Visual LLM workflow builder for everyone
flomoji enables AI conversations and workflow automation without complex setup. Supporting various AI providers including OpenAI, Anthropic, and local LLMs, it streamlines your work through prompt chaining and multi-step processing.
ใใฉใฆใถ๏ผใจAPIใญใผ๏ผใ ใใงๅใใ่ปฝ้LLMใจใผใธใงใณใ้็บใใผใซ
flomojiใฏใ่ค้ใช็ฐๅขๆง็ฏใชใใซAIใจใฎๅฏพ่ฉฑใใฏใผใฏใใญใผ่ชๅๅใๅฎ็พใใ่ปฝ้ใใผใซใงใใ OpenAIใAnthropicใใญใผใซใซLLMใชใฉๆงใ ใชAIใใญใใคใใผใซๅฏพๅฟใใใใญใณใใใใงใผใณใซใใๅคๆฎต้ๅฆ็ใงไฝๆฅญใๅน็ๅใงใใพใใ
- ๐ Browser-only: No installation required, runs directly in your browser.
- ๐ฌ AI Chat: Real-time conversations with AI, with chat history management.
- ๐ Workflows: Automate tasks using prompt chaining.
- โ๏ธ JavaScript Code Execution: Secure in-browser JavaScript processing with Web Worker isolation and built-in templates.
- ๐ Web Integration: HTTP requests, web search, and Web API connector for dynamic data processing.
- โฐ Scheduled Workflows: Automate workflows with cron-based scheduling.
- ๐จ Enhanced Node UI: Color-coded handles and toggleable labels for better visual clarity.
- ๐ Multi-LLM Support: Works with OpenAI, Anthropic, and local LLMs.
- ๐ Data Management: Easily manage and export your data.
Visual workflow builder with node-based interface for creating AI-powered automation workflows.
Instant access: ๐ Launch flomoji - No download required!
The easiest way to use flomoji locally is directly in your browser without any setup.
- Download: Click here to download the latest version.
- Unzip: Extract the downloaded ZIP file.
- Open: Open the
index.htmlfile in your web browser (like Chrome, Firefox, or Edge).
That's it! You can start using the application immediately.
- Settings (่จญๅฎ): Choose your LLM provider (like OpenAI) and enter your API key or endpoint. Test the connection to make sure it's working.
- Chat (ใใฃใใ): Start a conversation with the AI. Your chat history is saved automatically.
- Workflow (ใฏใผใฏใใญใผ): Define a series of prompts to automate a complex task. Input your data, select a workflow, and run it.
- Data (ใใผใฟ): View, export, or back up your conversation history and other data.
- โ๏ธ JavaScript Code Execution: Process data with custom JavaScript code in a secure sandboxed environment
- Built-in templates for JSON filtering, text processing, data aggregation
- Web Worker isolation for security and performance
- 5-second timeout protection and memory limits
- ๐ Web Integration Nodes:
- HTTP Request Node: Make API calls with customizable headers and methods
- Web Search Node: Search Google, Brave, or Bing directly from your workflows
- ๐ Control Flow: If/While conditions, variable management, and scheduled execution
- ๐ Structured Data Extraction: Extract and validate JSON-schema compliant data from unstructured text
- Rule-based extraction with regex patterns
- LLM-based extraction with automatic prompt generation
- Schema validation and automatic data repair
The Structured Extraction feature uses a two-node pattern to extract and validate data:
โโโโโโโโโโโโโโโโ โโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโ
โ Input Node โโโโโโถโStructuredโโโโโโถโ LLM Node โโโโโโถโ Schema โ
โ (Text Data) โ โExtractionโ โ (Process Prompt)โ โ Validator โ
โโโโโโโโโโโโโโโโ โโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโ
โ โ
โผ โผ
(Direct output (Validated JSON)
if rules match)
How it works:
- Structured Extraction Node attempts rule-based extraction first
- If rules fail or in LLM mode, it generates a prompt for the LLM
- LLM Node processes the prompt and returns JSON
- Schema Validator Node validates and optionally repairs the JSON
This DAG-compliant pattern ensures clean data flow without cycles. See Structured Extraction Guide for detailed documentation.
If you want to run the app locally or contribute to development:
Run locally:
# Install dependencies
pnpm install
# Start the development server
pnpm run devThen, open http://localhost:5173 in your browser.
Build for production:
# Create a production build
pnpm run buildThe output files will be in the dist folder. You can deploy this folder to any static hosting service.
You can programmatically run any workflow from an external script or your browser's developer console using the runFlomojiWorkflow global function.
Usage:
runFlomojiWorkflow(workflowId, inputData)
.then(result => {
console.log('Workflow completed successfully:', result);
})
.catch(error => {
console.error('Workflow execution failed:', error);
});Parameters:
workflowId(string): The ID of the workflow you want to run. You can find the ID in the workflow management screen.inputData(object, optional): An object containing the initial input data for the workflow. Defaults to an empty object.
Example:
To run a workflow with the ID my-awesome-workflow and provide it with some initial data, open your browser's developer console and enter:
runFlomojiWorkflow('my-awesome-workflow', {
userInput: 'This is a test message from the console.'
});This will execute the workflow and print the final output variables to the console upon completion.
This repository includes automated CI/CD workflows using GitHub Actions:
- Quality Checks: Automatic linting, testing, and building on every PR
- Auto Review: Comments with quality check results
- Auto Merge: Approved PRs are automatically merged after passing all checks
- Automatic Deployment: Every merge to
maintriggers automatic deployment to GitHub Pages - Build & Test: Ensures all tests pass before deployment
- Live Updates: Changes are automatically reflected at https://hama-jp.github.io/flomoji/
- Create a feature branch from
main - Make your changes and commit
- Create a Pull Request
- The CI system will automatically run quality checks
- Once approved, the PR will be auto-merged and deployed
- Your API keys are stored only in your browser's LocalStorage.
- All data processing happens on the client-side (in your browser).
- No data is ever sent to or stored on a server.
- JavaScript code execution is sandboxed using Web Workers with restricted global access.
- Timeout protection prevents infinite loops and resource exhaustion.
This project is licensed under the Apache 2.0 License.
If you have any issues or feature requests, please open an issue on GitHub.