Agent API Reference¶

Agent¶

Config-driven agent that auto-discovers handlers from YAML configuration.

from flowgentra_ai import Agent

Class Methods¶

Method	Returns	Description
`Agent.from_config_path(path)`	`Agent`	Create from a YAML config file
`Agent.from_config(config)`	`Agent`	Create from an `AgentConfig` object

Properties¶

Property	Type	Description
`state`	`SharedState`	The agent's current state
`config`	`AgentConfig`	The agent's configuration
`name`	`str`	Agent name from config

Methods¶

Method	Returns	Description
`run()`	`SharedState`	Execute the agent, returns final state
`run_with_thread(thread_id)`	`SharedState`	Execute with checkpointing
`set_state(key, value)`	`None`	Set a value in the agent's state

AgentConfig¶

Agent configuration loaded from YAML.

from flowgentra_ai import AgentConfig

Class Methods¶

Method	Returns	Description
`AgentConfig.from_file(path)`	`AgentConfig`	Load from a YAML file
`AgentConfig.from_yaml(yaml_str)`	`AgentConfig`	Load from a YAML string

Properties¶

Property	Type	Description
`name`	`str`	Agent name
`description`	`str \\| None`	Agent description

Methods¶

Method	Returns	Description
`validate()`	`None`	Validate the configuration (raises on error)
`to_json()`	`str`	Serialize to JSON string

AgentBuilder¶

Builder for creating prebuilt agents (ReAct, Conversational).

from flowgentra_ai import AgentBuilder, AgentType

Constructor¶

AgentBuilder(agent_type: AgentType)

Methods¶

Method	Description
`with_name(name)`	Set agent name
`with_llm_config(model)`	Set model (e.g., `"gpt-4"`)
`with_temperature(temp)`	Set temperature
`with_max_tokens(tokens)`	Set max tokens
`with_tool(tool_spec)`	Add a tool
`with_system_prompt(prompt)`	Set system prompt
`with_memory_steps(steps)`	Set memory window size
`with_evaluation()`	Enable evaluation
`with_retries(max_retries)`	Set retry count
`with_param(key, value)`	Set a custom parameter
`build_graph()`	Build into a `GraphBasedAgent`

AgentType¶

Prebuilt agent type selector.

from flowgentra_ai import AgentType

Method	Description
`AgentType.zero_shot_react()`	Reasoning + Action without examples
`AgentType.few_shot_react()`	Reasoning + Action with example demonstrations
`AgentType.conversational()`	Multi-turn dialogue with memory

GraphBasedAgent¶

A built prebuilt agent. Created by AgentBuilder.build_graph().

Methods¶

Method	Returns	Description
`execute_input(input)`	`str`	Execute with text input, returns text response
`node_names()`	`list[str]`	Get node names of the underlying graph

Properties¶

Property	Type	Description
`name`	`str`	Agent name

ToolSpec¶

Tool specification for prebuilt agents.

from flowgentra_ai import ToolSpec

Constructor¶

ToolSpec(name: str, description: str)

Methods¶

Method	Description
`add_parameter(name, param_type)`	Add a parameter (type as string: `"string"`, `"number"`, etc.)
`set_required(param_name)`	Mark a parameter as required

Properties¶

Property	Type	Description
`name`	`str`	Tool name
`description`	`str`	Tool description

MemoryAwareAgent¶

Agent with automatic conversation memory management.

from flowgentra_ai import MemoryAwareAgent

Class Methods¶

Method	Returns	Description
`MemoryAwareAgent.from_config(path)`	`MemoryAwareAgent`	Create from YAML config

Methods¶

Method	Returns	Description
`set_thread_id(thread_id)`	`None`	Set conversation thread
`thread_id()`	`str`	Get current thread ID
`run_turn(input)`	`str`	Run a conversation turn
`clear_memory()`	`None`	Clear memory for current thread
`memory_stats()`	`MemoryStats`	Get memory usage stats

MemoryStats¶

Memory usage statistics.

Properties¶

Property	Type	Description
`message_count`	`int`	Total messages stored
`user_messages`	`int`	User message count
`assistant_messages`	`int`	Assistant message count
`approximate_tokens`	`int`	Approximate token count

StateField¶

A field in the state schema (used in config).

from flowgentra_ai import StateField

Constructor¶

StateField(field_type: str, description: str)

Properties¶

Property	Type	Description
`field_type`	`str`	Type of the field
`description`	`str`	Field description