Trajectory Recording¶

Record agent execution trajectories for learning.

Ingesting Trajectories¶

zubbl.ingest(
    trajectory_id="traj-001",
    task_description="Analyze codebase for security issues",
    steps=[
        {"action": "read_file", "tool_name": "file_reader", "tool_output": "..."},
        {"action": "analyze", "tool_name": "ast_parser", "tool_output": "..."},
    ],
    status="success",  # "success" | "failure" | "partial"
    metadata={"tokens": 1240, "duration_ms": 3200},
)

IngestRequest Fields¶

Field	Type	Required	Description
`trajectory_id`	string	Yes	Unique ID for the trajectory
`tenant_id`	string	Auto	Extracted from API key
`task_description`	string	Yes	What the agent was doing
`steps`	list[Step]	Yes	Ordered list of actions taken
`status`	string	Yes	Outcome: success, failure, partial
`metadata`	dict	No	Extra data (tokens, duration, etc.)

Step Schema¶

Field	Type	Description
`action`	string	What the agent did
`tool_name`	string	Tool used (if any)
`tool_input`	string	Input to the tool
`tool_output`	string	Output from the tool
`reasoning`	string	Agent's reasoning (if captured)