CyborgDB - LangChain Integration

CyborgDB provides a seamless integration with LangChain, enabling developers to leverage encrypted vector search for large language models (LLMs) and RAG pipelines. This integration is implemented via the CyborgVectorStore class, which conforms to the VectorStore interface from langchain-core.

Installation

Embedded
Python SDK
JS/TS

Install cyborgdb-core with the langchain extra (Python 3.10 or newer; the python3 that ships with macOS is 3.9):

pip install "cyborgdb-core[langchain]"

Install the cyborgdb package with the langchain extra (Python 3.10 or newer):

pip install "cyborgdb[langchain]"

Install the cyborgdb npm package and @langchain/core (an optional peer of cyborgdb, only required for this integration):

npm install cyborgdb @langchain/core

To learn more about the differences between cyborgdb and cyborgdb-core, refer to the Deployment Models Guide.

Import

Embedded
Python SDK
JS/TS

from cyborgdb_core.integrations.langchain import CyborgVectorStore

from cyborgdb.integrations.langchain import CyborgVectorStore

import { CyborgVectorStore } from 'cyborgdb/integrations/langchain';

Key Features

Full VectorStore API compatibility (add_texts, similarity_search, etc.)
Async operations support
Configurable distance metrics (cosine, euclidean, squared_euclidean)
Document ID tracking and management
Max marginal relevance search
Factory methods (from_texts, from_documents) for quick setup

Key Classes

`CyborgVectorStore`

Implements the LangChain-compatible vector store interface. Supports:

.add_texts() / .add_documents()
.similarity_search() / .similarity_search_with_score()
.max_marginal_relevance_search()
.get() / .list_ids() / .delete()
.as_retriever() for use in LangChain chains
Async variants (.aadd_texts(), .asimilarity_search(), etc.)
Reversible document ID tracking
Configurable distance metric (cosine, euclidean, squared_euclidean)

`StorageConfig` (Embedded only)

Configuration object for defining where index data is stored, built with StorageConfig.disk(path) or StorageConfig.s3(bucket, ...). Used only with the Embedded library (cyborgdb-core).

Quick Start

Embedded
Python SDK
JS/TS

import secrets
from cyborgdb_core import StorageConfig
from cyborgdb_core.integrations.langchain import CyborgVectorStore

# Generate an encryption key for the index.
# WARNING: store this key securely — data encrypted with a lost key is unrecoverable.
index_key = secrets.token_bytes(32)

store = CyborgVectorStore.from_texts(
    texts=["hello world", "goodbye world"],
    embedding="all-MiniLM-L6-v2",   # built-in model: no extra install, no paid key
    index_key=index_key,
    api_key="",                     # free tier; pass your key to lift the 1M-item cap
    storage_config=StorageConfig.disk("./cyborgdb-langchain"),
    metric="cosine",
)

for doc in store.similarity_search("hello"):
    print(doc.page_content)

# Delete the index so the example can be run again (the key above is not saved).
store.delete(delete_index=True)

To keep the index between runs, store index_key and construct

CyborgVectorStore(index_name="langchain_index", index_key=index_key, api_key="", embedding="all-MiniLM-L6-v2", storage_config=...)

directly: it reopens an existing index instead of creating one.To embed with OpenAI instead of the built-in model, install langchain-openai (it is not part of the langchain extra), set OPENAI_API_KEY, and pass embedding=OpenAIEmbeddings():

pip install langchain-openai

from cyborgdb.integrations.langchain import CyborgVectorStore

store = CyborgVectorStore.from_texts(
    texts=["hello world", "goodbye world"],
    embedding="all-MiniLM-L6-v2",
    index_key=CyborgVectorStore.generate_key(save=True),
    api_key="your-api-key",
    base_url="http://localhost:8000",
    metric="cosine"
)

docs = store.similarity_search("hello")

import { CyborgVectorStore } from 'cyborgdb/integrations/langchain';
import { OpenAIEmbeddings } from '@langchain/openai';

const store = await CyborgVectorStore.fromTexts(
    ["hello world", "goodbye world"],
    [{ source: "greeting" }, { source: "farewell" }],
    new OpenAIEmbeddings(),
    {
        baseUrl: "http://localhost:8000",
        apiKey: "your-api-key",
        indexName: "my_documents",
        indexKey: CyborgVectorStore.generateKey(),
        metric: "cosine"
    }
);

const docs = await store.similaritySearch("hello");

fromExistingIndex() is available as a JS/TS-only convenience factory method for connecting to an already-created index without re-adding documents.

Store index_key before you upsert any data. CyborgDB has no key-recovery path — losing the key means losing the data. For anything past evaluation, keep the key in AWS Secrets Manager, Vault, or your KMS. See Managing Keys.

generate_key(save=True) exists in the Python SDK only. The embedded library’s generate_key() returns the bytes and nothing else, so persist the key yourself as shown above.

Introduction

CyborgVectorStore

Types

CyborgDB - LangChain Integration

Installation

Import

Key Features

Key Classes

`CyborgVectorStore`

`StorageConfig` (Embedded only)

Quick Start

​Installation

​Import

​Key Features

​Key Classes

​CyborgVectorStore

​StorageConfig (Embedded only)

​Quick Start

Installation

Import

Key Features

Key Classes

`CyborgVectorStore`

`StorageConfig` (Embedded only)

Quick Start