Skip to content

Latest commit

 

History

15 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Prompter

Turn rough prompt material into a usable prompt from the terminal. From this checkout, start with the offline image assembler: it prints an assembled prompt to standard output and needs no provider credentials.

If you want to… Start with… What happens
Assemble an image prompt without a network call image <subject> Prints an assembled prompt built from local components.
Improve, critique, rewrite, or apply a prompt refine, critique, rewrite, or apply Sends the prepared input to the configured LLM provider.
Find or maintain local prompt files browse or prompts status|upgrade Searches the prompt vault or reports/stages starter-prompt changes.
Inspect or change resolved settings configure Opens the terminal wizard, or prints non-secret settings when output is redirected.
Refresh model choices models refresh Fetches catalog data and updates the local model-choice cache.

First use: assemble an offline image prompt

Prerequisite: this module declares Go 1.26.3. The command below is safe to run from a checkout; GOWORK=off makes it use this module rather than a parent workspace.

GOWORK=off go run . image "desert observatory" --profile minimal

It prints an image-generation prompt to standard output. In the checked runtime, the result was:

desert observatory, clean composition, concept art

The command selects a local assembly profile and combines the supplied subject with embedded components. It builds prompt text only; it does not generate an image.

Source-checked variation (not executed for this README): request structured output with --json.

GOWORK=off go run . image "desert observatory" --json

Configuration and local state

Prompter resolves settings in this order:

CLI flags > environment variables > ~/.config/prompter/config.json > defaults

prompter configure opens a configuration form on an interactive terminal. When standard output is redirected, it prints resolved non-secret settings instead. Configuration includes the active provider, model, provider endpoint, prompt locations, image component file, timeout, output-token budget, retry count, and buffered-result clipboard preference.

The prompt vault is configured through prompts_dir and prompts_dirs. browse searches local Markdown prompt files; an empty primary vault can be seeded with starter prompts. models refresh stores its catalog cache at ~/.config/prompter/models-dev.json.

Non-goals and operating limits

  • image assembles an image-generation prompt; it does not create an image.
  • models refresh is a networked catalog refresh, unlike the offline image path above.
  • A streamed provider response can write partial text before the provider reports a failed or incomplete terminal state. Treat streamed output from a nonzero exit as unusable.
  • Prompt-output validation requires buffered output, so validated apply prompts reject --stream.

Capability reference

LLM prompt operations

refine, critique, rewrite, and apply accept input from arguments, --file, or standard input. apply selects a catalog prompt by exact name or alias; its prompt body becomes the system prompt and its frontmatter is not sent to the provider.

Use command help to inspect supported flags before making a remote request:

prompter refine --help

The LLM commands support provider/model selection, file input, output-file writing, buffered clipboard copying, dry-run inspection, streaming, endpoint overrides, and verbose timing. --output writes the buffered result both to the named file and standard output; it cannot be combined with --stream.

Local prompt vault

prompter prompts status
prompter prompts upgrade --dry-run

status classifies starter prompts. upgrade --dry-run previews missing-prompt installation and versioned replacement candidates without writing them. A non-dry-run upgrade installs missing prompts and stages replacements rather than overwriting existing files.

Configuration and model catalog

prompter configure
prompter models refresh

Run configure on an interactive terminal for the form. Run models refresh only when a networked catalog update is intended; it refreshes choices from Models.dev, OpenRouter, and a local OMLX endpoint when available.

Verification and contribution

From the repository root, run:

GOWORK=off go test -count=1 ./...
GOWORK=off go build -o prompter .
GOWORK=off go vet ./...
gofmt -l .

The first command runs the repository test packages. The build produces a local prompter binary. A clean gofmt -l . produces no path output.

For detailed command flags, setup notes, prompt-file format, providers, and troubleshooting, see:

Limits and non-goals

Remote prompt operations need a configured provider and its applicable authentication. The offline image command is the credential-free path.

About

Zero-dependency, high-performance CLI for prompt enhancement, markdown restructuring, offline image prompt assembly, and prompt vault search.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages