Login

Integration

Long-term memory for LangChain

LangChain's `ConversationBufferMemory` is in-process: your worker restarts and the memory's gone. `VectorStoreRetrieverMemory` requires a vector DB you have to host. `EngramChatMessageHistory` is a direct replacement that gives you BM25 plus vector plus knowledge-graph retrieval against a hosted backend, with explanation traces.

Get an API key (free) All integrations →

Install

Three steps: sign up for an Engram API key, paste a BYOK LLM-provider key on /models, then drop the snippet below into LangChain.

Three steps to memory in your agent

  1. Sign up. Free, no card. You'll land on a Getting Started page that walks the next two steps.
  2. Add your LLM key. Engram is BYOK. Paste an OpenAI / Anthropic / Groq / Together / Fireworks key and we'll route every extraction and query call through your provider. You pay your provider directly. We never see your inference.
  3. Paste the snippet below into your agent and restart it. Use Authorization: Bearer <api-key>with the API key from your portal.

engram-langchain: durable BaseChatMessageHistory

Drop-in replacement for ConversationBufferMemory / VectorStoreRetrieverMemory. One hosted backend with hybrid retrieval (BM25 + vector + KG) and an explanation trace on every recall. Source: github.com/lumetra-io/engram-langchain.

  1. Install the Python client + LangChain:
    2. Terminal
    pip install lumetra-engram langchain langchain-core
  2. Vendor engram_langchain.py from the repo (~30 LOC; PyPI release coming) and export your API key:
    3. Terminal
    curl -fsSL https://raw.githubusercontent.com/lumetra-io/engram-langchain/main/engram_langchain.py -o engram_langchain.py
export ENGRAM_API_KEY="<api-key>"
  3. Use it as a BaseChatMessageHistory on any runnable:
    4. Python
    from langchain_core.runnables.history import RunnableWithMessageHistory
from engram_langchain import EngramChatMessageHistory

chain_with_memory = RunnableWithMessageHistory(
    chain,
    lambda session_id: EngramChatMessageHistory(bucket=f"user-{session_id}"),
    input_messages_key="input",
    history_messages_key="history",
)

What you can do once memory's wired in

  • Replace `ConversationBufferMemory` in an existing `RunnableWithMessageHistory` with a one-line swap
  • Use per-session buckets with `lambda session_id: EngramChatMessageHistory(bucket=f'user-{session_id}')`
  • Call `.query()` for semantic recall across the full history when the recent-window view isn't enough
  • Stack with other LangChain memory classes if you want both summary memory and full retrieval

FAQ

Does this work with both `langchain` and `langchain-core`?

Yes. `EngramChatMessageHistory` extends `BaseChatMessageHistory` from `langchain-core`, so anything that consumes that interface works.

How do I get semantic recall instead of recent-window?

Call `.query('some question')` directly on an `EngramChatMessageHistory` instance. `.messages` returns the recent window (good for prompt-stuffing), while `.query()` runs hybrid retrieval against the entire bucket.

Is the file PyPI-installable yet?

For now you vendor `engram_langchain.py` from the repo (~30 LOC). PyPI release is coming, and the file is small enough to copy in the meantime.

Related integrations

LangGraph Durable BaseStore replacing InMemoryStore. Tuple namespaces map to Engram buckets. Read more → LlamaIndex BaseMemory implementation that replaces LlamaIndex's chat-history buffer with hybrid retrieval. Read more → Haystack Haystack 2.x components: EngramWriter and EngramRetriever for any Pipeline. Read more →

Ship durable memory in LangChain today

Free tier: 10K memories and 50K retrievals per month. No credit card. Same Engram backend powers all 41 integrations, so memories you write from one client are immediately queryable from the rest.

Start free See pricing