Extra configuration to pass to the LLM when instantiated.

Note: This will be validated against the LLM config's config_schema.

The IDs of the tool configs associated with the agent.

Must be provided if bound_tool_args are provided.

List of arguments to tools that have been hardcoded.

The order and length of this list is the same as that of tool_configs. The
dictionaries in this list contain key-value pairs, where the key is the name
of a tool argument and the value is what should be hardcoded for that tool.
If the dict is empty, then there are no hardcoded values for the tool.

Must be provided if tool_config_ids are provided.

The published status of the agent.

The input schema for the agent.

Specifies the inputs that the agents require conforming to the JSON schema
standard.
ref: https://json-schema.org/

Note that name of parameters are used to inject dependencies into the tools.
If a parameter name matches the argument name of a tool, then it will be
injected into the tool. It is up to the developer to ensure that the names
match those required by the tools. The names are not validated.

If the parameters don't match the arguments of the tools, then the LLM may
still be able to manually include them in tool calls.

The output schema for the agent.

Specifies the output that the agents output conforming to the JSON schema
standard.
ref: https://json-schema.org/

If None, the agent will output a string.