arhupy 2.4.0

A prompt engineering toolkit for LLMs

arhupy

A lightweight prompt engineering toolkit for LLMs like Claude, GPT, and Gemini.

arhupy gives you a practical set of tools for writing, scoring, comparing, saving, exporting, improving, and reusing prompts. It works from both Python and the command line, uses only the Python standard library, and stores local data as simple JSON files in your current working directory.

Write prompt -> Score it -> Compare versions -> Save or export -> Reuse later
                         -> Improve with Claude
                         -> Serve through CLI, web, or local API

Why arhupy?

Prompt engineering can get messy fast. arhupy keeps the useful parts small and organized:

Need	arhupy gives you
Fill reusable prompts	Prompt("You are a {role}")
Chain prompt steps	PromptChain and arhupy chain
Score prompt quality	score_prompt() and arhupy score
Compare prompt versions	compare_prompts() and arhupy diff
Save prompt libraries	Local arhupy_library.json storage
Share prompts	JSON export and import helpers
Track sessions	Local prompt history with reuse
Improve with AI	Claude-powered prompt improvement
Generate prompts	Turn short ideas into structured prompts
Run locally	CLI, web dashboard, and HTTP API
Share prompts	Local share links with scored prompt pages
Extend behavior	Simple plugin system

Installation

Install from PyPI:

pip install arhupy

Install from source:

git clone https://github.com/Typeshi-dotcom/arhupy.git
cd arhupy
pip install -e .

Check the CLI:

arhupy --help

Quick Start

from arhupy import Prompt

prompt = Prompt("You are a {role}. Explain {topic} in {style}.")
filled = prompt.fill(
    role="fitness coach",
    topic="progressive overload",
    style="simple terms",
)

print(filled)

Output:

You are a fitness coach. Explain progressive overload in simple terms.

Feature Map

Feature	Python API	CLI command
Prompt templates	Prompt	arhupy fill coding
Prompt chains	PromptChain, build_chain()	arhupy chain
Prompt scoring	score_prompt()	arhupy score "..."
Prompt comparison	compare_prompts()	arhupy diff "..." "..."
Prompt storage	save(), load()	arhupy save, arhupy list
Sharing	export_prompt(), import_prompt()	arhupy export, arhupy import
History	add_history(), get_history()	arhupy history, arhupy reuse
Sharing	save_shared(), get_shared()	arhupy share "..."
Claude integration	ClaudeClient, improve_prompt()	arhupy improve
AI generation	generate_prompt()	arhupy generate "..."
Web dashboard	run_server()	arhupy web
Local API	run_api_server()	arhupy api
Plugins	ArhupyPlugin, get_plugin()	arhupy plugin echo "hello"

CLI Command Reference

Score a prompt

arhupy score "You are a fitness coach. Explain progressive overload step by step."

What it checks:

Check	What it looks for
Length	Too short, ideal length, or too long
Role	Phrases like you are, act as, or role
Task clarity	Words like write, explain, generate, analyze, or give
Structure	Placeholders like {role}, {task}, or {question}
Output format	Instructions like bullet points, step by step, or format
Constraints	Words like limit, max, within, or only

Example output:

Score: 8/10

Strengths:
- Good prompt length
- Role is defined
- Clear task defined

Improvements:
- Use placeholders like {role}, {task}, or {question}.

Compare two prompts

arhupy diff "You are a coach" "You are a strict fitness coach"

The comparison shows:

• Length difference
• Word count difference
• Common words
• Words unique to each prompt
• Score for each prompt
• Which prompt looks stronger

Improve a prompt with Claude

arhupy improve "You are a coach" --api-key YOUR_KEY

arhupy uses the built-in ClaudeClient with urllib, so no external request library is needed. For safe local testing, the placeholder YOUR_KEY returns a demo improvement without making a real API call.

Generate a prompt from an idea

arhupy generate "fitness coach" --api-key YOUR_KEY

The generator expands a short idea into a full prompt with a role, task, structure, and constraints. With YOUR_KEY, it returns a realistic local demo prompt. With a real Claude API key, it asks Claude to generate the prompt.

Save, list, export, and import prompts

arhupy save workout "You are a fitness coach. Create a {plan} for {goal}."
arhupy list
arhupy export prompts.json
arhupy import prompts.json

Saved prompts live in:

arhupy_library.json

The file is created in your current working directory.

