Skip to main content
The ModelResponse class contains the response from a language model, including content, tool calls, media outputs, and metrics.

​
Core Fields

​
role
str
default:"None"
Role of the response message.
​
content
Any
default:"None"
Text content of the response.
​
parsed
Any
default:"None"
Parsed structured output (when using output_schema).

​
Media Outputs

​
audio
Audio
default:"None"
Audio generated by the model.
​
images
List[Image]
default:"None"
Images generated by LLM or tools.
​
videos
List[Video]
default:"None"
Videos generated by tools.
​
audios
List[Audio]
default:"None"
Multiple audio outputs from tools.
​
files
List[File]
default:"None"
Files generated by tools.

​
Tool Calls

​
tool_calls
List[Dict[str, Any]]
Model tool calls (function calls to execute).
​
tool_executions
List[ToolExecution]
Actual tool execution results.

​
Reasoning

​
reasoning_content
str
default:"None"
Reasoning/thinking content from the model.
​
redacted_reasoning_content
str
default:"None"
Redacted reasoning content.

​
Metadata

​
event
str
default:"assistant_response"
Event type: “assistant_response”, “tool_call_started”, etc.
​
provider_data
Dict[str, Any]
default:"None"
Provider-specific metadata.
​
citations
Citations
default:"None"
Citations from the model.
​
response_usage
MessageMetrics
default:"None"
Token usage and timing metrics.
​
created_at
int
Unix timestamp of response creation.

​
State Updates

​
updated_session_state
Dict[str, Any]
default:"None"
Session state updates from tool executions.

​
Compression

​
compression_stats
Dict[str, Any]
default:"None"
Statistics about context compression if applied.

​
Model Metrics

​
input_tokens
int
default:"None"
Number of input tokens.
​
output_tokens
int
default:"None"
Number of output tokens.
​
total_tokens
int
default:"None"
Total token count.
​
time_to_first_token
float
default:"None"
Time to first token in seconds.
​
reasoning_tokens
int
default:"None"
Number of reasoning tokens.
​
cache_read_tokens
int
default:"None"
Tokens read from cache.
​
cache_write_tokens
int
default:"None"
Tokens written to cache.

​
Methods

​
to_dict()

Serialize response to dictionary. 
data = response.to_dict()

​
from_dict()

Deserialize response from dictionary. 
response = ModelResponse.from_dict(data)

​
ToolExecution

Represents the execution of a tool.
​
tool_call_id
str
default:"None"
ID of the tool call.
​
tool_name
str
default:"None"
Name of the tool executed.
​
tool_args
Dict[str, Any]
default:"None"
Arguments passed to the tool.
​
result
str
default:"None"
Result returned by the tool.
​
tool_call_error
bool
default:"None"
Whether the tool call resulted in an error.
​
child_run_id
str
default:"None"
If the tool created a run (agent/team/workflow), its run ID.

​
HITL Fields

​
requires_confirmation
bool
default:"None"
Whether the tool requires user confirmation.
​
confirmed
bool
default:"None"
Whether the tool has been confirmed.
​
requires_user_input
bool
default:"None"
Whether the tool requires user input.
​
external_execution_required
bool
default:"None"
Whether the tool requires external execution.

​
Example Usage

response = agent.run("What's 2+2?")

print(response.content)  # "The answer is 4."
print(response.response_usage.input_tokens)
print(response.response_usage.output_tokens)

Build docs developers (and LLMs) love

Get started for freeTalk to us