No description
- TypeScript 97.7%
- JavaScript 2.3%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
- Add containsImageInput() to detect image DataParts before the request is sent, so text-only models fail fast with a clear message instead of an opaque API error at runtime - The error lists the vision-capable models available in the picker (e.g. GLM-5.3 Flash) so the user knows where to switch - Verified with a standalone test harness (vscode stub): 5/5 cases — text-only, image in current/earlier message, non-image data, tool result |
||
| images | ||
| src | ||
| .editorconfig | ||
| .gitignore | ||
| .prettierrc.cjs | ||
| .vscodeignore | ||
| bundle.mjs | ||
| eslint.config.cjs | ||
| eslint.ignores.cjs | ||
| LICENSE | ||
| package-lock.json | ||
| package.json | ||
| README.md | ||
| tsconfig.json | ||
| tsconfig.tsbuildinfo | ||
GLM Chat Provider
Z.AI GLM models as a VS Code Language Model Chat Provider for the Coding Plan.
Text Models
| Model | Context | Output | Tool Calling |
|---|---|---|---|
| GLM-5.3 | 1M | 128K | Yes |
| GLM-5.2 | 1M | 131K | Yes |
| GLM-5.1 | 200K | 131K | Yes |
| GLM-5 | 205K | 131K | Yes |
| GLM-5-Turbo | 200K | 131K | Yes |
| GLM-4.7 | 205K | 131K | Yes |
| GLM-4.7-Flash | 200K | 131K | Yes |
| GLM-4.7-FlashX | 200K | 131K | Yes |
| GLM-4.6 | 205K | 131K | Yes |
| GLM-4.5 | 131K | 98K | Yes |
| GLM-4.5-Flash | 131K | 98K | Yes |
| GLM-4.5-Air | 131K | 98K | Yes |
Vision Models
| Model | Context | Output | Image Input | Tool Calling |
|---|---|---|---|---|
| GLM-5.3-Flash | 1M | 128K | Yes (image/video) | Yes |
| GLM-5V-Turbo | 200K | 131K | Yes | Yes |
| GLM-4.6V | 128K | 33K | Yes | Yes |
| GLM-4.5V | 64K | 16K | Yes | Yes |
Commands
GLM: Set API Key-- Store your Z.AI API key in VS Code secretsGLM: Clear API Key-- Remove the stored API keyGLM: Manage Provider-- Open provider management optionsGLM: Set Thinking Effort-- Toggle thinking mode (Auto, Enabled, Low, High, Max, Disabled)
Thinking Mode
For models that support it (GLM-4.5 and above), you can control whether the model uses its reasoning/thinking capability.
Run GLM: Set Thinking Effort from the Command Palette to choose between:
- Auto -- Let the model decide when to think (default)
- Enabled -- Always use thinking mode
- Disabled -- Never use thinking mode
The selected value is persisted in your VS Code settings under glm-chat-provider.defaultThinkingMode.
How to Use
- Open the Command Palette and run
GLM: Set API Keyto configure your API credentials - Use the provider from VS Code's Language Model Chat UI and select Z.AI GLM
License
MIT (c) Denizhan Dakilir