Reference

Look up the contract for your SDK and backend. Start with Backend capabilities and data scope when comparing Bolt and NAMS. These pages describe the checked source; verify availability in the SDK artifact you install.

Python API

Adapters reference — Constructor signatures for every LLM adapter shipped in neo4j_agent_memory.llm.adapters.
Provider factory reference — The from_provider string-shorthand factory for embedding and LLM providers.
LLM provider API reference — Protocols, types, factory, and exception hierarchy in neo4j_agent_memory.llm.
LongTermMemory API reference — Entity, preference, fact, and relationship operations exposed through client.long_term.
- Entity and relationship operations — add_entity, search_entities, add_relationship, and the Entity/Relationship models.
- Preference and fact operations — add_preference, search_preferences, add_fact, and the Preference/Fact models.
- Deduplication and provenance — Ingest-time duplicate detection and extraction provenance tracking.
- Geospatial operations and NAMS behavior — Location search, geocoding, get_context, and the NAMS wrapper differences.
MemoryClient API reference — Connects to the configured backend and exposes its memory stores.
ReasoningMemory API reference — Trace, step, and tool-call operations exposed through client.reasoning.
- Trace operations — start_trace, complete_trace, get_trace_with_steps, and related lookups.
- Step, tool-call, and streaming operations — add_step, record_tool_call, and the StreamingTraceRecorder context manager.
- Search, statistics, and NAMS lifecycle — get_similar_traces, get_tool_stats, and the NAMS synthetic-trace lifecycle.
- Models and structured outcomes — ReasoningTrace, ReasoningStep, ToolCall, and TraceOutcome.
ShortTermMemory API reference — Conversation and message operations exposed through client.short_term.
- Conversation and message operations — add_message, add_messages_batch, get_conversation, and the NAMS conversation operations.
- Message search and context — search_messages and get_context.
- Sessions, summaries, and maintenance — list_sessions, get_conversation_summary, migrate_message_links, and related maintenance calls.

Configuration, service and SDK contracts

Configuration and CLI

Backend capabilities and data scope — Which operations and data are available on Bolt versus NAMS.
Configuration reference — MemorySettings fields, defaults, precedence, and provider shapes.
- Neo4j and NAMS connection settings — Neo4jConfig and NamsConfig.
- Embedding, LLM, schema, extraction, and resolution settings — EmbeddingConfig, LLMConfig, SchemaConfig, ExtractionConfig, ResolutionConfig.
- Memory, search, geocoding, and enrichment settings — MemoryConfig, SearchConfig, GeocodingConfig, EnrichmentConfig, and the integration-limits summary.
- Deduplication, observability, and CLI settings — The DeduplicationConfig dataclass and how it follows ResolutionConfig, the observability factory, and CLI configuration.
Environment variables reference — Every NAM_ environment variable MemorySettings and its integrations recognize.
CLI reference — Commands for local extraction, bolt schema/statistics, and MCP serving.
Self-hosted Python MCP reference — Tool parameters, resources, and prompts for the self-hosted MCP server.
Documentation publication — How this documentation site is built and published.
Deploy a memory application — Compares deployment paths for an application built on the SDKs.
Extractor classes reference — Local Python extraction classes and result models.
- Built-in extractor classes — SpacyEntityExtractor, GLiNER2Extractor, and LLMEntityExtractor.
- Relation extractors — How GLiNER2.5 and the LLM extractor produce typed relations, and how validate_relations checks them against an ontology.
- Extraction pipeline — ExtractionPipeline, merge strategies, and the pipeline/batch result models.
- Extractor builder — ExtractorBuilder fluent construction.
- Streaming extraction — StreamingExtractor for chunked long-document extraction.
Domain schemas reference — Built-in domain schemas and custom schema creation.
- Built-in domain schemas — The eight registered GLiNER2.5 schemas, the relationships their ontology templates declare, and how to select one.
- Custom schemas and persistence — DomainSchema and OntologyDocument, the Bolt-only SchemaManager, and the label-to-POLE+O type mapping.
- SchemaManager persistence API — save_schema, load_schema, list_schemas, and the rest of the CRUD surface.
Schema objects reference — Constraints, regular, vector, and point indexes the bolt backend creates on connection.

Hosted service

Hosted NAMS MCP server — The MCP server the hosted NAMS backend ships, independent of the SDK.
Authentication and API keys — How NAMS authenticates requests and where API keys come from.
REST API reference — The NAMS REST API endpoints at /v1.
NAMS limits and behavior — Operational limits, quotas, and runtime behavior for the hosted backend.
Ontology API reference — client.ontology on both backends: NamsOntology on NAMS and BoltOntology on your own Neo4j, with the differences between them.
Skills API (Preview) — The NAMS-hosted Agent Skills REST and MCP surface.

TypeScript

TypeScript SDK API reference — The shape of the public surface of @neo4j-labs/agent-memory, linking into the TypeDoc reference.
NAMS AI provider API reference — Hand-maintained reference for the separate @neo4j-labs/nams-ai-provider package: exports, options, and defaults per mode.