itch.io

ChatGBT made me a local proxy that converts the response format into a compatible one with deepseek, so I can actually use DeepSeek without needing to get it through OpenRoutner. So no worries.

2026-09-23 16:44:53.437 main     LOG   [llm] → Custom endpoint models: https://api.deepseek.com/v1/models

2026-09-23 16:44:58.757 main     LOG   [llm] → Custom endpoint deepseek-flash (thinking=medium, tier=standard, maxTokens=65536, schema=connection_test)

2026-09-23 16:44:58.757 main     LOG   [llm] → system: (truncated, 20 chars)

2026-09-23 16:44:58.757 main     LOG   [llm] → user: Ping.

2026-09-23 16:44:59.089 main     LOG   [llm] ✕ Custom endpoint deepseek-flash HTTP 400 after 332ms

2026-09-23 16:44:59.089 main     ERROR [ipc] llm:testWriter failed: {

    code: 'LLM_REQUEST_REJECTED',

    message: 'Custom endpoint rejected the request (HTTP 400).',

    detail: 'invalid_request_error This response_format type is unavailable now'

  }

DeepSeek API gives HTTP 400 with OpenAI-compatible endpoint

Hi! I really like the game so far. I tried using my paid DeepSeek API key instead of Gemini.

Settings I used:

• Endpoint: https://api.deepseek.com/v1/chat/completions
• Model: deepseek-flash

When I click Test connection, I get: “Custom endpoint rejected the request (HTTP 400).”

DeepSeek's API is OpenAI-compatible, so I was wondering whether there is something in the request that isn't compatible with DeepSeek yet.

It would be great to have DeepSeek working as a custom provider, since I already have a paid DeepSeek API key and really like the game.

Thanks!