Deveraux-Parker

Qwen3-Coder-30B-A3B-Monkey-Wrenches. Efforts toward giving Qwen 3 Coder 30B A3B proper agentic tool calling capabilities at or near 100% reliability.

Parakeet_Multitalk. A realtime speech to text diarization system to gather and interleave speech from multiple speaker audio.

Nvidia_parakeet-tdt-0.6b-v2-FAST-BATCHING-API-1200x-RTFx. Python

nanoGPT_1GPU_SPEEDRUN. Would you like to recreate GPT-2 124m in a cave with a box of scraps and a 4090 in less than two hours? LETS SPEEDRUN!

GPT-OSS-MONKEY-WRENCHES. HTML

VibeVoice-Low-Vram. Getting VibeVoice 7b working with 10 gb of vram.

supertonic-2-afterburner-fastAPI. A batching fastAPI for a custom safetensors Supertonic 2 audio model.

FlappyAI. Testing various AI's ability to recreate flappy bird.

GPT-OSS-20b-TAKE-THE-WHEEL. Experiments with giving GPT-OSS-20b situational control and access to an old fashioned Asteriods spaceship.

Adaptive_Sparse_Attention_MoE. Python

glm-code-api. A simple GLM code plan api for agentic coding.

browser_parakeet_onnx_realtime. A simplistic demo of a browser-based parakeet speech to text transcription service that lives in the browser and runs locally on the user's CPU. It will automatically download and prepare the model for use. Click the button to speak, speak, click the button to stop speaking and send the audio. Feel free to implement more complex VAD if you like!