hof-engine 0.1.86

Full-stack Python + React framework for building applications with workflows, tables, UIs, and LLM integration.

hof-engine

Full-stack Python + React framework for building applications with workflows, database tables, UIs, and LLM integration -- all defined as code.

Features

• Tables: Define database schemas as Python classes with auto-generated CRUD APIs
• Functions: Backend operations exposed as API endpoints and CLI commands
• Flows: Workflow DAGs with parallel execution, LLM nodes, and human-in-the-loop
• UIs: Native React components with hot reload via Vite
• Cron Jobs: Scheduled tasks with Celery Beat
• CLI: Full CLI access to all features
• Admin Dashboard: Visual flow viewer, table browser, execution history, and logs
• LLM Integration: First-class support via llm-markdown with structured outputs
• Self-Contained Docs: Bundle Markdown documentation with your app, served and rendered in the admin UI

Quick Start

pip install hof-engine
hof new project my-app
cd my-app

Define a table:

# tables/task.py
from hof import Table, Column, types

class Task(Table):
    title = Column(types.String, required=True)
    done = Column(types.Boolean, default=False)

Define a function:

# functions/greet.py
from hof import function

@function
def greet(name: str) -> dict:
    return {"message": f"Hello, {name}!"}

Define a flow:

# flows/onboarding.py
from hof import Flow

onboarding = Flow("onboarding")

@onboarding.node
def create_task(user_name: str) -> dict:
    return {"task": f"Welcome {user_name}"}

@onboarding.node(depends_on=[create_task])
def notify(task: str) -> dict:
    return {"notified": True}

Run:

hof db migrate
hof dev

hof fn output: By default, hof fn <name> prints human-friendly tables (for list-like results such as { "rows", "total" }) or a key/value layout for plain dicts. Use --format json when you need machine-readable JSON for scripts or pipes. The -j / --json flag supplies input to the function only, not the output format.

Self-Contained Docs

Every hof application can ship its own documentation. Place Markdown files in a docs/ directory at the project root and they are automatically served at /api/docs and rendered in the admin UI at /docs.

my-app/
  docs/
    index.md          # Overview
    data-model.md     # Table schemas
    api.md            # Function reference

Use optional YAML frontmatter to control titles, section grouping, and sort order:

---
title: Data Model
section: Reference
order: 1
---

No configuration needed — docs_dir defaults to "docs". Set docs_dir="" in hof.config.py to disable. See the Configuration reference for full details.

Requirements

• Python 3.11+
• Node.js 18+ (for React UI)
• PostgreSQL
• Redis

Documentation

See the docs/ directory:

• Getting Started
• Tables Reference
• Functions Reference
• Flows Reference
• UI Reference
• LLM Reference
• Agent Reference
• CLI Reference
• Configuration

Ecosystem

hof-engine is part of the bithof platform:

Repo	Role
hof-engine (this repo)	Core framework (pip package)
hof-components	Reusable modules and templates, copied via hof add
hof-os	Agency operations: deployment, provisioning, billing, design system generation
design-system-<customer>	Per-customer design tokens + Tailwind theme (git submodule in project repos)
customer-acme-test	Example customer project

For application examples, see hof-components/docs/examples/.

Contributing (PR hygiene)

• Prefer squashing noisy WIP commits (duplicate messages, one-line “fix” iterations) on a feature branch before merging to main, so history stays readable.
• When syncing the components artifact (hof/components-manifest.json and related generated output), batch updates into fewer commits instead of many back-to-back chore: update components artifact pushes.

License

MIT

Releasing to PyPI

This repository is configured for Trusted Publishing via GitHub Actions:

• .github/workflows/publish.yml for PyPI

• .github/workflows/publish-testpypi.yml for TestPyPI

• The repository can stay private.

• No PYPI_TOKEN secret is required.

• Publishing to PyPI happens when a GitHub Release is published (or manually via workflow dispatch).

• Publishing to TestPyPI is manual via workflow dispatch.

One-time setup in your PyPI account:

1. Go to PyPI Trusted Publishers.
2. Add a publisher with:
   • PyPI project name: hof-engine
   • Owner/repo: your GitHub repo for this project
   • Workflow: publish.yml (filename only)
   • Environment: pypi
3. In GitHub, create an environment named pypi (recommended to require manual approval).

Optional setup for TestPyPI:

1. Go to TestPyPI Trusted Publishers.
2. Add a publisher with:
   • TestPyPI project name: hof-engine
   • Owner/repo: your GitHub repo for this project
   • Workflow: publish-testpypi.yml (filename only)
   • Environment: testpypi
3. In GitHub, create an environment named testpypi.

Release flow:

1. Bump/update code as needed and push to default branch.
2. Create and push a git tag like v0.1.1 (git tag v0.1.1 && git push origin v0.1.1).
3. The workflow builds and uploads the package to PyPI under your account/project ownership.
4. Optional: also publish a GitHub Release for changelog/visibility.

Manual publish options:

• Publish to PyPI now (no release): run the Publish to PyPI workflow via GitHub Actions Run workflow.
• Publish to TestPyPI: run the Publish to TestPyPI workflow via GitHub Actions Run workflow.