redis_openai_agents.RedisAgentRunner#

class RedisAgentRunner(cache=None, metrics=None, session=None)[source]#

Integrated runner with automatic caching, metrics, and session management.

This class provides a unified interface that combines: - SemanticCache: Check/store LLM responses - AgentMetrics: Record latency, tokens, cache hits - AgentSession: Store conversation history

Example

>>> from redis_openai_agents import (
...     RedisAgentRunner, SemanticCache, AgentMetrics, AgentSession
... )
>>>
>>> runner = RedisAgentRunner(
...     cache=SemanticCache(redis_url="redis://localhost:6379"),
...     metrics=AgentMetrics(name="my_agent"),
...     session=AgentSession(user_id="user_123"),
... )
>>>
>>> result = runner.run(agent, "What is Redis?")
>>> # Cache checked, metrics recorded, session updated automatically
Parameters:

  • cache (SemanticCache | None) – Optional SemanticCache for response caching

  • metrics (AgentMetrics | None) – Optional AgentMetrics for observability

  • session (AgentSession | None) – Optional AgentSession for conversation history

Initialize the runner with Redis components.

Parameters:

  • cache (SemanticCache | None) – Optional SemanticCache for response caching

  • metrics (AgentMetrics | None) – Optional AgentMetrics for observability

  • session (AgentSession | None) – Optional AgentSession for conversation history

__init__(cache=None, metrics=None, session=None)[source]#

Initialize the runner with Redis components.

Parameters:

  • cache (SemanticCache | None) – Optional SemanticCache for response caching

  • metrics (AgentMetrics | None) – Optional AgentMetrics for observability

  • session (AgentSession | None) – Optional AgentSession for conversation history

Return type:

None

Methods

__init__([cache, metrics, session])

Initialize the runner with Redis components.

arun(agent, input, **kwargs)

Async version of run() - run agent with automatic caching and metrics.

run(agent, input, **kwargs)

Run an agent with automatic caching and metrics.

run_streamed(agent, input, **kwargs)

Run an agent with streaming (no caching for streamed responses).