$ ask anything
tprompter is a developer-friendly, CLI-first tool that streamlines interactions with AI language models.
It comes with built-in agents and templates to help you solve your tasks using large language models (LLMs).
The tool automatically generates prompts, embeds local file content (for code reviews, test generation, etc.), and archives each request for future reference. Its “Actions” feature allows you to copy or print results, open them in your text editor, or feed them directly into ChatGPT—all while providing easy configuration (e.g., setting your OpenAI API key) via a simple CLI interface.
Work in Progress
git diff | tprompter generate commit_message --after=chatgpt)tprompter ask --model o3-mini "What's on your mind?"
tprompter ask --after=print_raw "I need plain output"
The model is aware of the OS and shell version
tprompter includes built-in agents to assist you with various tasks. For example, the code_review agent generates a code review prompt.
tprompter agent code_review
Shell pipes are supported, allowing you to pass the output of any command directly to an agent.
tprompter supports chaining agents to refine LLM output. For example:
$ cat /path/to/file.ts | tprompter agent unit_test | tprompter agent code_review
$ tprompter archive 1 --after=print_raw # to see the unit_tests' original answer
File embedding is supported: simply drag and drop a file from your IDE, Finder, or Explorer into the terminal. Folder embedding is also supported.
See the File Embedding section for more details.
If you need to quickly generate a prompt based on a template, use the generate command. You can then copy the prompt to your clipboard and paste it into the ChatGPT page (or any other LLM interface).
You can directly pass the prompt to the ChatGPT web-page by using the --after=chatgpt option. See the How "Open in ChatGPT" works? section for more details.
This feature is particularly useful when working with large codebases on high-cost models such as o1. It saves time and resources while providing a full ChatGPT experience.
For example, to generate a starter prompt:
tprompter generate starter --after=chatgpt
Then, use ChatGPT to assist with code reviews, unit tests, or other tasks.
There is no specific configuration file; you can set the configuration via the CLI:
tprompter config openAIApiKey <your_api_key>
The tprompter will suggest you set the openAIApiKey if you try to use the ask or agent without proper configuration.
Configuration CLI suggestions work, so you can set the configuration on the fly.
npm install -g tprompter
git clone https://github.com/pinkiesky/tprompter.git
cd tprompter
npm install
npm run install:local
| Name | Expected Input | Purpose |
|---|---|---|
| write_unit_tests | Code followed by unit test examples | Generates unit tests for the provided code. Including examples improves the LLM’s output quality. |
| code_review | Code | Generates a code review for the provided code. |
| starter | Code or any text of your choice | Generates a starter prompt (e.g., "I will write you questions or ask for favors based on the code provided"). |
| commit_message | Code (recommended: output from git diff --staged) | Generates a commit message based on your code changes. |
| changelog | Code (recommended: output from git log --oneline --since="2 weeks ago" or git diff v1..main) | Generates a changelog summarizing the changes. |
| summary | Any text | Generates a brief, structured summary of the provided text. |
Display the help information to see all available commands and options:
$ tprompter help
tprompter <command> [options]
Commands:
tprompter list List available templates
tprompter generate <templateNameOrFile> Generate a prompt
tprompter template <subcommand> Manage templates
tprompter archive <index> Archive of previously generated prompt
tprompter version Show version
tprompter assets <subcommand> List and get assets
tprompter agent <template> Ask a question
tprompter ask <question..> Ask a question
tprompter config <configKey> [value] Show or set current configuration
tprompter completion generate completion script (bash, zsh, fish)
Options:
--help Show help [boolean]
-v, --verbose Increase verbosity [boolean]
-q, --quiet Do not output anything [boolean]
tprompter list
# or
tprompter template list
tprompter generate <template> --after=<action>
# tprompter generate code_review --after=chatgpt
tprompter generate <file_path> --after=<action>
# tprompter generate /path/to/file.md --after=print
🫤 archive feature is quite messy, I'm working on it.
tprompter archive 0 --after=<action>
# tprompter archive 0 --after=copy
If you prefer to use your own templates instead of the built-in ones, you can install a custom template. The template format is a simple Markdown file with placeholders.
You can override a built-in template by installing a custom one with the same name.
Example template:
You are an expert code converter. Please convert the following code to TypeScript.
