_instructions` | `bool` | False | Add current date/time to instructions |
| `stream` | `bool` | False | Stream the response as it's generated |
| `stream_intermediate_steps` | `bool` | False | Stream intermediate reasoning steps |
| `show_full_reasoning` | `bool` | False | Show all reasoning steps in the response |
| `show_reasoning_stream` | `bool` | False | Stream reasoning in human-readable format |
## Agent.run() Parameters
| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `message` | `str` | Required | The user message or query |
| `user_id` | `str` | None | User identifier for memory and personalization |
| `stream` | `bool` | False | Stream the response as it's generated |
| `max_tokens` | `int` | None | Override agent's max_tokens for this run |
| `temperature` | `float` | None | Override temperature for this run |
| `top_p` | `float` | None | Override top_p for this run |
| `top_k` | `int` | None | Override top_k for this run |
| `knowledge_search_params` | `dict` | None | Override knowledge search settings |
| `show_full_reasoning` | `bool` | False | Show complete reasoning process |
| `stream_intermediate_steps` | `bool` | False | Stream each reasoning step |
| ` debug_mode` | `bool` | False | Show a complete debug log |
## Example Usage
```python
from agno.agent import Agent
from agno.models.anthropic import Claude
from agno.tools.reasoning import ReasoningTools
from agno.tools.yfinance import YFinanceTools
from agno.memory.v2.db.sqlite import SqliteMemoryDb
from agno.memory.v2.memory import Memory
from agno.storage.sqlite import SqliteStorage
# Comprehensive agent with multiple parameter categories
agent = Agent(
# Core parameters
name="FinanceAdvisor",
model=Claude(id="claude-4-sonnet-latest"),
role="You are a helpful financial advisor",
instructions=[
"Provide accurate financial data",
"Explain concepts clearly",
"Format data in tables when appropriate"
],
expected_output=["A curated list of data"]
# Tools and functionality
tools=[
ReasoningTools(add_instructions=True),
YFinanceTools(stock_price=True, company_info=True)
],
show_tool_calls=True,
debug_mode=True
# Memory and state
storage=SqliteStorage(
table_name="agent_sessions",
db_file="data/examople.db"
),
memory=Memory(
model=Claude(id="claude-4-sonnet-latest"),
db=SqliteMemoryDb(
table_name="user_memories",
db_file="data/finance.db"
)
),
enable_agentic_memory=True,
add_history_to_messages=True,
num_history_runs=3,
# Response configuration
markdown=True,
temperature=0.7,
# Runtime behavior
debug_mode=True,
add_datetime_to_instructions=True
)
# Running the agent with specific parameters
agent.print_response(
"What are the recent trends in tech stocks?",
user_id="john_doe@example.com",
stream=True,
show_full_reasoning=True,
stream_intermediate_steps=True
)
```
