FunctionConfig

The FunctionConfig class defines functional behavior configuration for the Agent runtime.

Properties

• use_minimal_context (bool): Default False. Whether to use minimal context, i.e. system prompt + user's last message. Disabling this option uses all context from the message list, which may consume a large amount of Tokens during Agent workflow execution; enabling it may effectively reduce token usage
• no_tokenizer (bool): Default False. Disable built-in tokenizer when response is not returning a token count
• tokenizer_used (str): Default "simple". Tokenizer to use
• agent_tool_call_limit (int): Default 10. Tool call limit when calling tools
• agent_middle_message (bool): Default True. Whether to allow Agent to send intermediate messages to users during tool calling
• agent_mcp_client_enable (bool): Default False. Whether to enable MCP client
• agent_mcp_server_scripts (list[str]): Default []. List of MCP server scripts

Description

The FunctionConfig class inherits from BaseModel and is exposed as AmritaConfig.function_config. It controls runtime behavior of the agent: context usage, tokenizer selection, tool call limits, and MCP client integration.

Example

from amrita_core.config import FunctionConfig

func_config = FunctionConfig(
    tokenizer_used="my_tokenizer",
    agent_tool_call_limit=5,
    agent_mcp_client_enable=True,
    agent_mcp_server_scripts=["path/to/mcp_server.py"],
)