Installation Guide

Connecting your application to the Datafuse Active Integration network requires installing our lightweight client SDK. Datafuse provides first-class support for both Node.js/TypeScript and Python environments.

Prerequisites

Before installing the SDK, ensure you have:

• A Datafuse developer account.
• An active API Key (e.g. df_live_... or df_test_...), which you can fetch from your developer dashboard.
• Node.js (version 18 or higher) or Python (version 3.8 or higher) installed in your workspace.

1. Python SDK Installation

For Python-based AI agent setups (supporting LangChain, LlamaIndex, CrewAI, AutoGen, and custom LLM loops), install our unified SDK via pip, poetry, or uv.

Core Package

pip install datafuse-sdk

From Local Source (For Developers)

If you are developing locally inside the Datafuse monorepo and want to link the development SDK directly:

pip install -e sdks/python

Quick Verification

Verify that the Python client can be imported without errors:

python -c "import datafuse; print('Python SDK successfully installed!')"

2. TypeScript SDK Installation

!IMPORTANT@datafuse/sdk is not yet published to npm. Until it is, install it directly from the repository source shown below. If you only need Node/serverless calls today, the TypeScript client guide shows the equivalent fetch-based calls with zero dependencies.

From Source (Current Method)

Clone the monorepo (or copy the sdks/typescript directory), build it, and install it as a local file dependency:

git clone https://github.com/ApertaCodex/datafuse.git
cd datafuse/sdks/typescript
npm install
npm run build

Then reference it from your project with a local file: dependency:

npm install /absolute/path/to/datafuse.xyz/sdks/typescript

Quick Verification

Check that the built package resolves and exports the client correctly:

node -e "const { Datafuse } = require('@datafuse/sdk'); console.log('TypeScript SDK successfully installed!');"

Environment Configuration

Datafuse clients automatically look for standard environment variables to initialize credentials, keeping your secrets out of source control.

Configure the following variables in your local .env file:

# API Key from your Datafuse developer console (df_live_... or df_test_...).
# Used ONLY for the /api/v1/connected_accounts routes. NOT accepted by the
# integration resolve-tools/invoke routes or most other authenticated routes
# (those require DF_TOKEN, the JWT bearer).
DATAFUSE_API_KEY=df_live_your_secret_key_here

# User access token (JWT) from the app's authentication flow.
# Required for integration, resolve-tools, invoke, and most authenticated routes.
DF_TOKEN=your_user_access_token

# Base API Server Endpoint (defaults to https://api.datafuse.xyz)
DATAFUSE_API_URL=https://api.datafuse.xyz

!IMPORTANT The integration routes (POST /api/v1/integrations, …/resolve-tools, …/invoke) currently accept a JWT bearer token (DF_TOKEN) only — they do not accept DATAFUSE_API_KEY. Use DATAFUSE_API_KEY only for the /api/v1/connected_accounts routes; every other authenticated route (including /api/v1/api-keys and /api/v1/audit/logs) requires a JWT bearer token. The curl and client guides below use DF_TOKEN for the integration flow.

!TIP If you are running Datafuse inside a private self-hosted VPC, change DATAFUSE_API_URL to point to your secure internal load balancer (e.g. http://datafuse-gateway.internal).

Next Steps

Now that the SDK is successfully installed:

• Follow the 5-Minute Quickstart to list integrations, resolve tools, and invoke a tool.
• Prefer the terminal? Use the 5-Minute Quickstart with curl.
• Explore our TypeScript client guide or Python client guide.