Long-term memory storage foundation
Long-term memory is built on LangGraph stores, which save data as JSON documents organized by namespace and key.
LangChain · Agents · all subjects
middleware & retrieval
6 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.
Memory storage organization structure
Each memory in a store is organized under a custom namespace (similar to a folder) and a distinct key (like a file name). Namespaces often include user or org IDs or other labels that make it easier to organize information. This structure enables hierarchical organization of memories. Cross-namespace searching is supported through content filters.
Memory store backends
Long-term memory can use InMemoryStore for in-process storage or PostgreSQL with langgraph-checkpoint-postgres for persistent database storage. The PostgreSQL backend requires installation of psycopg (Psycopg 3) with optional binary extras.
PostgreSQL package installation
To use PostgreSQL for long-term memory in Python, install langgraph-checkpoint-postgres and psycopg[binary] using pip or uv. The psycopg[binary] extra is recommended for most users.
PostgreSQL package installation for JavaScript
To use PostgreSQL for long-term memory in JavaScript, install @langchain/langgraph-checkpoint-postgres using npm.
Three levels of prompt caching in LangChain
Prompt caching can be engaged at three levels: (1) Implicit provider caching where providers automatically pass on cost savings with no configuration required (examples: OpenAI, Gemini); (2) Provider-level explicit controls where you manually indicate cache points for greater control (examples: ChatOpenAI via prompt_cache_key, Anthropic content-block cache_control, Gemini, AWS Bedrock cachePoint blocks); (3) LangChain middleware where middleware optimizes caching of stable system prompt and tool content (examples: AnthropicPromptCachingMiddleware, BedrockPromptCachingMiddleware).