@dakera-ai/langchain

LangChain.js integration for Dakera — persistent semantic memory and server-side vector search, no local embedding model required.

| Class | Description | |-------|-------------| | DakeraMemory | Drop-in BaseMemory for LangChain.js conversation chains | | DakeraVectorStore | VectorStore backed by Dakera's server-side embedding engine |

Quick Start

Step 1 — Run Dakera

Dakera is a self-hosted memory server. Spin it up with Docker:

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

For a production setup with persistent storage, use Docker Compose:

curl -sSfL https://raw.githubusercontent.com/Dakera-AI/dakera-deploy/main/docker-compose.yml \
  -o docker-compose.yml
DAKERA_API_KEY=dk-mykey docker compose up -d

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

Full deployment guide: github.com/Dakera-AI/dakera-deploy

Step 2 — Install the integration

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

Step 3 — Use it

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);

Installation

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

Requirements: Node.js ≥ 20, a running Dakera server (see Step 1 above)

DakeraMemory

Persistent conversation memory for LangChain.js chains. Stores and recalls conversation history using Dakera's hybrid search.

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."

Options

| Option | Type | Default | Description | |--------|------|---------|-------------| | apiUrl | string | — | Dakera server URL (e.g. http://localhost:3300) | | apiKey | string | "" | Dakera API key | | agentId | string | — | Agent identifier for memory namespacing | | recallK | number | 5 | Number of memories to recall per turn | | minImportance | number | 0 | Minimum importance threshold for recall | | memoryKey | string | "history" | Key injected into the chain prompt | | inputKey | string | first key | Input key used as recall query | | importance | number | 0.7 | Importance score assigned to stored memories |

DakeraVectorStore

Server-side embedded vector store for RAG. Dakera handles embeddings — no local model, 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 persistent memory platform for AI agents.", metadata: { source: "docs" } },
  { pageContent: "It supports Python, JavaScript, Rust, and Go SDKs.", metadata: { source: "docs" } },
]);

// Semantic search
const results = await vectorstore.similaritySearch("What languages are supported?", 4);
console.log(results[0].pageContent);

RAG chain

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

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

From texts (LangChain convention)

const store = await DakeraVectorStore.fromTexts(
  ["Document one content", "Document two content"],
  [{ source: "a" }, { source: "b" }],
  null, // embeddings param — unused, Dakera handles server-side embedding
  { apiUrl: "http://localhost:3300", apiKey: "dk-mykey", namespace: "docs" },
);

Options

| Option | Type | Default | Description | |--------|------|---------|-------------| | apiUrl | string | — | Dakera server URL | | apiKey | string | "" | Dakera API key | | namespace | string | — | Vector namespace to read/write | | embeddingModel | string | namespace default | Server-side embedding model override |

Related packages

| Package | Framework | Language | |---------|-----------|----------| | langchain-dakera | LangChain | Python | | crewai-dakera | CrewAI | Python | | llamaindex-dakera | LlamaIndex | Python | | autogen-dakera | AutoGen | Python |

Links

• Dakera Server — self-hosted memory server
• Dakera JS SDK — low-level API client
• Integration guide — full setup walkthrough
• All integrations

License

MIT © Dakera AI

dakera.ai · Documentation · Request Early Access