cognitive_settings
Cognitive settings module
Classes
AnalyzeSettings
Bases: BaseDataModel
Analyze service settings
Attributes:
-
collection_names
(List[str]
) –List of collection names to to be used as memory sources when building definitions
-
top_k_vectors
(int
) –Soft maximum number of vectors to return per memory collection. This is a soft limit since the number of vectors requested is divided by the number of context chunks.
-
relevance_alpha
(float
) –Relevance alpha
-
importance_alpha
(float
) –Importance alpha
-
recency_alpha
(float
) –Recency alpha
-
min_score
(float
) –Minimum score needed to be considered for recall
-
max_memories
(int
) –Max memories to recall
-
max_memory_strategy
(str
) –Max memory strategy used for recall operation
Attributes
collection_names
class-attribute
instance-attribute
collection_names: List[str] = Field(
...,
title="Collection Names",
description="List of collection names to to be used as memory sources when building definitions",
)
importance_alpha
class-attribute
instance-attribute
importance_alpha: float = Field(
default=0.2,
le=1.0,
ge=0.0,
title="Importance Alpha",
description="Importance alpha",
)
max_memories
class-attribute
instance-attribute
max_memories: int = Field(
default=100,
gt=0,
title="Max Memories",
description="Max memories to recall",
)
max_memory_strategy
class-attribute
instance-attribute
max_memory_strategy: str = Field(
default="HARD_LIMIT",
title="Max Memory Strategy",
description="Max memory strategy used for recall operation",
)
min_score
class-attribute
instance-attribute
min_score: float = Field(
default=0.6,
le=1.0,
gt=0.0,
title="Minimum Score",
description="Minimum score needed to be considered for recall",
)
recency_alpha
class-attribute
instance-attribute
recency_alpha: float = Field(
default=0.0,
le=1.0,
ge=0.0,
title="Recency Alpha",
description="Recency alpha",
)
relevance_alpha
class-attribute
instance-attribute
relevance_alpha: float = Field(
default=0.8,
le=1.0,
ge=0.0,
title="Relevance Alpha",
description="Relevance alpha",
)
top_k_vectors
class-attribute
instance-attribute
top_k_vectors: int = Field(
default=200,
gt=0,
title="Top K Vectors",
description="Soft maximum number of vectors to return per memory collection. This is a soft limit since the number of vectors requested is divided by the number of context chunks.",
)
CognitiveSettings
Bases: BaseModel
Eleanor framework cognitive settings
Attributes:
-
add_memories_vectorize_timeout
(int
) –When invoking the add_memories service operation, this is a total timeout value for the vectorization component.
-
analyze_settings
(AnalyzeSettings
) –Analyze service settings
-
max_integration_attempts
(int
) –The maximum number of tries the integration process will attempt on memories created in an agent/session before giving up. This will prevent continuous failures for session that contain corrupt data that for whatever reason cannot complete the integration process. One observed instance is described in #143. Note that once the session turn limit increases to trigger integration eligibility again, another attempt will be made.
-
chains
(Dict[str, IoCFactoryModel]
) –Chain definitions used by the framework. This setting is lazy loaded on startup and validation + initialization is
Attributes
add_memories_vectorize_timeout
class-attribute
instance-attribute
add_memories_vectorize_timeout: int = Field(
default=90,
gt=0,
title="Add Memories Vectorize Timeout",
description="When invoking the add_memories service operation, this is a total timeout value for the vectorization component.",
)
analyze_settings
class-attribute
instance-attribute
analyze_settings: AnalyzeSettings = Field(
..., title="Analyze Settings", description=__doc__
)
chains
class-attribute
instance-attribute
chains: Dict[str, IoCFactoryModel] = Field(
default_factory=dict,
title="Chains",
description="Chain definitions used by the framework. This setting is lazy loaded on startup and validation + initialization is performed ",
)
max_integration_attempts
class-attribute
instance-attribute
max_integration_attempts: int = Field(
default=3,
title="Maximum Integration Attempts",
description="The maximum number of tries the integration process will attempt on memories created in an agent/session before giving up. This will prevent continuous failures for session that contain corrupt data that for whatever reason cannot complete the integration process. One observed instance is described in #143. Note that once the session turn limit increases to trigger integration eligibility again, another attempt will be made.",
)