Do not output any explanation; you should just return a code!
Code:
'''
${readStdin('Enter the code')}
To install the custom template, use one of the following methods:
tprompter prompt install <name> <filepath>
# tprompter prompt install code_review /path/to/file.md
# or
cat /path/to/file.md | tprompter prompt install <name>
curl <url> | tprompter prompt install <name>
# curl -s https://raw.githubusercontent.com/pinkiesky/tprompter/main/prompts/code_review.md | tprompter prompt install code_review
tprompter prompt uninstall <name>
# tprompter prompt uninstall code_review
To open the templates folder in your default file manager, run:
tprompter prompt open_folder
tprompter enhances your prompts by embedding the content of files. When you provide a file path, it reads the file and includes its content in the prompt. If the path points to a directory, tprompter recursively reads all files within that directory and embeds their content.
$ tprompter generate starter --after=copy -v
❯ Input required (Press CTRL+D to finish): Enter the source code files
'/tprompter/src' '/logs/tprompter.log'
class Info {
# You can mix plain code/data with file paths
# The tprompter will try to read the file and embed its content into the prompt
# Even if you mixed everything
}
^D
debug: Reading directory: /tprompter/src {"name":"TextDataEnricher"}
debug: Found 14 items in directory: /tprompter/src {"name":"TextDataEnricher"}
debug: Reading directory: /tprompter/src/actions {"name":"TextDataEnricher"}
debug: Found 1 items in directory: /tprompter/src/actions {"name":"TextDataEnricher"}
debug: Reading file: /tprompter/src/actions/Actions.ts {"name":"TextDataEnricher"}
debug: Reading directory: /tprompter/src/archive {"name":"TextDataEnricher"}
...
info: Tokens count: 18158 {"name":"MainController"}
debug: Executing action: copy {"name":"Actions"}
$
To configure how tprompter reads directories, you can include a .promptconfig.yaml file in the desired directory. This file allows you to specify which files or directories to exclude or include during the embedding process.
Example .promptconfig.yaml:
exclude:
- node_modules
- .git
- dist
- package-lock.json
include:
- *.ts
This configuration instructs tprompter to exclude node_modules, .git, dist, and package-lock.json from the embedding process. The logic is similar to that of a .gitignore file.
Note that node_modules, .git, and package-lock.json are excluded by default.
For the --after option in the generate, archive and agent commands, you can use the following actions:
print - Display the prompt on the terminal.print_raw - Display the raw markdown prompt on the terminal.chatgpt - Open the ChatGPT page with the prompt.copy - Copy the prompt to the clipboard.edit - Open the prompt in your default editor.The ChatGPT web app does not provide an option to load a prompt directly on it. To overcome this limitation, tprompter opens the ChatGPT page and then injects the prompt into it using JavaScript. This JavaScript code should be installed via a browser extension such as Tampermonkey.
The script is designed to be simple and secure. It injects the prompt into the page.
But how does the script know what to inject? tprompter generates the prompt and starts a "disposable HTTP server" that serves the prompt content on any request. The script retrieves the prompt from this server and injects it into the ChatGPT page. The server stops automatically after the prompt is injected, and it uses a randomly assigned port that changes every time.
tprompter assets get chatgptloader
The script is straightforward, so you can easily review it before installing.
To enable shell completion for bash or zsh, run:
tprompter completion >> ~/.bashrc
After installation, restart your terminal or run source ~/.bashrc to apply the changes.
tprompter assets get fish > ~/.config/fish/completions/tprompter.fish
After installation, restart your terminal to apply the changes.
https://docs.github.com/en/get-started/exploring-projects-on-github/contributing-to-a-project is a good starting point.
Any contribution is welcome and makes me happy.
I'm open to any suggestions and help and would encourage you to provide me with a critical review.
FAQs
```bash $ ask anything ```
We found that tprompter demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 0 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Multiple high-impact npm maintainers confirm they have been targeted in the same social engineering campaign that compromised Axios.
Security News
Axios compromise traced to social engineering, showing how attacks on maintainers can bypass controls and expose the broader software supply chain.
Security News
Node.js has paused its bug bounty program after funding ended, removing payouts for vulnerability reports but keeping its security process unchanged.