Create a chatbot using Google Drive. Make it smart and store data by connecting it to you or your visitors' google drive account.
Select your AI Model and optionally connect you and/or your users' google drive. Node developers can npm install langDrive to access our DriveUtils and DriveChatbot classes. Individuals may use a free deployed instance of this service by vising addy-ai.com. ChatGPT, HuggingFace, or more LLM Google OAuth2 keys are required to run your own instance (create credentials for a web app oAuth). Read our tutorial on OAuth2 on our blog.
Get a chatbot up and running NOW!
-
Click here to
-
Set Heroku Secret Variables to gain access to their Service
GOOGLE_WEB_CLIENT_IDandGOOGLE_WEB_CLIENT_SECRETwith Google OAuth2 Keys instructions are needed for user login and to connect Google Drive to their chatbot.OPENAI_API_KEYfor ChatGPT4.HUGGINGFACE_API_KEYto use a HuggingFace LLM.- ``
LangDrive was built for Node.js.
npm install langdriveexposesDriveChatbotand it's underlying utility classDriveUtils.DriveChatbot- Uses an agent conditionally integrated with drive_utils.DriveUtils- Syncs data to a users google drive folder if the drive_chatbot is initialized with a OAuth2 access token. The extent of the data sync is conditioned on Auth scope.
The Chatbot returns Async Promises.
Chatbot's minimal initalization is like so:
chatbot = new langdrive.DriveChatbot({model_config:{HuggingFaceAPIKey:<KEY>}})
or like so:
chatbot = new langdrive.Chatbot({model_config:{openAIApiKey:<KEY>}})
Get started with a sample script by created the following files:
npm install langdrive dotenv
node test.js
.env File:
OPENAI_API_KEY=<YOUR_KEY_HERE>
GOOGLE_DESKTOP_CLIENT_KEYFILE_PATH=<YOUR_KEY_HERE>
test.js File:
require("dotenv").config();
const langdrive = require("langdrive");
// LangDrive returns promises
(async()=>{
// To initialize Langdrive, give it a model to use and any associated config information.
// Here we select openAi and pass it an API key (hidden behind .env)
let chatbot = await new langdrive.DriveChatbot({
verbose: true,
drive: {
verbose: false,
...(!GOOGLE_DESKTOP_KEYFILE_PATH
? {}
: {
server: {
embed_from_folder: "chatbot",
embed_to_folder: "chatbot/embeddings",
scopes: ["https://www.googleapis.com/auth/drive"],
// serviceKeyFile: __dirname + "/../" + GOOGLE_SERVICE_KEYFILE_PATH
// OR
desktopKeyFile: __dirname + GOOGLE_DESKTOP_KEYFILE_PATH
// ( Alternately:) desktopKeyFileContents: GOOGLE_DESKTOP_CLIENT_KEYFILE_CONTENTS
// OR
// desktopTokenFile: GOOGLE_DESKTOP_CLIENT_TOKEN_PATH:
// ( Alternately:) desktopTokenFileContents: GOOGLE_DESKTOP_CLIENT_TOKEN_CONTENTS
// OR
//client_id: GOOGLE_DESKTOP_CLIENT_ID, // and
//client_secret: GOOGLE_SERVICE_CLIENT_SECRET //and
//client_redirect_uri: xyz
}
})
},
model: {
service: !!HUGGINGFACE_API_KEY ? "huggingFace" : "chatOpenAi",
model_config: !!HUGGINGFACE_API_KEY
? {
model_id: "meta-llama/Llama-2-30b",
huggingFaceApiKey: HUGGINGFACE_API_KEY
}
: {
modelName: "gpt-3.5-turbo", // default = "text-davinci-003"
// maxTokens: 256, // default = 256
openAIApiKey: OPENAI_API_KEY,
temperature: 0.9
}
},
agent: {
type: "chat-conversational-react-description",
memory_length: 2,
vector_length: 2,
verbose: false,
tools: [],
agent_config: {}
// prefix
// suffix
}
});
// LangDrive returns a promise, so let's await those.
let prompt = "My name is Michael, What can you do for me.";
console.log("> " , await chatbot.sendMessage(prompt));
prompt = "What can you do for me in google drive?";
console.log("> " , await chatbot.sendMessage(prompt));
prompt = "What is my name?";
console.log("> " , await chatbot.sendMessage(prompt));
})()
You can also clone the repo and get started with our demo chatbot
- Download Repo
-
npm install
- Create Google OAuth2 Keys
- .env.examples -> .env + Keys
- npm run start
The props used in DriveChatbot(props) configure your chatbot. Available settings and their default values are shown below.
DriveUtils =
~~~~~~~~~~ class methods
getDrive()
listFiles(mimeType)
getFileInfo(filename, mimeType)
getFileById(fileId)
getFileByName(filename, mimeType = false)
createFile(filename, mimeType, message)
createAndOrGetFile(filename, mimeType, message)
updateFile(fileId, mimeType, message)
~~~~~~~~~~ static methods
getAuthUrl({redirect_uri, client_id, scope})
handleAuthCallback({client_secret, access_token, code, redirect_uri})
refreshToken({client_secret, access_token, refresh_token, client_id})
checkAndRefresh ({access_token, timestamp, refresh_token, client_id, client_secret})
Create an instance of the DriveUtils class like so:
const myDriveTool = new DriveUtils(GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET, { ACCESS_TOKEN, verbose })
The DriveUtils class has some helping static functions to help you work with google OAuth2 access tokens.
Provides a URL for a user-consent page.
let authUrl = DriveUtils.getAuthUrl({
redirect_uri: "https://localhost:3000/auth/google/callback",
scope: "https://www.googleapis.com/auth/drive",
client_id: GOOGLE_CLIENT_ID
});
res.redirect(authUrl);
Success will redirect you to the redirect_uri you with a code included in as a query parameter.
This `redirect_uri`` must be approved in the project's google cloud console oAuth2 admin panel.
We take the resulting req.query.code and issue anothe request to google to retrieve our ACCESS_TOKEN
let await DriveUtils.handleAuthCallback({
code,
client_id: GOOGLE_CLIENT_ID,
client_secret: GOOGLE_CLIENT_SECRET
});
The resulting data will return a refresh_token. If you save that, when the access_token expires, you can use checkAndRefresh to get a new one.
await DriveUtils.checkAndRefresh(
access_token,
req.session.timestamp,
req.session.refresh_token,
GOOGLE_CLIENT_ID,
GOOGLE_CLIENT_SECRET
);
With that, you can create a new DriveUtils which provides Drive access.
myDriveTool.listFiles(mimeType)