Skip to content

Cloud‐LLM‐API‐Setup

Jump to bottom
ic005k edited this page Jul 29, 2026 · 1 revision

Cloud LLM API Configuration Guide

Overview

Knot supports third-party online LLM services compatible with the OpenAI v1 API standard (SiliconFlow as demonstration example). Cloud AI functions are fully separated from local embedding models; they can be enabled or disabled independently.

UI Field Reference (Match AI Preference Screenshot)

Master Toggle: Enable AI

The checkbox labeled "开启AI / Enable AI" is the global switch for cloud API features:

  • Unchecked: All cloud LLM functions are disabled. Endpoint, API Key and model ID will not take effect.
  • Checked: Activates cloud API configuration, all input fields below become functional.

Cloud API Config Area

Setting Description
Endpoint URL Root API address of your service provider, example: https://api.siliconflow.cn/v1
API Key Private authentication key obtained from your service dashboard
Model ID Target model identifier provided by the vendor, example: THUDM/GLM-4-9B-0414

Operation Buttons

  1. Select: Quick pick preset service endpoints and model IDs;
  2. Test: Verify network connection and key validity after filling all parameters.

Local Model Zone Note

This guide only covers cloud API setup. The "Local Model List" section below is an independent offline embedding module, which works without relying on the "Enable AI" toggle.

Step-by-Step Setup

  1. Open Knot Settings and switch to the AI tab;
  2. Tick the "Enable AI" checkbox to unlock cloud API input fields;
  3. Fill in 3 parameters provided by your LLM service: Endpoint URL, API Key, Model ID;
  4. Click the "Test" button to validate connection; the configuration is ready if the test passes, and all current settings will be saved automatically;
  5. To disable cloud AI at any time, uncheck the "Enable AI" toggle, all online requests will be terminated immediately.

Additional Notes

  1. Isolation Logic: When "Enable AI" is unchecked, Knot will send zero network requests to cloud APIs. Local embedding models still work normally on their own.
  2. Compatibility: Only services following the OpenAI v1 REST API specification are supported.
  3. Troubleshooting for failed tests: Double-check your API Key, remove extra spaces in the endpoint URL, and confirm the Model ID is available on your vendor’s platform.

Clone this wiki locally