Sections

Section builders for MemoryManagerConfigBuilder.

This module contains the fluent subsection builders used by the public memory manager builder.

MemoryManagerEmbeddingBuilder(root_builder)

Configure embedding sections for MemoryManagerConfigBuilder.

Attributes:

Name	Type	Description
_root_builder	RootBuilderProtocol	Root builder used for fluent writes.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

openai(*, model=DEFAULTS.openai_embedder_model, api_key=None)

Configure OpenAI embedding provider.

Parameters:

Name	Type	Description	Default
model	str	OpenAI embedding model name. Defaults to DEFAULTS.openai_embedder_model.	openai_embedder_model
api_key	str | None	Optional OpenAI API key. Defaults to None.	None

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

register(em_invoker, *, embedding_dims=None, invoke_kwargs=None, **kwargs)

Register one gllm_inference embedding invoker.

Parameters:

Name	Type	Description	Default
em_invoker	Any	Embedding invoker instance.	required
embedding_dims	int | None	Optional embedding dimension count. Defaults to None.	None
invoke_kwargs	dict[str, Any] | None	Optional invoke kwargs. Defaults to None.	None
**kwargs	Any	Deprecated or unexpected keyword arguments.	{}

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

Raises:

Type	Description
ValueError	If em_invoker is None.
TypeError	If one deprecated or unexpected keyword argument is provided.

MemoryManagerKnowledgeGraphBuilder(root_builder)

Knowledge-graph builder for MemoryManagerConfigBuilder.

Attributes:

Name	Type	Description
text_to_graph	MemoryManagerKnowledgeGraphTextToGraphBuilder	Text-to-graph section builder.
graph_data_store	MemoryManagerKnowledgeGraphDataStoreBuilder	Graph data store section builder.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

enable(*, lm_invoker, graph_store)

Configure the common KG happy path with one LM invoker and one Neo4j store.

Parameters:

Name	Type	Description	Default
lm_invoker	Any	Language-model invoker used for KG extraction.	required
graph_store	Neo4jGraphStoreConfig	Caller-facing Neo4j graph-store config object.	required

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

Raises:

Type	Description
ValueError	If lm_invoker or graph_store is missing.
TypeError	If graph_store is not a Neo4jGraphStoreConfig instance.

MemoryManagerKnowledgeGraphDataStoreBuilder(root_builder)

Configure graph data store settings for knowledge-graph integration.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

neo4j(*, uri, user, password, max_connection_pool_size=DEFAULTS.knowledge_graph_neo4j_max_connection_pool_size)

Configure a Neo4j graph data store.

Parameters:

Name	Type	Description	Default
uri	str	Neo4j connection URI.	required
user	str	Neo4j username.	required
password	str	Neo4j password.	required
max_connection_pool_size	int	Maximum Neo4j connection pool size. Defaults to 100.	knowledge_graph_neo4j_max_connection_pool_size

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

MemoryManagerKnowledgeGraphTextToGraphBuilder(root_builder)

Configure text-to-graph settings for knowledge-graph integration.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

openai(*, model_id, api_key=None, base_url=None, allowed_nodes=None, allowed_relationships=None, strict_mode=DEFAULTS.knowledge_graph_strict_mode, use_structured_output=DEFAULTS.knowledge_graph_use_structured_output)

Configure text-to-graph extraction using an explicit OpenAI LM invoker.

Parameters:

Name	Type	Description	Default
model_id	str	OpenAI model name for KG extraction.	required
api_key	str | None	Optional OpenAI API key. Defaults to None.	None
base_url	str | None	Optional OpenAI-compatible base URL. Defaults to None.	None
allowed_nodes	list[str] | None	Optional allowed node labels. Defaults to None.	None
allowed_relationships	list[str] | list[tuple[str, str, str]] | None	Optional allowed relationship constraints. Defaults to None.	None
strict_mode	bool	Whether extraction should apply strict filtering. Defaults to True.	knowledge_graph_strict_mode
use_structured_output	bool	Whether structured model output should be requested. Defaults to False.	knowledge_graph_use_structured_output

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

register(lm_invoker, *, model_id=None, credentials=None, config=None, allowed_nodes=None, allowed_relationships=None, strict_mode=DEFAULTS.knowledge_graph_strict_mode, use_structured_output=DEFAULTS.knowledge_graph_use_structured_output)

Register one LM invoker-backed text-to-graph configuration.

Parameters:

Name	Type	Description	Default
lm_invoker	Any	Language-model invoker used by the graph transformer.	required
model_id	str | None	Optional provider-qualified model identifier. Defaults to None.	None
credentials	str | dict[str, Any] | None	Optional model credentials payload. Defaults to None.	None
config	dict[str, Any] | None	Optional model runtime config. Defaults to None.	None
allowed_nodes	list[str] | None	Optional allowed node labels. Defaults to None.	None
allowed_relationships	list[str] | list[tuple[str, str, str]] | None	Optional allowed relationship constraints. Defaults to None.	None
strict_mode	bool	Whether extraction should apply strict filtering. Defaults to True.	knowledge_graph_strict_mode
use_structured_output	bool	Whether structured model output should be requested. Defaults to False.	knowledge_graph_use_structured_output

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

Raises:

Type	Description
ValueError	If lm_invoker is None.

MemoryManagerLLMBuilder(root_builder)

Configure LLM sections for MemoryManagerConfigBuilder.

Attributes:

Name	Type	Description
_root_builder	RootBuilderProtocol	Root builder used for fluent writes.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

register(lm_invoker, *, model=None, temperature=0.1, invoke_kwargs=None)

Register one gllm_inference LM invoker.

