LLM providers

eunha supports three providers. All descriptions use the same prompt contract regardless of provider.

OpenAI

• Recommended model: gpt-4o-mini (fast, cheap, reliable JSON mode)
• Requires OPENAI_API_KEY or a token in config
• Uses the Chat Completions API with response_format: { type: "json_object" }

[llm]
provider = "openai"
api_key = "sk-..."
model = "gpt-4o-mini"

Anthropic

• Recommended model: claude-haiku-4-5-20251001 (fast + affordable)
• Uses the Messages API
• JSON output is enforced via the prompt — Claude follows the format reliably

[llm]
provider = "anthropic"
api_key = "sk-ant-..."
model = "claude-haiku-4-5-20251001"

Ollama (local / offline)

Ollama lets you run open-source models locally. Privacy-first, no API costs, works offline.

Requirements:

• Ollama installed and running
• A model with JSON mode support (Ollama --format json)

Tested models: llama3.2, mistral, qwen2.5

[llm]
provider = "ollama"
model = "llama3.2"
base_url = "http://localhost:11434"

Note

eunha uses Ollama’s /api/generate endpoint with format: "json". If the model produces invalid JSON, eunha fails with a clear error — there is no silent fallback.

Switching providers

Edit ~/.eunha/config.toml and change provider. Existing descriptions are not affected. Re-describe with shift-D (single) or shift-A (all stale) after switching.