Skip to content

INIT.md

Latest commit

 

History

History
210 lines (133 loc) · 3.54 KB

INIT.md

File metadata and controls

210 lines (133 loc) · 3.54 KB

Initialization (init)

The init command bootstraps ExplainThisRepo by creating a local, persistent configuration file. Runnig It configures:

  • your selected LLM provider
  • optional GitHub access (for private repos and higher rate limits)

What init does

  • Prompts you to select an LLM provider:

    • Gemini
    • OpenAI
    • Ollama
    • Anthropic
    • Groq
    • OpenRouter

  • Overwrites any existing configuration file

  • Collects only the configuration required for that provider

  • Optionally prompts for a GitHub token

  • Writes a minimal config.toml file to the OS-appropriate config directory

  • Writes the selected provider as the active provider

  • Exits immediately

GitHub access

During setup, you can configure a GitHub token.

This enables:

  • Access to private repositories
  • Higher API rate limits for public repositories

If skipped:

  • public repositories still work
  • rate limits are lower
  • private repositories will fail

If not set in config, ExplainThisRepo will also check:

  • GITHUB_TOKEN
  • GH_TOKEN

environment variables.

If both config and environment variables are set, the config value takes precedence.

Per provider input

Depending on the provider selected:

Gemini

  • Prompts for api_key

OpenAI

  • Prompts for api_key

Ollama

  • Prompts for model (e.g. llama3, gemma3:4b, glm-5:cloud)
  • Prompts for host (default: http://localhost:11434)

Anthropic

  • Prompts for api_key

Groq

  • Prompts for api_key
  • Prompts for model selection

OpenRouter

  • Prompts for api_key
  • Prompts for model selection or manual input

Only the selected provider configuration and optional GitHub configuration are written. Empty input for optional fields skips configuration safely.

What init does NOT do

  • No repository analysis is performed during initialization.
  • No model execution
  • No API key validation
  • No dependency installation
  • No environment variable modification

The configuration is written locally only.

Input handling

  • API keys and tokens are read using hidden terminal input
  • Characters are not echoed
  • Paste works normally
  • Ctrl+C exits cleanly without writing partial state

Config location

A single authoritative config path is used per OS.

Windows

%APPDATA%\ExplainThisRepo\config.toml

macOS / Linux

$XDG_CONFIG_HOME/explainthisrepo/config.toml

Fallback: ~/.config/explainthisrepo/config.toml

Example resulting config

GitHub token

[github]
token = "ghp_xxx"

Gemini

[llm]
provider = "gemini"

[providers.gemini]
api_key = "..."

OpenAI

[llm]
provider = "openai"

[providers.openai]
api_key = "..."

Ollama

[llm]
provider = "ollama"

[providers.ollama]
model = "<user-selected>"
host = "http://localhost:11434"

Anthropic

[llm]
provider = "anthropic"

[providers.anthropic]
api_key = "..."

Groq

[llm]
provider = "groq"

[providers.groq]
api_key = "..."
model = "<user-selected>"

OpenRouter

[llm]
provider = "openrouter"

[providers.openrouter]
api_key = "..."
model = "<user-selected>"

Design intent

init exists to separate configuration from execution.

After initialization:

  • All analysis commands run without repeated prompts

  • GitHub authentication is resolved automatically from config or environment

  • Provider selection can be overridden via --llm without re-running init.

  • Only one provider is active at a time

This establishes a stable foundation for:

  • single-provider execution with optional runtime override

  • authenticated GitHub access