Agent Configuration Schema Reference¶

Complete reference for all agent configuration fields based on Agno Agent schema.

Schema Overview¶

Agent configurations are dictionaries that map to Agno Agent parameters. All fields are optional unless specified, and follow Agno's default behavior when omitted.

Core Agent Fields¶

Identity & Metadata¶

Field	Type	Required	Default	Description
`name`	`str`	No	Auto-generated	Agent name (usually set during creation)
`id`	`str`	No	Auto-generated	Agent ID
`description`	`str`	No	`None`	Agent description
`instructions`	`List[str]`	Yes	`[]`	Agent instructions (system prompt)

Model Settings¶

Field	Type	Required	Default	Description
`model`	`Model`	Yes	Set during creation	Model instance (usually passed separately)

Memory Fields¶

Agentic Memory¶

Field	Type	Required	Default	Description
`enable_agentic_memory`	`bool`	No	`False`	Enable agentic memory (can create/update/delete)
`memory_manager`	`MemoryManager`	No	`None`	Memory manager instance (usually passed separately)

User Memories¶

Field	Type	Required	Default	Description
`enable_user_memories`	`bool`	No	`False`	Enable user memories (can create/update only)
`add_memories_to_context`	`bool`	No	`False`	Add memories to context

Reference: Agno Memory Documentation

Storage & Sessions Fields¶

Chat History¶

Field	Type	Required	Default	Description
`add_history_to_context`	`bool`	No	`False`	Add chat history to context
`num_history_runs`	`int`	No	`0`	Number of history runs to include (0-100 recommended)
`read_chat_history`	`bool`	No	`False`	Read chat history

Session Management¶

Field	Type	Required	Default	Description
`session_id`	`str`	No	`None`	Session ID (usually set at runtime)
`cache_session`	`bool`	No	`False`	Cache session state
`search_session_history`	`bool`	No	`False`	Search across sessions
`num_history_sessions`	`int`	No	`0`	Number of sessions to search (0-100 recommended)
`enable_session_summaries`	`bool`	No	`False`	Generate session summaries
`add_session_summary_to_context`	`bool`	No	`False`	Add session summary to context

Agent State¶

Field	Type	Required	Default	Description
`enable_agentic_state`	`bool`	No	`False`	Enable agentic state management
`add_session_state_to_context`	`bool`	No	`False`	Add session state to context
`session_state`	`Dict`	No	`None`	Session state (usually set at runtime)

Reference: Agno Storage Documentation

Knowledge (Agentic RAG) Fields¶

Field	Type	Required	Default	Description
`knowledge`	`Knowledge`	No	`None`	Knowledge instance (usually passed separately)
`search_knowledge`	`bool`	No	`False`	Enable knowledge search at runtime
`knowledge_filters`	`List`	No	`None`	Knowledge filters
`enable_agentic_knowledge_filters`	`bool`	No	`False`	Let agent decide when to filter
`add_knowledge_to_context`	`bool`	No	`False`	Add knowledge to context
`references_format`	`str`	No	`"json"`	Format for references ("json", "markdown", etc.)

Reference: Agno Knowledge Documentation

Tools Fields¶

Field	Type	Required	Default	Description
`tools`	`List[Tool]`	No	`[]`	List of tool instances
`tool_call_limit`	`int`	No	`None`	Maximum tool calls per run
`tool_choice`	`str`	No	`"auto"`	Tool choice strategy: "auto", "required", "none"
`tool_hooks`	`List[Callable]`	No	`None`	Tool execution hooks

Reference: Agno Tools Documentation

Reasoning Fields¶

Field	Type	Required	Default	Description
`reasoning`	`bool`	No	`False`	Enable reasoning mode
`reasoning_model`	`Model`	No	`None`	Model for reasoning (if different from main model)
`reasoning_agent`	`Agent`	No	`None`	Reasoning agent instance
`reasoning_min_steps`	`int`	No	`1`	Minimum reasoning steps (1-100 recommended)
`reasoning_max_steps`	`int`	No	`10`	Maximum reasoning steps (1-100 recommended)

Reference: Agno Reasoning Documentation

Dependencies Fields¶

Field	Type	Required	Default	Description
`dependencies`	`List[Agent]`	No	`None`	List of agent dependencies
`add_dependencies_to_context`	`bool`	No	`False`	Add dependencies to context

Reference: Agno Dependencies Documentation

Context Engineering Fields¶

Field	Type	Required	Default	Description
`build_context`	`bool`	No	`True`	Enable context building
`resolve_in_context`	`bool`	No	`True`	Resolve variables in context
`add_name_to_context`	`bool`	No	`False`	Add agent name to context
`add_datetime_to_context`	`bool`	No	`False`	Add current datetime to context
`add_location_to_context`	`bool`	No	`False`	Add location to context
`timezone_identifier`	`str`	No	`None`	Timezone for datetime context (e.g., "America/New_York")
`additional_context`	`str`	No	`None`	Additional context to include

Reference: Agno Context Engineering Documentation

Hooks & Guardrails Fields¶

Field	Type	Required	Default	Description
`pre_hooks`	`List[Callable]`	No	`None`	Functions to run before agent execution
`post_hooks`	`List[Callable]`	No	`None`	Functions to run after agent execution
`guardrails`	`List[Callable]`	No	`None`	Functions to validate/constrain agent behavior

Reference: - Agno Pre-hooks & Post-hooks Documentation - Agno Guardrails Documentation

Output & Formatting Fields¶

Field	Type	Required	Default	Description
`output_schema`	`Type[BaseModel]`	No	`None`	Pydantic BaseModel for structured output
`input_schema`	`Type[BaseModel]`	No	`None`	Pydantic BaseModel for structured input
`use_json_mode`	`bool`	No	`False`	Use JSON mode for structured output
`markdown`	`bool`	No	`False`	Enable markdown formatting
`parse_response`	`bool`	No	`True`	Parse response

Pre-defined Output Schemas¶

Located in: dto/output/

EarlyExitDecision¶

Output schema for early exit decision analysis. Used with early_exit_agent_config.

Fields: - action: Literal["HARD_EXIT", "SOFT_EXIT", "KEEP_GOING"] (Required) - HARD_EXIT: Must exit the conversation - SOFT_EXIT: Caller decides whether to exit - KEEP_GOING: Continue the conversation - reason: str (Required) - Brief explanation of why exit is or isn't needed - Common values: "EMPTY_INPUT", "HAS_ENDING_PHRASE", "GREETING_MSG", "NO_EXIT", etc.

Usage:

from bast_ai_agno.dto.agent_configs import early_exit_agent_config, EarlyExitDecision

agent = factory.create_openai_chat_agent(
    name="early_exit_agent",
    model=model,
    db=storage,
    config_raw=early_exit_agent_config  # Includes output_schema
)

response = agent.run(input="Thank you!")
decision: EarlyExitDecision = response.content

Advanced Fields¶

Retries & Error Handling¶

Field	Type	Required	Default	Description
`retries`	`int`	No	`0`	Number of retries on failure
`delay_between_retries`	`float`	No	`1.0`	Delay between retries (seconds)
`exponential_backoff`	`bool`	No	`False`	Use exponential backoff for retries

Streaming¶

Field	Type	Required	Default	Description
`stream`	`bool`	No	`None`	Enable streaming (defaults to model setting)
`stream_intermediate_steps`	`bool`	No	`False`	Stream intermediate reasoning steps

Media¶

Field	Type	Required	Default	Description
`send_media_to_model`	`bool`	No	`True`	Send media to model
`store_media`	`bool`	No	`True`	Store media in database

Debugging & Telemetry¶

Field	Type	Required	Default	Description
`debug_mode`	`bool`	No	`False`	Enable debug mode
`debug_level`	`int`	No	`1`	Debug level (1-5)
`telemetry`	`bool`	No	`True`	Enable telemetry

Team & Workflow¶

Field	Type	Required	Default	Description
`role`	`str`	No	`None`	Agent role in team/workflow
`team_id`	`str`	No	`None`	Team ID
`workflow_id`	`str`	No	`None`	Workflow ID

Field Dependencies¶

Required Dependencies¶

add_datetime_to_context=True requires build_context=True
add_dependencies_to_context=True requires dependencies to be set
add_session_summary_to_context=True requires enable_session_summaries=True
Reasoning fields require reasoning=True to be effective

Recommended Combinations¶

Memory: Enable both enable_agentic_memory and enable_user_memories for full memory capabilities
Sessions: Use cache_session=True with search_session_history=True for effective session management
Knowledge: Set search_knowledge=True when providing a knowledge instance
Context: Enable build_context=True when using context engineering features

Validation Rules¶

Value Ranges¶

num_history_runs: 0-100 (recommended)
num_history_sessions: 0-100 (recommended)
reasoning_min_steps: 0-100 (recommended)
reasoning_max_steps: 1-100 (recommended, must be >= min_steps)
retries: 0-10 (recommended)
debug_level: 1-5

Logical Constraints¶

reasoning_max_steps >= reasoning_min_steps
num_history_runs >= 0
num_history_sessions >= 0
If reasoning=True, reasoning_min_steps and reasoning_max_steps should be set

Common Patterns¶

Pattern 1: Basic Agent¶

{
    "instructions": ["You are a helpful assistant."],
    "enable_agentic_memory": False,
    "enable_user_memories": False,
    "add_history_to_context": True,
    "num_history_runs": 3
}

Pattern 2: RAG Agent¶

{
    "instructions": ["Search knowledge base before answering."],
    "search_knowledge": True,
    "enable_agentic_memory": True,
    "enable_user_memories": True,
    "add_memories_to_context": True
}

Pattern 3: Conversational Agent¶

{
    "instructions": ["Maintain conversation context."],
    "enable_agentic_memory": True,
    "enable_user_memories": True,
    "add_history_to_context": True,
    "num_history_runs": 10,
    "cache_session": True,
    "search_session_history": True
}

Pattern 4: Reasoning Agent¶

{
    "instructions": ["Think through problems carefully."],
    "reasoning": True,
    "reasoning_min_steps": 2,
    "reasoning_max_steps": 15,
    "enable_agentic_memory": True
}