Browser-based tool for optimizing, testing and comparing LLM prompts with your own model API keys
- Deploy
- One-click
- License
- AGPL-3.0-only
- GitHub stars
- 36.7k stars
- Last commit
Self-hosted alternatives to OpenAI Prompt Optimizer
The prompt optimizer in the OpenAI dashboard rewrites a prompt according to current best practices, and OpenAI says its dataset-backed version is being deprecated. One listed app does its main job and runs in your own Cloudflare account.
Each answer comes from the app’s README and source code, compared with the features OpenAI Prompt Optimizer describes on its own site.
OpenAI Prompt Optimizer features, app by app
| Feature | Prompt Optimizer |
|---|---|
| Chat interface that optimizes a prompt using current best practices | Yes |
| Evaluation datasets of test cases and responses | No |
| Annotations and critiques that guide the rewrite | Partly |
| Repeated optimize, test and re-optimize cycles | Yes |
| Testing the optimized prompt before deployment | Yes |
Browser-based tool for optimizing, testing and comparing LLM prompts with your own model API keys
- What a OpenAI Prompt Optimizer user would miss
- It has no stored datasets or good and bad annotations, and the Cloudflare build is a static site that needs your own model API keys and has no password page.
- What it adds
- AGPL-3.0, keeps prompts in browser storage with requests going straight to the model provider, and also ships as a desktop app, a Chrome extension and a Docker image.
- Yes Chat interface that optimizes a prompt using current best practices. One-click optimization with multi-round improvement for system prompts and user prompts, on models from several providers.
- No Evaluation datasets of test cases and responses. No stored datasets of test cases; testing uses variables and multi-turn conversations.
- Partly Annotations and critiques that guide the rewrite. Analysis, single-result evaluation and compare evaluation drive an evaluation-driven smart rewrite; no good or bad annotations on stored outputs found.
- Yes Repeated optimize, test and re-optimize cycles. Multi-round iterative optimization with evaluation of whether a prompt improved.
- Yes Testing the optimized prompt before deployment. Advanced testing mode with context variables, multi-turn tests and function calling, plus a compare mode.
Sources:github.com/linshenkx/prompt-optimizer#readmealways200.comdocs.always200.comdevelopers.openai.com/api/docs/guides/prompt-optimizer