Skip to content

Python SDK for AI agent monitoring, LLM cost tracking, benchmarking, and more. Integrates with most LLMs and agent frameworks including CrewAI, Agno, OpenAI Agents SDK, Langchain, Autogen, AG2, and CamelAI

3.4 KB
Raw Blame
---
title: 'Recording Events'
description: 'Record Events using the AgentOps SDK'
---

To get the most out of AgentOps, it is best to carefully consider what events to record - 
not simply record whatever your agent is logging. AgentOps offers two ways to record events 
outside of logs.


## `record_action` Decorator
To make AgentOps easier to integrate, we also provide a function decorator to automatically creates 
and records an event for your function.

```python python
@ao_client.record_action('sample function being record')
def sample_function(...):
    ...
```

The decorator will record the function's parameters, returns, and the time duration. We suggest using this on functions that take a long time and contain nested functions. For example, if you decorate a function that makes several OpenAI calls, then each openai call will show in the replay graph as a child of the decorated function.


record_action:

Arguments
        * event_type (str): Type of the event
        * tags (List[str], optional): Any tags associated with the event. Defaults to None.


## `record()` Method
From this point, simply call the .record() method in the AgentOps client:

```python python
ao_client.record(Event("event_type1"))
```

In AgentOps, each session is associated with a number of "Events". Events have 
must have an "event_type" which is any arbitrary string of your choice. It might be something
like "OpenAI Call". Events can also have other information such as the parameters of the operation,
the returned data, alongside tags, etc.


Events:
Represents a discrete event to be recorded.

Args:
        event_type (str): Type of the event, e.g., "API Call". Required.
        params (Optional[Dict[str, Any]], optional): The parameters passed to the operation. Defaults to None.
        returns (str, optional): The output of the operation. Defaults to None.
        result (str, optional): Result of the operation, e.g., "Success", "Fail", "Indeterminate". Defaults to "Indeterminate".
        action_type (str, optional): Type of action of the event e.g. 'action', 'llm', 'api', 'screenshot'. Defaults to 'action'.
        model (Models, optional): The model used during the event if an LLM is used (i.e. GPT-4).
            For models, see the types available in the Models enum. 
            If a model is set but an action_type is not, the action_type will be coerced to 'llm'. 
Arguments
        * event_type (str): Type of the event, e.g., "API Call". Required.
        * params (str, optional): The parameters passed to the operation.
        * returns (str, optional): The output of the operation.
        * result (EventState, optional): Result of the operation, Enum: "Success", "Fail", "Indeterminate".
        * action_type (ActionType, optional): Type of action of the event e.g. 'action', 'llm', 'api'
        * model (Models, optional): The model used during the event if an LLM is used (i.e. GPT-4).
            For models, see the types available in the Models enum. 
            If a model is set but an action_type is not, the action_type will be coerced to 'llm'. 
            Defaults to None.
        * prompt (str, optional): The input prompt for an LLM call when an LLM is being used.
        * tags (List[str], optional): Tags that can be used for grouping or sorting later. e.g. ["my_tag"].


Attributes
        * timestamp (float): The timestamp for when the event was created, represented as seconds since the epoch.