This path is kept for direct-invoker compatibility. New integrations that need prompt ownership or fallback behavior should prefer register_component(...) and build the LMComponent on the caller side.

Parameters:

Name	Type	Description	Default
lm_invoker	Any	Language-model invoker instance.	required
model	str | None	Optional model name override. Defaults to None.	None
temperature	float	Default model temperature. Defaults to 0.1.	0.1
invoke_kwargs	dict[str, Any] | None	Optional invoke kwargs. Defaults to None.	None

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

Raises:

Type	Description
ValueError	If lm_invoker is None.

register_component(lm_component)

Register one memory-compatible LM component runtime object.

This is the preferred path for new integrations. The recommended input is gllm_memory's MemoryLMComponent or another BaseMemoryLMComponent subclass. Legacy raw LMComponent objects remain supported for compatibility.

Parameters:

Name	Type	Description	Default
lm_component	Any	LM component exposing a callable run method.	required

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

Raises:

Type	Description
ValueError	If lm_component is missing or invalid.

MemoryManagerRerankerBuilder(root_builder)

Configure reranker sections for MemoryManagerConfigBuilder.

Attributes:

Name	Type	Description
_root_builder	RootBuilderProtocol	Root builder used for fluent writes.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

llm_reranker(*, llm_provider=PROVIDERS.openai, model=None, api_key=None, top_k=None, temperature=0.0, max_tokens=100, scoring_prompt=None, llm_config=None)

Configure an LLM-based reranker.

Parameters:

Name	Type	Description	Default
llm_provider	str	Reranker LLM provider name. Defaults to PROVIDERS.openai.	openai
model	str | None	Optional model name override. Defaults to None.	None
api_key	str | None	Optional provider API key. Defaults to None.	None
top_k	int | None	Optional reranker output limit. Defaults to None.	None
temperature	float	Reranker model temperature. Defaults to 0.0.	0.0
max_tokens	int	Maximum token budget for reranker generation. Defaults to 100.	100
scoring_prompt	str | None	Optional custom scoring prompt. Defaults to None.	None
llm_config	dict[str, Any] | None	Optional nested LLM config. Defaults to None.	None

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

register(provider, config=None)

Register one generic reranker provider.

Parameters:

Name	Type	Description	Default
provider	str	Provider identifier to register.	required
config	dict[str, Any] | None	Provider config payload. Defaults to None.	None

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

similarity_based(em_invoker, *, top_k=None, invoke_kwargs=None)

Configure one local similarity-based reranker.

Parameters:

Name	Type	Description	Default
em_invoker	Any	Embedding invoker instance used by the reranker.	required
top_k	int | None	Optional final output limit applied after local reranking. Defaults to None.	None
invoke_kwargs	dict[str, Any] | None	Optional keyword arguments forwarded to the embedding invoker during reranking. Defaults to None.	None

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

Raises:

Type	Description
ValueError	If em_invoker is None.

MemoryManagerStoreBuilder(root_builder)

Configure memory-store sections for MemoryManagerConfigBuilder.

Attributes:

Name	Type	Description
_root_builder	RootBuilderProtocol	Root builder used for fluent writes.

Initializes a new instance of the builder.

Parameters:

Name	Type	Description	Default
root_builder	RootBuilderProtocol	Root builder.	required

elasticsearch(*, host, port, collection_name, embedding_model_dims, user=None, password=None, auto_create_index=DEFAULTS.elasticsearch_auto_create_index, use_ssl=DEFAULTS.elasticsearch_use_ssl, verify_certs=DEFAULTS.elasticsearch_verify_certs)

Configure Elasticsearch-backed memory storage.

Parameters:

Name	Type	Description	Default
host	str	Elasticsearch host name or URL prefix.	required
port	int	Elasticsearch port number.	required
collection_name	str	Target Elasticsearch index name.	required
embedding_model_dims	int	Embedding vector dimension count.	required
user	str | None	Optional Elasticsearch username. Defaults to None.	None
password	str | None	Optional Elasticsearch password. Defaults to None.	None
auto_create_index	bool	Whether the index may be created automatically. Defaults to DEFAULTS.elasticsearch_auto_create_index.	elasticsearch_auto_create_index
use_ssl	bool	Whether SSL should be used. Defaults to DEFAULTS.elasticsearch_use_ssl.	elasticsearch_use_ssl
verify_certs	bool	Whether SSL certificates should be verified. Defaults to DEFAULTS.elasticsearch_verify_certs.	elasticsearch_verify_certs

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

register(provider, config=None)

Register one generic memory-store provider.

Parameters:

Name	Type	Description	Default
provider	str	Provider identifier to register.	required
config	dict[str, Any] | None	Provider config payload. Defaults to None.	None

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

RootBuilderProtocol

Bases: Protocol

Define root-builder operations required by subsection builders.

Subsection builders only depend on this minimal protocol so they stay decoupled from the concrete MemoryManagerConfigBuilder implementation.

merge_backend_option(key, value)

Merge one backend option section on the root builder.

Parameters:

Name	Type	Description	Default
key	str	Backend option key.	required
value	dict[str, Any]	Partial backend option payload to merge.	required

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.

set_provider_config(section, provider, config)

Set one provider section on the root builder.

Parameters:

Name	Type	Description	Default
section	str	Target configuration section name.	required
provider	str	Provider identifier for the section.	required
config	dict[str, Any]	Provider-specific config payload.	required

Returns:

Name	Type	Description
Any	Any	Root builder instance for fluent chaining.