LangChain.js Integration

Persistent semantic memory and server-side vector search for LangChain.js. Full TypeScript types — no local embedding model required.

Quick Start

docker run -d \
  --name dakera \
  -p 3300:3300 \
  -e DAKERA_ROOT_API_KEY=dk-mykey \
  ghcr.io/dakera-ai/dakera:latest

curl http://localhost:3300/health  # → {"status":"ok"}

npm install @dakera-ai/langchain @dakera-ai/dakera @langchain/core

import { DakeraMemory } from "@dakera-ai/langchain";
import { ConversationChain } from "langchain/chains";
import { ChatOpenAI } from "@langchain/openai";

const memory = new DakeraMemory({
  apiUrl: "http://localhost:3300",
  apiKey: "dk-mykey",
  agentId: "my-agent",
});

const chain = new ConversationChain({
  llm: new ChatOpenAI({ model: "gpt-4o" }),
  memory,
});

// Memory persists across sessions and restarts
const response = await chain.call({ input: "My project is called NeuralBridge." });
console.log(response.response);

DakeraMemory

Persistent conversation memory for LangChain.js chains. Stores and recalls conversation history using Dakera's hybrid search (BM25 + vector).

import { DakeraMemory } from "@dakera-ai/langchain";
import { ConversationChain } from "langchain/chains";
import { ChatOpenAI } from "@langchain/openai";

const memory = new DakeraMemory({
  apiUrl: "http://localhost:3300",
  apiKey: process.env.DAKERA_API_KEY!,
  agentId: "my-agent",
  recallK: 5,      // how many past memories to surface per turn
  importance: 0.7, // importance score for stored memories
});

const chain = new ConversationChain({
  llm: new ChatOpenAI({ model: "gpt-4o" }),
  memory,
});

// First session
await chain.call({ input: "My name is Alice and I'm building a chatbot." });

// Later session — memory persists across restarts
const { response } = await chain.call({ input: "What was I building?" });
console.log(response); // "You mentioned you were building a chatbot."

DakeraMemory options

DakeraVectorStore

Server-side embedded vector store for RAG. Compatible with VectorStore from @langchain/core. Dakera handles all embeddings — no OpenAI embeddings API needed.

import { DakeraVectorStore } from "@dakera-ai/langchain";

const vectorStore = new DakeraVectorStore({
  apiUrl: "http://localhost:3300",
  apiKey: process.env.DAKERA_API_KEY!,
  namespace: "my-docs",
});

// Index documents (server handles embedding)
await vectorStore.addDocuments([
  { pageContent: "Dakera is a self-hosted memory server.", metadata: {} },
  { pageContent: "It scores 87.6% on the LoCoMo benchmark.", metadata: {} },
]);

// Similarity search
const results = await vectorStore.similaritySearch("benchmark score", 3);
console.log(results);

RAG chain with retrieval

import { DakeraVectorStore } from "@dakera-ai/langchain";
import { RetrievalQAChain } from "langchain/chains";
import { ChatOpenAI } from "@langchain/openai";

const vectorStore = new DakeraVectorStore({
  apiUrl: "http://localhost:3300",
  apiKey: process.env.DAKERA_API_KEY!,
  namespace: "product-docs",
});

const chain = RetrievalQAChain.fromLLM(
  new ChatOpenAI({ model: "gpt-4o" }),
  vectorStore.asRetriever({ k: 4 }),
);

const { text } = await chain.call({ query: "How does memory decay work?" });
console.log(text);

DakeraVectorStore options

Related integrations

Links

• GitHub — dakera-langchain-js
• Dakera deploy — Docker Compose setup
• Dakera full documentation
• All integrations