CLI Reference
Usage
vibex <file.vibe> [flags]
Flags
| Flag | Default | Description |
|---|---|---|
-o <output> | a.out | Output binary name |
--arch <arch> | arm64 | Target architecture: arm64 or x86_64 |
--model <model> | gpt-4o | Override the LLM model |
--work | off | Keep temp directory and print its path |
Environment Variables
Configure via .env in your project directory, ~/.vibex.env globally, or as shell env vars.
| Variable | Default | Description |
|---|---|---|
VIBEX_BASE_URL | https://api.openai.com/v1 | OpenAI-compatible API endpoint |
VIBEX_API_KEY | — | API key (OPENAI_API_KEY also works) |
VIBEX_MODEL | gpt-4o | Model to use as compiler |
VIBEX_ARCH | arm64 | Default target architecture |
Examples
# Basic usage
vibex hello.vibe -o hello
# Use a specific model
vibex hello.vibe -o hello --model gpt-4o
# Compile for x86_64 (requires Rosetta on Apple Silicon)
vibex hello.vibe -o hello --arch x86_64
# Keep temp files for debugging
vibex hello.vibe -o hello --work
# Use a local model
VIBEX_BASE_URL=http://localhost:11434/v1 VIBEX_MODEL=llama3.2 VIBEX_API_KEY=ollama \
vibex hello.vibe -o hello
Inspecting Generated Assembly
Use --work to keep the temp directory:
vibex hello.vibe -o hello --work
# [vibex] WORK=/var/folders/.../vibex_work_xxx/
# assembly is at /var/folders/.../vibex_work_xxx/program.s