Prompt Improvement Studio — turn rough AI prompts into production-ready ones automatically
Project description
Prompt Improvement Studio
Turn rough AI prompts into production-ready ones — automatically.
Improve system prompts for ChatGPT, Claude, Copilot, Cursor, and any AI agent. No API key required. One command, copy-paste result.
Try in 60 seconds (offline):
pip install recursive-intelligence && ri-engine improve --template customer-support
Copy the improved prompt into ChatGPT, Claude, Cursor, or your AI tool. Full discoverability guide →
Run in the cloud (free tier): Codespaces + Colab guide — use Colab for quick API experiments, Codespaces for the full CLI, runbook, and plateau workflow.
License: MIT · Changelog: CHANGELOG.md · Grow usage: discoverability guide
Who it's for
| You are… | Start here |
|---|---|
| Developer (Cursor, Copilot, etc.) | ri-engine improve --template coding-assistant |
| Support or ops lead | ri-engine improve --template customer-support |
| Evaluating the tool (no API key) | ri-engine demo |
| Python integrator | from ri_engine import improve |
Install (2 minutes)
From PyPI (recommended):
pip install recursive-intelligence
ri-engine improve --template customer-support
From source:
git clone https://github.com/russfranky/recursive-intelligence.git
cd recursive-intelligence
pip install -e ".[all]"
Works offline immediately — no API key needed for the built-in demo.
Repository setup: see docs/standalone_repo.md — clone, Codespaces, or ./scripts/publish-private-repo.sh for mirrors.
3-Step Quick Start
1. See what's available
ri-engine templates
2. Improve a prompt
ri-engine improve --template customer-support
Or use your own starting prompt:
ri-engine improve \
--seed "You are a helper. Answer questions." \
--goal "Help customers resolve billing issues in one conversation"
3. Copy the result
The tool shows Your Improved Prompt — Copy This. Paste it into your AI assistant as the system prompt. Done.
Results are also saved to output/your_improved_prompt.json.
See the proof
ri-engine demo
Runs 6 real business scenarios (support, sales, security, coding, research, code review) and shows before/after quality — no jargon.
Ready-made templates
| Template | Best for |
|---|---|
customer-support |
Help desks and support teams |
code-review |
Engineering teams reviewing pull requests |
coding-assistant |
Developers using Cursor, Copilot, and similar coding agents |
sales-outreach |
Sales emails that get replies |
security-response |
Security alert triage and response |
research-analyst |
Structured research and analysis |
publication-agent |
Preparing a repo for public GitHub release |
discoverability-agent |
Growing repo adoption — README, keywords, CTAs |
occams-razor |
Add simplicity pressure across the VSR loop |
ri-engine improve --template occams-razor
Using your own AI provider (optional)
For semantic rewrites with GPT or Claude:
pip install -e ".[openai]"
export OPENAI_API_KEY=sk-...
ri-engine improve --template sales-outreach --provider openai
What happens behind the scenes (plain English)
You don't need to watch or understand this — but for the curious:
- Style match — We pick the best writing tone for your task (clear, friendly, formal, etc.)
- Try versions — The system drafts multiple improved versions of your prompt
- Pick the best — Each version is scored for clarity, usefulness, and completeness
- Keep what works — Winning patterns carry forward to the next round
- Deliver — You get one polished prompt, ready to paste
For developers & power users
ri-engine improve --template customer-support --expert # technical output
ri-engine expert benchmark # full benchmark
Python API
from ri_engine import improve
result = improve(
seed_prompt="You are a helper.",
objective="Help customers resolve billing issues in one conversation.",
)
print(result.improved_prompt)
See docs/publication.md for the publication guide, red-team review, and API reference.
See docs/technical_reference.md for architecture, VSR loop, and research background.
Support
ri-engine # welcome & quick help
ri-engine templates # list templates
ri-engine demo # see proof it works
Quality guarantee: Every template scenario in our demo improves from failing grade (F) to production-ready (A).
Powered by recursive improvement research. Built for teams who need results, not jargon.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file recursive_intelligence-0.1.2.tar.gz.
File metadata
- Download URL: recursive_intelligence-0.1.2.tar.gz
- Upload date:
- Size: 117.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e5dbd165c06fa89213f3dff21d21d300d9fa7252d0abae61f5b61e83d67fb387
|
|
| MD5 |
0d4675398e454a094bf37fdf80f2700d
|
|
| BLAKE2b-256 |
e5445bfcac870f1ecc2d87350b3bd359d2001aaab79eeba84ff7997e1eb275b1
|
Provenance
The following attestation bundles were made for recursive_intelligence-0.1.2.tar.gz:
Publisher:
publish.yml on russfranky/recursive-intelligence
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
recursive_intelligence-0.1.2.tar.gz -
Subject digest:
e5dbd165c06fa89213f3dff21d21d300d9fa7252d0abae61f5b61e83d67fb387 - Sigstore transparency entry: 1825903808
- Sigstore integration time:
-
Permalink:
russfranky/recursive-intelligence@d4c1694c256cbccc0d6e675da41d09e778c50f2d -
Branch / Tag:
refs/tags/v0.1.2 - Owner: https://github.com/russfranky
-
Access:
private
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@d4c1694c256cbccc0d6e675da41d09e778c50f2d -
Trigger Event:
push
-
Statement type:
File details
Details for the file recursive_intelligence-0.1.2-py3-none-any.whl.
File metadata
- Download URL: recursive_intelligence-0.1.2-py3-none-any.whl
- Upload date:
- Size: 138.5 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
0786e283d0c282922bb69da5b8784c641b27a2570a3c10fc14f5bfd236dbcb42
|
|
| MD5 |
8230c2ab801a44401453ff24b90e42bb
|
|
| BLAKE2b-256 |
c005184c98bbbff3a49f7f65b54e026570eddb5ecc1f5090f0aedd35628dc202
|
Provenance
The following attestation bundles were made for recursive_intelligence-0.1.2-py3-none-any.whl:
Publisher:
publish.yml on russfranky/recursive-intelligence
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
recursive_intelligence-0.1.2-py3-none-any.whl -
Subject digest:
0786e283d0c282922bb69da5b8784c641b27a2570a3c10fc14f5bfd236dbcb42 - Sigstore transparency entry: 1825903831
- Sigstore integration time:
-
Permalink:
russfranky/recursive-intelligence@d4c1694c256cbccc0d6e675da41d09e778c50f2d -
Branch / Tag:
refs/tags/v0.1.2 - Owner: https://github.com/russfranky
-
Access:
private
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@d4c1694c256cbccc0d6e675da41d09e778c50f2d -
Trigger Event:
push
-
Statement type: