AbstractFlow documentation

AbstractFlow is a Python library (plus a reference web UI) for authoring and executing durable AI workflows.

This doc set is intentionally: - actionable (commands and entrypoints you can run) - evidence-based (each page points to the code that implements the described behavior)

Start here

• Published documentation site: https://www.lpalbou.info/AbstractFlow/
• Project overview + install: ../README.md
• Getting started (quickstarts + “waiting” runs): getting-started.md

Find what you need

• API reference (high-level): api.md
• Architecture (how the pieces fit): architecture.md
• FAQ (common questions): faq.md
• VisualFlow JSON format: visualflow.md
• Visual editor (run the reference UI): web-editor.md
• CLI (bundle, serve): cli.md

Repo policies

• Changelog: ../CHANGELOG.md
• Contributing: ../CONTRIBUTING.md
• Security reporting: ../SECURITY.md
• Acknowledgments: ../ACKNOWLEDGMENTS.md

Code map (evidence)

• Public Python API exports: ../abstractflow/init.py
• Programmatic flows (IR re-export from AbstractRuntime): ../abstractflow/core/flow.py
• Flow execution convenience: ../abstractflow/runner.py (FlowRunner)
• VisualFlow schema (portable JSON): ../abstractflow/visual/models.py
• VisualFlow host wiring + execution: ../abstractflow/visual/executor.py
• VisualFlow interface contracts: ../abstractflow/visual/interfaces.py
• WorkflowBundle helpers (thin wrapper): ../abstractflow/workflow_bundle.py
• CLI entrypoint: ../abstractflow/cli.py
• Web backend (FastAPI): ../web/backend/main.py, ../web/backend/routes/
• Web frontend (React): ../web/frontend/src/