Providers

Tokori is provider-agnostic. Pick whatever LLM backend you have keys for (or run one locally) — the chat UI, flashcards, and reader all work the same regardless.

Built-in providers

Provider	Local?	Streaming	Notes
Ollama	✓	✓	Easiest local setup. Pulls models from https://ollama.com.
OpenAI	–	✓	GPT-4o, GPT-4 mini, etc.
Anthropic	–	✓	Claude Sonnet 4.6, Opus 4.7.
Gemini	–	✓	Gemini 1.5 Pro / Flash.
OpenRouter	–	✓	Aggregates 200+ models behind one key.
MiniMax	–	✓	Cheap, fast Chinese-tuned.

Switching the active provider

Settings → Providers → click Use on a configured row. The active provider applies immediately to chat, the reader's "Explain this" button, and the flashcards AI helper.

Ollama

1. Install Ollama from https://ollama.com/download.
2. ollama pull llama3.1 (or any model).
3. In Tokori: Settings → Providers → Add → Ollama.
4. Host: http://localhost:11434 (default — change if you're reaching a remote daemon).
5. Pick the model from the dropdown (Tokori queries /api/tags to list pulled models).

Cold-start latency

Ollama lazy-loads weights on the first real request. Tokori warms the model the moment you click Use on an Ollama provider — the first chat after a session start should be hot. Subsequent restarts re-pay the cold load on the first message.

OpenAI

1. Get a key from https://platform.openai.com/api-keys.
2. Settings → Providers → Add → OpenAI.
3. Paste the key. Pick a model.
4. Optional: set a custom base_url to point at any OpenAI-compatible endpoint (DeepSeek, Together, vLLM, …).

Anthropic

1. Get a key from https://console.anthropic.com/.
2. Add as Anthropic with model id like claude-sonnet-4-6.
3. The streaming uses the Messages API + Anthropic-version header under the hood.

Gemini

1. Get a key from https://aistudio.google.com/apikey.
2. Add as Gemini with model gemini-1.5-pro or -flash.

OpenRouter

1. Get a key from https://openrouter.ai/keys.
2. Add as OpenAI (it's OpenAI-compatible) with base_url: https://openrouter.ai/api/v1.
3. Model id can be any of OpenRouter's catalog (e.g. anthropic/claude-sonnet-4-6).

MiniMax

1. Get a key from https://www.minimaxi.com.
2. Add as MiniMax with model MiniMax-M2-7B-Instruct or whichever you've selected.

Adding a new provider

The provider abstraction lives in src-tauri/src/providers.rs — one ProviderConfig enum variant + a streaming function per kind. New providers are usually a 100-line PR. See Contributing for the full process.