corpus_reader_settings
Corpus reader settings module
Classes
CorpusReaderSettings
Bases: BaseDataModel
Corpus reader configuration settings
Attributes:
-
source
(str
) –Source Identifier
-
default_namespace
(str
) –Default namespace
-
default_user_name
(str
) –Default User
-
default_agent_name
(str
) –Default Agent
-
system_message_map
(Dict[str, str]
) –System Message Map
-
new_collection_vectordb
(str
) –New Collection VectorDB Name
-
session_defaults
(SessionResourceSettings
) –Session Defaults
-
splitter
(SplitterSettings
) –Splitter Settings
-
dynamic_session
(DynamicSessionSettings
) –Dynamic Session Settings
-
resume_info_path
(str | None
) –Resume Info Path
-
unbind_other_collections
(bool
) –Unbind Other Collections
-
chunk_retry_limit
(int
) –Chunk Retry Limit
Attributes
chunk_retry_limit
class-attribute
instance-attribute
chunk_retry_limit: int = Field(
default=3,
title="Chunk Retry Limit",
description="Number of times to retry a chunk operation that fails to process",
)
default_agent_name
class-attribute
instance-attribute
default_agent_name: str = Field(
...,
title="Default Agent",
description="Agent that is selected when one is not provided via the CLI",
)
default_namespace
class-attribute
instance-attribute
default_namespace: str = Field(
"default",
title="Default namespace",
description="Namespace that is selected when one is not provided via the CLI",
)
default_user_name
class-attribute
instance-attribute
default_user_name: str = Field(
...,
title="Default User",
description="User that is selected when one is not provided via the CLI",
)
dynamic_session
class-attribute
instance-attribute
dynamic_session: DynamicSessionSettings = Field(
default_factory=DynamicSessionSettings,
title="Dynamic Session Settings",
description=__doc__,
)
new_collection_vectordb
class-attribute
instance-attribute
new_collection_vectordb: str = Field(
default="default",
title="New Collection VectorDB Name",
description="When auto-creating new memory collections, this is the name/reference to the vectordb configuration. See settings.yaml for a list of available VectorDB configurations.",
)
resume_info_path
class-attribute
instance-attribute
resume_info_path: str | None = Field(
default=None,
title="Resume Info Path",
description="Path to a file where resume information is written when an error is raised during the corpus read process.",
)
session_defaults
class-attribute
instance-attribute
session_defaults: SessionResourceSettings = Field(
default_factory=SessionResourceSettings,
title="Session Defaults",
description="Session settings to use for corpus reading operations",
)
source
class-attribute
instance-attribute
source: str = Field(
...,
title="Source Identifier",
description="A freeform string that is added to the Eleanor Framework header to identify the source system when using the chat API",
)
splitter
class-attribute
instance-attribute
splitter: SplitterSettings = Field(
default_factory=SplitterSettings,
title="Splitter Settings",
description=__doc__,
)
system_message_map
class-attribute
instance-attribute
system_message_map: Dict[str, str] = Field(
default_factory=dict,
title="System Message Map",
description="Top-level system message to use. This dictionary is keyed by the namespace name which - by convention - is the same as the model name on the Eleanor Framework chat API. This allows users to map specific system messages per LLM.",
)
unbind_other_collections
class-attribute
instance-attribute
unbind_other_collections: bool = Field(
default=False,
title="Unbind Other Collections",
description="Then true, all other memory collections will be unbound from the agent to prevent conceptual cross-over. This may or may not be desirable depending on the use case.",
)
DynamicSessionSettings
Bases: BaseDataModel
Dynamic session settings configuration.
Attributes:
-
enabled
(bool
) –Flag that when true enables dynamic session creation
-
score_dividend
(int
) –Score dividend
-
spacy_model
(str
) –SpaCy model to used when determining chunk information density
Attributes
enabled
class-attribute
instance-attribute
score_dividend
class-attribute
instance-attribute
SplitterSettings
Bases: BaseDataModel
Corpus splitter configuration settings
Attributes:
-
sample_packing_enabled
(bool
) –Flag that when true enables sample packing
-
sample_packing_tokenizer
(str
) –Reference to a LLM tokenizer to use for sample packing. This tokenizer must be attached to a LLM that is loaded by the framework. Required whenever sample packing is enabled
-
splitter_kwargs
(KwargsModel
) –Additional keyword arguments to pass to the splitter when creating new sessions. This is applicable whether or not sample packing is enabled.
Attributes
sample_packing_enabled
class-attribute
instance-attribute
sample_packing_enabled: bool = Field(
default=False,
title="Sample Packing Enabled",
description="Flag that when true enables sample packing",
)
sample_packing_tokenizer
class-attribute
instance-attribute
sample_packing_tokenizer: str = Field(
default="",
title="Sample Packing Tokenizer",
description="Reference to a LLM tokenizer to use for sample packing. This tokenizer must be attached to a LLM that is loaded by the framework. Required whenever sample packing is enabled",
)
splitter_kwargs
class-attribute
instance-attribute
splitter_kwargs: KwargsModel = Field(
default_factory=KwargsModel,
title="Splitter Kwargs",
description="Additional keyword arguments to pass to the splitter when creating new sessions. This is applicable whether or not sample packing is enabled.",
)