Copilot/Settings
Copilot
Copilot Settings
Configure runtime options, vendor profiles, and prompt shortcuts for ApiHug Copilot.
ApiHug Copilot uses Langchain4j as its model integration layer. That means the settings model is based on vendor profiles, runtime defaults, and reusable prompts.
Toolbar > ApiHug > Settings > AI
What It Is
The settings page has three responsibilities:
- define runtime defaults shared by chat sessions
- store one or more vendor profiles
- manage reusable prompt shortcuts
Why It Matters
Clean configuration prevents two common problems:
- low-signal answers caused by unstable sampling or missing prompts
- environment confusion when teams switch between cloud, internal, and local models
Runtime Settings
Toolbar > ApiHug > Settings > AI > Settings
These values apply to normal chat execution unless a session overrides them.
| Name | What it controls |
|---|---|
| Chat Memory Size | how many recent messages stay in working memory |
| Top-P | probability sampling window |
| Max Output Tokens | upper bound for response length |
| Timeout(s) | request timeout |
| Max Attempts | retry count |
| Temperature | response variability |
| Stream | streamed output |
| System Prompt | default system instruction |
| Show Detail | timing and token detail in the UI |
Vendor Profiles
Toolbar > ApiHug > Settings > AI > Vendor
Copilot supports multiple profiles per provider, which is useful for development, staging, production, or model comparisons.
Use straightforward names such as OpenAI-dev, Azure-prod, or Local-Ollama.
Common provider families include:
- OpenAI-style APIs
- Azure OpenAI
- Gemini
- Anthropic
- DeepSeek
- Qianfan
- Zhipu
- Ollama
- LM Studio
- Jan
- OpenRouter
- Tongyi and Hunyuan
Profile fields
| Name | Purpose |
|---|---|
| Name | the profile label shown in the UI |
| Description | optional explanation for the profile |
Typical connection fields
| Provider type | Common fields |
|---|---|
| OpenAI-style | URL, API Key, optional model |
| Azure AI | endpoint, API Key, deployment |
| Local runtime | local URL and optional model |
| Qianfan | API Key, Secret Key, model |
| Zhipu | API Key, model |
Any provider that exposes a compatible OpenAI-style API can usually be modeled as a custom vendor profile.
Prompt Library
Toolbar > ApiHug > Settings > AI > Prompt
Prompt shortcuts let teams standardize repeated asks such as /explain, /review, or /api.
| Name | Purpose |
|---|---|
| name | the slash command, such as /explain |
| prompt | the full instruction body behind that command |
Keep prompt names short, unique, and easy to recognize.
Prompt guidance
- write prompts for one clear job
- include project-specific constraints when needed
- keep reusable prompts stable and versionable
- prefer prompt libraries over repeating long instructions by hand
Next Step
- Continue to Copilot Features
- Review Skills & Rules if you want framework-aware prompt workflows