ChatGPT JavaScript Toolkit is a single JavaScript file to wrap the OpenAI's ChatGPT access. The main purpose is to enable ChatGPT access from the browser's client side.
The usage is extremely simple: Just download the file chatgpt.js or chatgpt.min.js. No npm or similar tools are required.
- Chrome Browser Extension
- Visual Studio Code Extension
- Other use cases are welcome!
Storing the OpenAI API key in the client side is supposed to be insecure. However, we give this option to the user: the privacy and the cost.
For most ChatGPT-enhanced applications, the service providers deploy their own service endpoints as untrusted proxies and have the capability to take advantage of the user's private data.
On the other hand, we believe that the decentralized applications have the bright future. The application developers build the prompts and handlers while all requests are sent directly to the OpenAI endpoints. In this way, our API key and private data are not leaky to the application providers.
The following table summarizes the pros and cons:
| Centralized App | Decentralized App | |
|---|---|---|
| Privacy | Untrusted App Providers. | Trusted Extension Code (Open-Source). |
| Cost | ChatGPT usage + application profits. | ChatGPT actual usage. |
| Motivation of Development | Benefits from extra charges and privacy data (?) | Love (?) |
Store credentials.
let account = new OAccount();
await account.load(); // Load account from storage;
account.api_key = "new key";
await account.save(); // Save account to storage.
const headers = account.headers(); // Generate the header to send queries.Get the information of models.
const models = account.models();
const [model_list, errorcode, error_message] = models.list();
const [model, errorcode, error_message] = models.retrieve("gpt-3.5-turbo");Store history, token usages, and finish reasons (Internal use).
Create and manage a chat.
const chat = new OChat(account);
const [message, errorcode, error_message] = await chat.create("ChatGPT is "); // Send a query. There are more options in the source code.
const history = chat.history; // Get the OHistory object.We provide a unit-test extension in /tests. We use Qunit.js as our testing framework for its simplicity.
cp credentials-template.js credentials.js
And customize the credentials.js. A valid OpenAI account is required for testing.
Load the root directory of this project to the Chrome extensions.
The pop-up of this testing extension will directly run the tests and show all results.
- Support a minimal set of functionalities to create a Chat.
- Error handling and robustness.
- Extend and support more functionalities.
Any Issue or Pull Request is welcome. You can also contact me via Homepage.