Workflow-Use
Semantic browser automation with deterministic workflow generation and variables.
Quick Start
1. Test Deterministic Workflow Generation (NEW!)
python examples/scripts/deterministic/run_complete_test.py
Generate workflows without LLM for step creation - 10-100x faster, guaranteed semantic steps.
2. Create Your Own Workflow
from workflow_use.healing.service import HealingService
from browser_use.llm import ChatBrowserUse
llm = ChatBrowserUse(model_name="bu-latest")
service = HealingService(llm=llm, use_deterministic_conversion=True)
workflow = await service.generate_workflow_from_prompt(
prompt="Go to GitHub, search for browser-use, get star count",
agent_llm=llm,
extraction_llm=llm
)
3. Run a Workflow
cd /path/to/workflow-use/workflows
python cli.py run-workflow-no-ai my_workflow.json
# If the workflow has variables, the CLI will prompt you interactively:
# Enter value for repo_name (required, type: string): browser-use
Key Features
🚀 Deterministic Workflow Generation
- Direct Action Mapping:
input_text→inputstep (no LLM) - Guaranteed Semantic Steps: 0 agent steps (instant execution, $0/run)
- 10-100x Faster: 5-10s vs 20-40s for LLM-based
- 90% Cheaper: Minimal LLM usage
🎯 Semantic-Only Multi-Strategy Element Finding
- No CSS/XPath: 100% semantic strategies (text, role, ARIA, placeholder, etc.)
- 7 Fallback Strategies: text_exact → role_text → aria_label → placeholder → title → alt_text → text_fuzzy
- Works WITH Browser-Use: Finds element index in DOM state, then uses browser-use's controller
- Fast & Robust: Direct index lookup when strategies match, falls back to AI when needed
- Human-Readable: Workflow YAML contains semantic strategies, not brittle selectors
🔄 Variables in Workflows
- Reusable Workflows: Parameterize dynamic values
- Semantic Targeting: Use
{variable}intarget_text - Auto-Extraction: LLM suggests variables automatically
📊 Real-time Progress Tracking (NEW!)
- Step-by-Step Visibility: See each browser action as it's recorded
- Status Updates: Track workflow processing phases in real-time
- Cloud Integration Ready: Store progress in database for live UI updates
- Debug Friendly: Know exactly where workflow generation fails
- Zero Overhead: Optional callbacks, fully backward compatible
# Track workflow generation progress in real-time
workflow = await service.generate_workflow_from_prompt(
prompt="Search for Python docs",
agent_llm=llm,
extraction_llm=llm,
on_step_recorded=lambda s: print(f"Step {s['step_number']}: {s['description']}"),
on_status_update=lambda msg: print(f"Status: {msg}"),
)
Documentation
- docs/DETERMINISTIC.md - Deterministic workflow generation
- docs/VARIABLES.md - Variables guide
- docs/PROGRESS_TRACKING.md - Real-time progress tracking ⭐ NEW
- QUICK_START_PROGRESS_TRACKING.md - 5-minute integration guide
- examples/README.md - Example scripts
Project Structure
workflows/
├── workflow_use/ # Main package
│ ├── healing/ # Workflow generation & healing
│ │ ├── deterministic_converter.py # NEW: Deterministic conversion
│ │ ├── variable_extractor.py # Auto variable detection
│ │ └── service.py # Main workflow generation
│ ├── workflow/ # Workflow execution
│ │ └── semantic_executor.py # Semantic step execution
│ ├── controller/ # Workflow controller
│ ├── recorder/ # Workflow recording
│ ├── storage/ # Storage logic
│ ├── mcp/ # MCP integration
│ ├── schema/ # Schema definitions
│ └── builder/ # Workflow builder
│
├── backend/ # FastAPI backend service
│ ├── api.py # API entry point
│ ├── routers.py # API routes
│ └── service.py # Business logic
│
├── examples/ # Examples organized by feature
│ ├── scripts/
│ │ ├── deterministic/ # Deterministic workflow examples
│ │ │ ├── run_complete_test.py # ⭐ Test deterministic generation
│ │ │ └── create_deterministic_workflow.py
│ │ ├── variables/ # Variable feature examples
│ │ ├── demos/ # Advanced demos
│ │ └── runner.py # Generic workflow runner
│ ├── progress_tracking_example.py # ⭐ NEW: Real-time progress tracking
│ └── workflows/ # Example workflow JSON files
│ ├── basic/ # Basic workflow examples
│ ├── form_filling/ # Form filling examples
│ ├── parameterized/ # Parameterized workflows
│ └── advanced/ # Advanced workflows
│
├── tests/ # Test files
│ ├── test_button_click.py
│ └── test_recorded_workflow.py
│
├── docs/ # Documentation
│ ├── DETERMINISTIC.md # Deterministic workflows
│ └── VARIABLES.md # Variables guide
│
├── data/ # Runtime & test data
│ └── test_data/ # Test data (tracked in git)
│ ├── form-filling/
│ └── flight-test/
│
├── cli.py # CLI entry point
├── pyproject.toml # Project configuration
└── README.md # This file
Comparison: Deterministic vs LLM-Based
| Feature | Deterministic | LLM-Based |
|---|---|---|
| Generation Speed | ⚡ 5-10s | 🐌 20-40s |
| Generation Cost | 💰 $0.01-0.05 | 💸 $0.10-0.30 |
| Agent Steps | ✅ 0 guaranteed | ❌ Variable |
| Deterministic | ✅ Yes | ❌ No |
| Execution Speed | ⚡ Instant | 🐌 5-45s |
| Execution Cost | 💰 $0/run | 💸 $0.03-0.30/run |
Recommendation: Use deterministic for most workflows (search, click, input, navigate).
Testing
# Test deterministic generation
python examples/scripts/deterministic/run_complete_test.py
# Test variables
python examples/scripts/variables/create_workflow_with_variables.py
# Compare approaches
python examples/scripts/deterministic/test_deterministic_workflow.py
Next Steps
- ✅ Run
examples/run_complete_test.py - ✅ Review the generated workflow JSON
- ✅ Try creating your own workflow
- ✅ Add variables to make it reusable
Metadata
Release files for workflow-use 0.2.11
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| workflow_use-0.2.11.tar.gz | 156.4 kB | Without provenance | Details for workflow_use-0.2.11.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| workflow_use-0.2.11-py3-none-any.whl | Python 3 | none | any | Details for workflow_use-0.2.11-py3-none-any.whl |
Total release size: 338.0 kB
Release files / workflow_use-0.2.11.tar.gz
| Download URL | workflow_use-0.2.11.tar.gz |
|---|---|
| Size | 156.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
43d6f9647905768be21f0909395fd65c374bd65001425c13a956de7660f3c5e7
|
|
BLAKE2b-256 checksum How to use checksums |
3fa56d677abe9955720087ab07a289a530b0cd03e5fd3e0a33bbd376ec419fd4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.9.10 {"installer":{"name":"uv","version":"0.9.10"},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / workflow_use-0.2.11-py3-none-any.whl
| Download URL | workflow_use-0.2.11-py3-none-any.whl |
|---|---|
| Size | 181.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
3e805310450ac3d340733c26830274296b1e5ee4d2c6c1b699569e85e49f652b
|
|
BLAKE2b-256 checksum How to use checksums |
59573b99f6811e9b23c273935eaf6a99ee90e172261ce494a511ac33c1209c13
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.9.10 {"installer":{"name":"uv","version":"0.9.10"},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for workflow-use 0.2.11
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.