Repository navigation
dlgo v0.1.0
Fast LLM inference in pure Go. These are portable, CPU-build binaries (no cgo, no Vulkan SDK required). Vulkan GPU acceleration currently requires building from source with -tags vulkan on a machine with the Vulkan SDK.
Included commands
| Binary | Purpose |
|---|---|
dlgo |
Chat CLI (dlgo run model.gguf), model manager (dlgo models / pull / list / rm), dlgo info |
dlgo-server |
HTTP server with web UI, OpenAI-compatible chat API, and System One typed decisions (/v1/systemone) |
dlgo-vision |
Multimodal demo: mmproj projector + text model image description |
dlgo-image |
Text-to-image diffusion pipeline (Z-Image/Lumina2-style DiT + VAE) |
mmproj-encode |
Encode/inspect vision projector (mmproj) output |
Downloads
| Platform | Files |
|---|---|
| Windows (amd64) | dlgo-windows-amd64.exe, dlgo-server-windows-amd64.exe, dlgo-vision-windows-amd64.exe, dlgo-image-windows-amd64.exe, mmproj-encode-windows-amd64.exe |
| Linux (amd64) | dlgo-linux-amd64, dlgo-server-linux-amd64, dlgo-vision-linux-amd64, dlgo-image-linux-amd64, mmproj-encode-linux-amd64 |
| macOS (Apple Silicon) | dlgo-darwin-arm64, dlgo-server-darwin-arm64, dlgo-vision-darwin-arm64, dlgo-image-darwin-arm64, mmproj-encode-darwin-arm64 |
Quick start
# Download a supported model (Unsloth GGUF from Hugging Face)
./dlgo models
./dlgo pull qwen3.5:0.8b-q4_k_m
# Chat
./dlgo run ~/.cache/dlgo/models/Qwen3.5-0.8B-Q4_K_M.gguf
# Start the web UI + API
./dlgo-server --model ~/.cache/dlgo/models/Qwen3.5-0.8B-Q4_K_M.gguf --port 8080On macOS first run you may need to allow the unsigned binary: xattr -d com.apple.quarantine ./dlgo (adjust per binary). See the project README for full usage, model management, System One, vision, and diffusion documentation.