AI Profiles
AI profiles are named presets for BYOK translations. Each profile stores an engine (Groq or OpenAI BYOK), optional model, prompt template, sampling fields, and an optional per-profile API key. Assign profiles to specific features instead of maintaining one global model/prompt for everything.
AI profiles ship on the beta dashboard first. Production may lag behind beta — check your dashboard for an AI Profiles section under User Settings or Server Configuration.
What a profile contains
| Field | Required | Notes |
|---|---|---|
| Name | Yes | Display label (e.g. "Formal announcements", "Casual chat") |
| Engine | Yes | Groq BYOK or OpenAI BYOK |
| Model | No | Falls back to your account-level BYOK model when empty |
| Prompt template | No | Supports {source_language}, {target_language}, {text} |
| Temperature / max tokens | No | OpenAI profiles; Groq uses account defaults when omitted |
| API key | No | Optional override; otherwise uses your saved user or guild BYOK key |
| Default flag | No | One default profile per user or guild |
Not stored in profiles: protection delimiters and full sampling parameter sets still come from your main BYOK settings. Profiles override model, prompt, temperature (OpenAI), max tokens (OpenAI), and optionally the API key for the translation context they serve.
Creating and managing profiles
Personal profiles
- Open Dashboard → User Settings
- Scroll to AI Profiles (requires User BYOK premium)
- Click New profile, fill in fields, and save
- Mark one profile as default if you want a fallback when no context-specific profile is assigned
Guild profiles
- Open Dashboard → Server → Server Configuration
- Find AI Profiles at the bottom of the configuration section (requires Server BYOK)
- Create profiles the same way as personal profiles
- Set a guild default for server-wide BYOK translations that do not have a more specific assignment
Assigning profiles
Resolution order for a translation:
- Context assignment (mirror, default-language channel, flag reaction, or context menu)
- Default profile for that user or guild
- Base BYOK settings (engine dropdown + model/prompt fields on User Settings or Server Configuration)
| Context | Where to assign |
|---|---|
| Channel mirroring | Edit a mirror group → AI profile for this mirror group |
| Default language | Default Language channel rule → profile selector per channel |
| Flag reaction translations | Server → Customization → AI profile for flag translations |
| Translate context menu | Server → Customization → AI profile for context menu translations |
If the assigned profile is deleted, translations fall back to the default profile, then to base BYOK settings.
Per-profile API keys
Profiles can store their own encrypted API key. When set, that key is used only for translations resolved through that profile. Leave the key empty to reuse the main BYOK key from User Settings or Server Configuration.
Use cases:
- Separate billing or rate limits per mirror group
- Different OpenAI projects for announcements vs. casual channels
- Testing a new model/key without changing your primary BYOK configuration
Integration with BYOK sampling
When a profile is active for a translation:
- Model, prompt, temperature, max tokens, and optional API key come from the profile (with fallbacks described above)
- Top P, frequency/presence penalties, reasoning effort, and delimiters still come from your account or guild BYOK fields
See BYOK overview and BYOK dashboard settings for delimiter and sampling details.
Tips
- Create a default profile before assigning context-specific ones — it catches any unassigned flows
- Use short, purpose-driven names so staff know which profile applies where
- Keep prompts focused; mirror groups often need different tone than default-language announcement channels
- After changing assignments, test with a message in each context — engine changes apply on the next translation
Related
- BYOK overview
- BYOK dashboard settings
- Customization — flag and context-menu profile assignments
- Feature terminology
- Premium tiers