redis_openai_agents.RedisCachingModel#

class RedisCachingModel(model, redis_url='redis://localhost:6379', cache_prefix='model_cache', cache_ttl=3600, enable_semantic_cache=False, semantic_threshold=0.95)[source]#

Caching wrapper for OpenAI Agents SDK Model interface.

Wraps any Model implementation and adds 2-level caching: - Level 1: Exact match using hash of (system_instructions, input) - Level 2: Semantic similarity using vector embeddings (optional)

Caching is bypassed when: - Tools are provided (responses may depend on tool calls) - Handoffs are provided (complex agent interactions) - Output schema is provided (structured output validation)

_model#

The underlying model being wrapped.

cache_ttl#

Time-to-live for cache entries in seconds.

enable_semantic_cache#

Whether to use semantic similarity caching.

semantic_threshold#

Minimum similarity score for semantic cache hit.

Initialize the caching model wrapper.

Parameters:

  • model (Any) – The underlying Model to wrap.

  • redis_url (str) – Redis connection URL.

  • cache_prefix (str) – Prefix for cache keys in Redis.

  • cache_ttl (int) – Time-to-live for cache entries in seconds.

  • enable_semantic_cache (bool) – Enable Level 2 semantic caching.

  • semantic_threshold (float) – Minimum similarity for semantic cache hit.

__init__(model, redis_url='redis://localhost:6379', cache_prefix='model_cache', cache_ttl=3600, enable_semantic_cache=False, semantic_threshold=0.95)[source]#

Initialize the caching model wrapper.

Parameters:

  • model (Any) – The underlying Model to wrap.

  • redis_url (str) – Redis connection URL.

  • cache_prefix (str) – Prefix for cache keys in Redis.

  • cache_ttl (int) – Time-to-live for cache entries in seconds.

  • enable_semantic_cache (bool) – Enable Level 2 semantic caching.

  • semantic_threshold (float) – Minimum similarity for semantic cache hit.

Return type:

None

Methods

__init__(model[, redis_url, cache_prefix, ...])

Initialize the caching model wrapper.

check_cache(system_instructions, input_data)

Check if response is in cache.

close()

Close Redis connection.

get_metrics()

Get cache performance metrics.

get_response(system_instructions, input, ...)

Get a response, checking cache first.

initialize()

Initialize Redis connection and semantic cache if enabled.

stream_response(system_instructions, input, ...)

Stream a response from the model.