diff --git a/src/config.js b/src/config.js index 110b583..75e9360 100644 --- a/src/config.js +++ b/src/config.js @@ -77,6 +77,16 @@ export const VLLM_ENDPOINTS = [ 'hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M': 32768, } }, + { + id: "gpt-oss.ai.unturf.com", + url: "https://gpt-oss.ai.unturf.com/v1", + // Ollama endpoints don't return max_model_len, so we specify it here + // Note: Model supports 262144 but limited by GPU VRAM (24GB) + modelContextWindows: { + 'gpt-oss:latest': 64000, + 'deepseek-r1:14b-qwen-distill-q8_0': 64000, + } + }, ]; // Function to get API configuration (custom or default)