CoderAI
Local and cloud GPU inference orchestrator exposing an OpenAI-compatible API for multimodal models.
Tested on Linux
Use it
Linux x86_64
Clone source via HTTPS mirror
git clone --depth 1 https://github.com/nextime/coderai.git coderai-srcInstall torch-free core deps
pip install --quiet requests fastapi "uvicorn[standard]" pydantic python-multipart psutil setproctitle "huggingface-hub>=0.19" numpy Pillow pyyaml jinja2 zeroconf litellm 2>&1 python3 -c "import fastapi, uvicorn, pydantic, requests; print('core ok')"
Where the README and reality differ
- README: git clone git@git.nexlab.net:nexlab/coderai.git
What worked: cloned https://github.com/nextime/coderai.git with --depth 1; SSH URL needs a key/account
Check that it works
cd coderai-src && python3 coderai --helpThe output should contain “--download-model”.
What it does
- Serve models through an OpenAI-compatible APINot tested
- Generate text, images, video and audioNot tested
- Route requests across runtimes and machinesNot tested
- Send extra work to rented cloud GPUsNot tested
How far pickfit got
Linux x86_64
- Installs
- Starts
- 0 of 4 features work
- Works offline
Good for
- generate images from text prompts
- generate video from text prompts
- synthesize speech from text
- train LoRA adapters on own GPU
- run OCR on images or PDFs
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
Installs and starts
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
Differences the test machine hit
- README: curl run_oci.sh && ./run_oci.sh --nvidia -d
In the test: script only wraps docker/podman, both absent (which docker/podman not found), so candidate cannot run here - README: ./build.sh all plus full requirements.txt
In the test: python3 coderai --help works after only requests/fastapi/uvicorn/pydantic/python-multipart/psutil/setproctitle/huggingface-hub/numpy/Pillow/pyyaml/jinja2/zeroconf/litellm; full requirements.txt plus torch/llama builds not needed for --help - README: llama-cpp-python with Vulkan/CUDA for GGUF models
In the test: no prebuilt llama-cpp-python wheel and no gcc/cmake, so GGUF engine cannot be built here