AI Providers

The bot supports multiple AI providers through a hybrid 3-slot routing system. You can mix providers — for example use Gemini for main responses and Groq for lightweight intent analysis.

The 3 slots

• Main — The primary provider used for user-facing AI responses.
• Lite — Used for small tasks like intent detection, topic extraction, translation. If empty, Main is used.
• Fallback — Engaged only when Main fails (5xx, timeout, network error). Safety net.

Supported providers

Cloud providers

| Provider | Free tier? | Default model | Notes |

|-------------|------------|-----------------------------|--------------------------------------------|

| Gemini | Yes (generous) | gemini-2.5-flash | Best for starters. Image support included. |

| Anthropic | No | claude-sonnet-4-5 | Highest quality. Image support. |

| OpenAI | No | gpt-4o | Popular. Image + vision support. |

| DeepSeek | Yes | deepseek-chat | Very cheap. Quality comparable to GPT-4o. |

| Groq | Yes | llama-3.3-70b-versatile | Ultra-fast inference. Text only. |

Self-hosted

| Provider | Default model | Notes |

|------------|---------------|------------------------------------------------|

| Ollama | llama3.1 | Run any model locally. Base URL + model required. |

| Custom | (user) | Any OpenAI-compatible endpoint. |

Configuration

Go to Dashboard → AI → AI Yapılandırması. You'll see 3 cards — Main, Lite, Fallback.

For each card:

1. Select a provider from the dropdown
2. Enter the API key (required for cloud providers)
3. For Ollama / Custom, enter the base URL (e.g. http://localhost:11434/v1) and model name (e.g. gemma:2b)
4. Model field is optional for cloud providers — the default is used if empty
5. Click Bu Slot İçin Test Et to verify the connection

Example setups

Recommended (balanced)

• Main: Gemini (quality + free tier)
• Lite: Groq (fast + free, for intent)
• Fallback: (empty — uses env Groq)

Fully free

• Main: Groq
• Lite: (empty)
• Fallback: Gemini

Self-hosted with cloud backup

• Main: Ollama (your server, llama3.1)
• Lite: Groq (fast intent, your Ollama is busy generating answers)
• Fallback: Gemini (if your Ollama server goes down)

Enterprise

• Main: Anthropic (claude-sonnet-4-5)
• Lite: OpenAI (gpt-4o-mini)
• Fallback: DeepSeek (cheap backup)

API key safety

• Keys are stored per-server in guilds/<id>/config.json
• The bot never shares its own keys across servers
• Keys are write-only in the UI (you can update but not read back)
• Use the 🗑️ button next to a stored key to clear it