Skip to content

setup.md

Latest commit

 

History

History
92 lines (63 loc) · 2.51 KB

setup.md

File metadata and controls

92 lines (63 loc) · 2.51 KB
Raw

Setup

Purpose: install prompter, configure one provider, and verify a working run.

Prerequisites

  • Go installed and $(go env GOPATH)/bin in your PATH
  • API key for one remote provider, or a running local OMLX server

Main workflow

  1. Install:

Using Homebrew (macOS & Linux):

brew install dotcommander/tap/prompter

Or using the Go toolchain:

go install github.com/dotcommander/prompter@latest

Update to future releases with the same installation method, for example:

go install github.com/dotcommander/prompter@latest
  1. Run immediately after authenticating Google ADC and selecting your project:
export GOOGLE_CLOUD_PROJECT="your-project-id"
prompter refine "explain this code"
  1. Or use any remote provider using standard environment variables:
export OPENAI_API_KEY="sk-..."
prompter refine -p openai "explain this code"
  1. Interactive configuration wizard (optional):
prompter configure

prompter configure launches an interactive setup form in your terminal to configure:

  • Default LLM provider (Gemini, OpenAI, Groq, Cerebras, DeepSeek, OpenRouter, Zai, OMLX)
  • Default model identifier
  • API key environment variable constant names (e.g. $OPENAI_API_KEY)
  • Reasoning effort level (low, medium, high)
  • Default automatic clipboard copying

Settings are saved to ~/.config/prompter/config.json using portable machine paths. If the primary prompt vault is empty, configure also auto-seeds the starter prompts.

Re-running prompter configure refreshes the configuration file. It persists the active provider, key-variable name, model, base URL, effort, and clipboard preference. Inline api_key values are never written to disk — keys stay in environment variables or the configured key_env name.

  1. Browse your prompt vault (optional):
prompter browse

The browser searches configured prompt directories recursively. If the primary vault is empty, it seeds the curated starter prompts before opening.

  1. Refresh the model catalog (optional):
prompter models refresh

This fetches Models.dev and OpenRouter metadata, caches up to five affordable choices per provider, and prints them. configure uses the same cache to populate model choices, falling back to built-in choices when the fetch fails.

Verification checklist

  • Command prints enhanced text
  • prompter refine -v "test" includes timing output

Related docs

  • common-tasks.md
  • flags.md
  • troubleshooting.md
  • use-json-output.md