Use built-in templates

arhupy templates
arhupy template coding
arhupy fill coding

Built-in templates:

Name	Template
fitness	You are a fitness coach. Create a {plan} for {goal}.
coding	You are a senior developer. Explain {concept} in simple terms.
writing	You are a writer. Write a {type} about {topic}.
business	You are a business expert. Analyze {idea} and suggest improvements.

Build a prompt chain

arhupy chain

Enter one prompt per line. Submit an empty line to finish.

Example:

Enter prompt 1: You are a coach.
Enter prompt 2: Explain progressive overload.
Enter prompt 3: Use bullet points.
Enter prompt 4:

Output:

You are a coach.
Explain progressive overload.
Use bullet points.

Work with prompt history

Commands that score, diff, or improve prompts automatically save them to history.

arhupy history
arhupy history 5
arhupy reuse 1
arhupy reuse 1 --score
arhupy compare-history 1 2

History lives in:

arhupy_history.json

Export and import session history

arhupy export-history history.json
arhupy import-history history.json

Imported history entries are merged safely. Duplicate entries are skipped.

Start interactive mode

arhupy interactive

Interactive mode lets you choose actions from a menu:

1. Score prompt
2. Improve prompt
3. Compare with another prompt
4. Save prompt
5. Exit
6. Fill template
7. Build prompt chain
8. Compare history prompts
9. Export history
10. Import history
11. Share prompt
12. Generate prompt from idea

Start the web dashboard

arhupy web

Then open:

http://localhost:8000

The dashboard includes:

• Score, Compare, and Improve tabs
• Prompt scoring with strengths and improvements
• Prompt comparison
• AI prompt improvement with a Claude API key field
• Saved prompt display
• Save prompt button

Improved Web Dashboard

The local dashboard uses only Python's standard-library http.server, but gives you a cleaner browser interface for day-to-day prompt work.

arhupy web

Open http://localhost:8000 and use:

Tab	Inputs	Output
Score	One prompt textarea	Score, strengths, and improvements
Compare	Two prompt textareas	Differences, prompt scores, and better prompt
Improve	Prompt textarea plus API key field	Claude-improved prompt text

Start API mode

arhupy api

Then call the local API at:

http://localhost:8001

Endpoints:

Endpoint	Method	Body
/score	POST	{ "prompt": "You are a coach" }
/diff	POST	{ "p1": "Prompt one", "p2": "Prompt two" }
/improve	POST	{ "prompt": "You are a coach", "api_key": "YOUR_KEY" }

Example with curl on macOS/Linux:

curl -X POST http://localhost:8001/score \
  -H "Content-Type: application/json" \
  -d '{"prompt":"You are a coach"}'

Example with curl on Windows PowerShell:

curl.exe -X POST "http://localhost:8001/score" -H "Content-Type: application/json" --data-raw '{""prompt"":""You are a coach""}'

Run a plugin

arhupy plugin echo "hello"

Output:

Echo: hello

Share a prompt locally

arhupy share "You are a coach. Explain progressive overload step by step."

Output:

Share link: http://localhost:8000/share/abc123

Start the dashboard with arhupy web, open the share link, and arhupy shows the prompt with its score, strengths, and improvements.

Python API Examples

Prompt

from arhupy import Prompt

prompt = Prompt("You are a {role}. Speak in {language}.")

print(prompt.fill(role="coding assistant", language="English"))
prompt.preview()
prompt.reset()

PromptChain

from arhupy import Prompt, PromptChain

system = Prompt("System: {instruction}")
user = Prompt("User: {task}")

system.fill(instruction="Be concise and practical.")
user.fill(task="Explain prompt chaining.")

chain = PromptChain([system, user])
print(chain.build())

Simple prompt chain from strings

from arhupy import build_chain

final_prompt = build_chain([
    "You are a senior developer.",
    "Explain recursion in simple terms.",
    "Use one short example.",
])

print(final_prompt)

Prompt scoring

from arhupy import score_prompt

result = score_prompt(
    "You are a {role}. Explain {task} step by step in bullet points within 200 words only."
)

print(result["overall_score"])
print(result["strengths"])
print(result["feedback"])

Prompt comparison

from arhupy import compare_prompts

result = compare_prompts(
    "You are a coach",
    "You are a strict fitness coach",
)

print(result)

Save and load prompts

from arhupy import Prompt, load, save

prompt = Prompt("Summarize this in {style}: {text}")
save("summarizer", prompt)

loaded = load("summarizer")
print(loaded.fill(style="plain English", text="Prompt engineering is useful."))

