A web-based bash script formatter and linter with AI-powered autofix.

• Format: Format bash scripts using shfmt
• Lint: Check scripts with shellcheck
• Autofix: Apply shellcheck's automatic fixes
• Autofix (AI): Fix shellcheck issues using an AI provider (Groq, DeepSeek, or any OpenAI-compatible endpoint)

• Go 1.21+
• shfmt - github.com/mvdan/sh
• shellcheck - github.com/koalaman/shellcheck

./bash-script-tools

Open http://localhost:8085 in your browser. Formatting and linting work with zero configuration; AI autofix needs a config file (below).

All configuration lives in $XDG_CONFIG_HOME/bash-script-tools/config.toml (~/.config/bash-script-tools/config.toml when XDG_CONFIG_HOME is unset). Generate a starter config with:

./bash-script-tools init

Every key is optional and falls back to its default. ${VAR} in any string value is replaced with that environment variable (empty if unset) - use it to keep API keys out of the file:

# port = 8085
# shfmt_path = "shfmt"
# shellcheck_path = "shellcheck"

# Which provider handles AI autofix. Omit to disable AI autofix.
provider = "deepseek"

[providers.groq]
api_key = "${GROQ_API_KEY}"
# api_url = "https://api.groq.com/openai/v1/chat/completions"
# model = "openai/gpt-oss-120b"

[providers.deepseek]
api_key = "${DEEPSEEK_API_KEY}"
# api_url = "https://api.deepseek.com/chat/completions"
# model = "deepseek-v4-flash"

A provider table can also set extra_body — fields merged into the chat-completions request body, for provider-specific parameters. The built-in deepseek defaults use it to disable thinking mode (which would make autofix take minutes); re-enable it with:

[providers.deepseek.extra_body]
thinking = { type = "enabled" }

groq and deepseek have built-in defaults for api_url and model, so their tables only need an api_key. Any other OpenAI-compatible endpoint works too - add a table with all three keys:

provider = "ollama"

[providers.ollama]
api_url = "http://localhost:11434/v1/chat/completions"
api_key = "unused"
model = "qwen2.5-coder"

If the selected provider's key resolves empty, the server still starts - AI autofix is just disabled. A malformed config file or a provider naming something unknown is a startup error.

MIT