Skip to main content
Version: 2.0

Meta agents module - Debugger and other meta-agent utilities.

This module provides meta-agents that operate on top of other agents, such as the Debugger for analyzing agent responses.

Example usage: from aixplain import Aixplain

Initialize the client

aix = Aixplain("<api_key>")

Standalone usage

debugger = aix.Debugger() result = debugger.run("Analyze this agent output: ...")

Or with custom prompt

result = debugger.run(content="...", prompt="Focus on error handling")

From agent response (chained)

agent = aix.Agent.get("my_agent_id") response = agent.run("Hello!") debug_result = response.debug() # Uses default prompt debug_result = response.debug("Why did it take so long?") # Custom prompt

DebugResult Objects​

@dataclass_json

@dataclass
class DebugResult(Result)

[view_source]

Result from running the Debugger meta-agent.

Attributes:

  • data - The debugging analysis output.
  • session_id - Session ID for conversation continuity.
  • request_id - Request ID for tracking.
  • used_credits - Credits consumed by the debugging operation.
  • run_time - Time taken to run the debugging analysis.
  • analysis - The main debugging analysis text (extracted from data output).

analysis​

@property
def analysis() -> Optional[str]

[view_source]

Extract the debugging analysis text from the result data.

Returns:

The analysis text if available, None otherwise.

Debugger Objects​

class Debugger()

[view_source]

Meta-agent for debugging and analyzing agent responses.

The Debugger uses a pre-configured aiXplain agent to provide insights into agent runs, errors, and potential improvements.

Attributes:

  • context - The Aixplain client context for API access.

Example:

Create a debugger through the client

aix = Aixplain("<api_key>") debugger = aix.Debugger()

Analyze content directly

result = debugger.run("Agent returned: 'Error 500'")

Debug an agent response

agent_result = agent.run("Hello!") debug_result = debugger.debug_response(agent_result)

__init__​

def __init__() -> None

[view_source]

Initialize the Debugger.

The context is set as a class attribute by the Aixplain client when creating the Debugger class dynamically.

run​

def run(content: Optional[str] = None,
        prompt: Optional[str] = None,
        **kwargs: Any) -> DebugResult

[view_source]

Run the debugger on provided content.

This is the standalone usage mode where you can analyze any content or agent output directly.

Arguments:

  • content - The content to analyze/debug. Can be agent output, error messages, or any text requiring analysis.
  • prompt - Optional custom prompt to guide the debugging analysis. If not provided, uses a default debugging prompt.
  • **kwargs - Additional parameters to pass to the underlying agent.

Returns:

  • DebugResult - The debugging analysis result.

Example:

debugger = aix.Debugger() result = debugger.run("Agent returned: 'Error 500'") print(result.analysis)

debug_response​

def debug_response(response: "AgentRunResult",
                   prompt: Optional[str] = None,
                   execution_id: Optional[str] = None,
                   **kwargs: Any) -> DebugResult

[view_source]

Debug an agent response.

This method is designed to analyze AgentRunResult objects to provide insights into what happened during the agent execution.

Arguments:

  • response - The AgentRunResult to analyze.
  • prompt - Optional custom prompt to guide the debugging analysis.
  • execution_id - Optional execution ID override. If not provided, will be extracted from the response's request_id or poll URL. The execution_id allows the debugger to fetch additional information like logs from the backend.
  • **kwargs - Additional parameters to pass to the underlying agent.

Returns:

  • DebugResult - The debugging analysis result.

Example:

agent_result = agent.run("Hello!") debug_result = debugger.debug_response(agent_result, prompt="Why is it slow?")

Or with explicit execution ID

debug_result = debugger.debug_response(agent_result, execution_id="abc-123")