Export and import one prompt

from arhupy import Prompt, export_prompt, import_prompt

prompt = Prompt("Write a {tone} email about {topic}.")
prompt.fill(tone="friendly", topic="a project update")

export_prompt(prompt, "email_prompt.json")
restored = import_prompt("email_prompt.json")

print(restored)

Export and import a prompt chain

from arhupy import Prompt, PromptChain, export_chain, import_chain

system = Prompt("System: {instruction}")
user = Prompt("User: {request}")

system.fill(instruction="Be concise.")
user.fill(request="Summarize this report.")

chain = PromptChain([system, user])
export_chain(chain, "summary_chain.json")

restored = import_chain("summary_chain.json")
print(restored.build())

Token estimation

from arhupy import estimate_tokens

tokens = estimate_tokens("A short prompt for an LLM.")
print(tokens)

Built-in templates

from arhupy import Prompt, get_template, list_templates

print(list_templates())

template = get_template("coding")
prompt = Prompt(template)
print(prompt.fill(concept="recursion"))

Prompt history

from arhupy import add_history, get_history, get_prompt_by_index

add_history("You are a coach. Explain progressive overload.")

print(get_history(limit=1))
print(get_prompt_by_index(1))

History comparison

from arhupy import add_history, compare_history

add_history("You are a coach.")
add_history("You are a fitness coach. Explain warmups step by step.")

print(compare_history(1, 2))

Claude integration

from arhupy import Prompt, ClaudeClient

client = ClaudeClient(api_key="your-api-key")

prompt = Prompt("You are a {role}. Answer this: {question}")
response = client.ask_with_template(
    prompt,
    role="fitness coach",
    question="What is progressive overload?",
)

print(response)

AI prompt improvement

from arhupy import improve_prompt

improved = improve_prompt("You are a coach", api_key="your-api-key")
print(improved)

AI prompt generator

from arhupy import generate_prompt

prompt = generate_prompt("fitness coach", api_key="YOUR_KEY")
print(prompt)

Plugin system

Bundled echo plugin:

from arhupy import get_plugin

plugin = get_plugin("echo")
print(plugin.run("hello"))

Create a new plugin inside arhupy/plugins/:

from arhupy.plugins import ArhupyPlugin


class EchoPlugin(ArhupyPlugin):
    name = "echo"

    def run(self, text):
        return f"Echo: {text}"

Sharing prompts

from arhupy import get_shared, save_shared

share_id = save_shared("You are a coach. Explain warmups step by step.")
print(f"http://localhost:8000/share/{share_id}")
print(get_shared(share_id))

Local Files Created By arhupy

arhupy keeps local project data in your current working directory:

File	Purpose
arhupy_library.json	Saved prompt templates
arhupy_versions.json	Versioned prompt snapshots
arhupy_history.json	Prompt history
arhupy_shared.json	Local share-link prompts

These files are ignored by the included .gitignore so private prompt data does not get committed by accident.

Design Goals

• No external runtime dependencies
• Beginner-friendly Python APIs
• Simple JSON storage
• Clear command-line output
• Local-first workflows
• Easy GitHub and PyPI publishing
• Extensible plugin architecture

Development

Clone and install in editable mode:

git clone https://github.com/Typeshi-dotcom/arhupy.git
cd arhupy
pip install -e .

Run tests:

python -m unittest

Build package artifacts:

python -m build

Deployment

arhupy can run its web dashboard in deployment environments that provide a PORT environment variable.

The deployment entry point is:

python app.py

The server binds to 0.0.0.0 and reads the port from PORT, defaulting to 8000 locally:

PORT=8080 python app.py

On Windows PowerShell:

$env:PORT = "8080"
python app.py

requirements.txt is intentionally minimal because arhupy has no external runtime dependencies.

Contributing

Contributions are welcome. See CONTRIBUTING.md for setup instructions, test commands, and pull request guidance.

License

This project is licensed under the MIT License. See LICENSE for details.