Add a Janitor AI API configuration with proxy URL https://api.deepseek.com/chat/completions and model name deepseek-flash, then save and refresh.
Older guides use retired model IDs such as deepseek-chat or a :free OpenRouter model; OpenRouter lists no free DeepSeek model as of October 7, 2026.
On an assumed 8,000-token prompt, deepseek-flash costs $0.32 to $2.76 per 1,000 messages, depending on caching and peak hours.
Most errors come from a wrong URL, a retired model name, an empty DeepSeek balance, or not refreshing the page after saving.
Avoid third-party free proxy sites; they can see your prompts and key. Use DeepSeek's own endpoint or a trusted router.