https://github.com/chuanqisun/iter

Prompt Engineer's Playground
https://github.com/chuanqisun/iter

devtool gpt openai openai-api playground sandbox

Last synced: 2 months ago
JSON representation

Prompt Engineer's Playground

• Host: GitHub
• URL: https://github.com/chuanqisun/iter
• Owner: chuanqisun
• License: mit
• Created: 2023-05-09T18:36:06.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2025-08-06T00:04:32.000Z (2 months ago)
• Last Synced: 2025-08-06T02:25:30.784Z (2 months ago)
• Topics: devtool, gpt, openai, openai-api, playground, sandbox
• Language: TypeScript
• Homepage: https://chuanqisun.github.io/iter/
• Size: 1.17 MB
• Stars: 11
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Iter

A minimalist frontend for Gen AI Chat models, optimized for rapid prompt iteration.

- **🔒 Privacy first**: Credentials are stored in your browser. All requests directly sent to API with no middleman. Absolutely no tracking.

- **⚡ API endpoint and model hot-swap**: Switch between different APIs and models without losing any chat progress

- **🦉 Adapts to OS/Browser default theme**: Dark theme for happy night owls

- **💅 Markdown parser**: Built-in syntax highlight and copy button for code blocks

- **📋 Smart paste**: HTML pastes as markdown, images as input, and files as attachments

- **🧭 Artifacts**: Live edit and preview code blocks for SVG, HTML, Mermaid, TypeScript, and React

- **💻 Interpreter**: Process uploaded files with TypeScript, with access to NPM registry, a virtual File System, and an LLM prompt API.

- **🖱️ Cursor chat**: Precisely edit the selected text

- **📸 Vision input**: Handle visual inputs with multi-modal models

- **🎙️ Speech input**: Use microphone to input text that can be mixed with typed message

- **📋 Document input**: Interpret PDF and text files without conversion

## Screenshots

Create a runnable program from text

![Two screenshots of the app, one showing gpt generated code for a todo app, another showing the todo app running live](./designs/screenshots/artifact.png)

Recreate the UI of Airbnb with a single screenshot

![Two screenshots of the app, one showing gpt generated code based on user uploaded screen, another showing the code running live](./designs/screenshots/vision.png)

## Supported model providers

- OpenAI\*

  - ✅ GPT-5

  - ✅ GPT-5-mini

  - ✅ GPT-5-nano

  - ✅ codex-mini

  - ✅ o4-mini

  - ✅ o3-pro

  - ✅ o3

  - ✅ o3-mini

  - ✅ GPT-4.5-preview

  - ✅ GPT-4.1

  - ✅ GPT-4.1-mini

  - ✅ GPT-4.1-nano

  - ✅ GPT-4o

  - ✅ GPT-4o-mini

- Anthropic

  - ✅ Claude Opus 4.1

  - ✅ Claude Opus 4

  - ✅ Claude Sonnet 4

  - ✅ Claude 3.7 Sonnet

  - ✅ Claude 3.5 Sonnet

  - ✅ Claude 3.5 Haiku

- Google Generative AI

  - ✅ Gemini 2.5 Pro

  - ✅ Gemini 2.5 Flash

  - ✅ Gemini 2.5 Flash Preview

  - ✅ Gemini 2.5 Flash Lite

  - ✅ Gemini 2.0 Flash

  - ✅ Gemini 2.0 Flash Lite

  - ✅ Gemini 2.0 Flash Thinking

- xAI\*\*

  - ✅ Grok 4

  - ✅ Grok 3

  - ✅ Grok 3 Fast

  - ✅ Grok 3 Mini

  - ✅ Grok 3 Mini Fast

- OpenRouter

  - All chat models

\*See detailed support matrix for [Azure OpenAI](https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/responses?tabs=python-secure#responses-api)  

\*\*xAI models do not support PDF document

## Keyboard shortcuts

Mac users, please use ⌘ instead of Ctrl

| Action               | Shortcut                                                        |

| -------------------- | --------------------------------------------------------------- |

| Send message         | Ctrl + Enter (in any textarea)            |

| Abort action         | Escape (when streaming response)                     |

| Dictate              | Shift + Space (hold to talk)              |

| Open response editor | Enter or double click (when focusing response block) |

| Open artifact editor | Enter or double click (when focusing artifact block) |

| Toggle cursor chat   | Ctrl + K (in artifact or response editor) |

| Rerun artifact       | Ctrl + Enter (in artifact editor)         |

| Exit editor          | Escape (in artifact or response editor)              |

| Select up/down       | ↑ / ↓                                     |

| Create backup        | Ctrl + S                                  |

| Restore backup       | Ctrl + O                                  |

| Export               | Ctrl + Shift + E               |

| Import               | Ctrl + Shift + O               |

## Directives

Directives force the LLM to generate code that performs specific tasks. The code will take effect after manually run in the editor.

### `Run` directive

Include a `run` block in the user message to force the LLM to generate code and output files.

````

```run

```

````

The run block can take optional directives to expose additional APIs to the generated code

**`llm`**: Generate code that can prompt the active LLM model.

````

```run llm

```

````

### `Edit` directive

Include an `edit` block in the user message to force the LLM to generate code that edits the nearest assistant message. Other messages will be hidden.

````

```edit

```

````

## Attachments

You can copy/paste or upload files into each message in one of the following formats:

- **Embedded**: LLM can see the image, PDF, or text content. If using `run` directive, LLM can write code that accesses the content as a file.

- **External**: LLM can only see the metadata of the file (name, size, type, etc.). If using `run` directive, LLM can write code that accesses the content as a file without reading the content.