Chatting Over API

Introduction

TaskLift provides access to chats and assistants over API for users with Pro and higher plans.

This is a very powerful feature because our API is OpenAI-compatible which means that TaskLift connects & works out-of-the-box with a whole ecosystem of AI applications, agents, frameworks and libraries. That makes the whole UI/app side of TaskLift optional, giving you a choice to use any app that you prefer while still taking advantage of TaskLift's wide selection of models and tools, as well as access to all configured assistants.

Chat API is also useful when you want to create a custom embed of TaskLift assistant on your website like described on the  Embedding Assistants page.

For a direct access to TaskLift tools that allows to plug them directly into external AI models, check out the MCP-compatible Tool API too - learn more on the  Calling Tools Over API page.

Usage With Applications

In order to use TaskLift chats in arbitrary app you'll need to configure it with following options:

• for API/LLM provider/adapter, select Generic OpenAI or similar (wording may vary across apps)
• for API base URL, fill https://tasklift.ai/api
• for API key, fill the key from the API Keys section

Usage With Code And Frameworks

Here's how you can make an example chat call:

Request & Response Details

Our Chat API supports following subset of OpenAI API standard:

• Input message - by setting the "messages": [{"role": "user", "content": "message"}] parameter
• Model selection - by setting the "model": "<model-name>" parameter
• Streaming of the response - by setting the "stream": true parameter
• Returning of token counts - in the "usage" key in response (or last event when streaming)

In addition we introduce following TaskLift-specific features:

• Tools selection - by setting the "tools": ["<tool-name>", ...] parameter
• Assistant selection - by setting the "assistant_id": "<id>" parameter (it will use assistant's model and tools ignoring ones sent with the request)
• Chat continuation - by setting the "chat_id": "<id>" parameter (it will append each message to the existing chat without a need to re-send the whole conversation and de-duplicate multiple chats that would be created for each new message within conversation)

Model Discovery

Some AI apps can also discover the list of models supported by given API and for that sake TaskLift provides the OpenAI-compatible Model API that accompanies the Chat API.

Compatible AI apps or frameworks should just call this API when needed and you can call it on your own in following way:

All models accessible by your account will be returned.