Hephaes

Hephaes is a local-first open-source robotics log indexing and dataset conversion stack, built to turn raw ROS and MCAP logs into clean, searchable, reproducible datasets on your own machine.

Core Workflow

• register local .bag and .mcap logs by file path or directory scan
• index logs to extract duration, start and end time, topic summaries, message counts, sensor types, and raw metadata
• browse assets in a sortable/filterable inventory with tags and indexing status
• inspect each asset in a detail page with topic breakdowns, related jobs, and conversion history
• convert selected sessions to Parquet or TFRecord with custom mappings, compression, resampling, and manifests
• browse generated outputs and open artifact content directly
• track durable jobs for indexing and conversion

Quick Start

Requires Rust, Node.js, and Python 3.11+.

This launches the local desktop UI and starts the bundled backend sidecar on your machine.

pip install -r requirements.txt
cd frontend && npm install
npm run tauri:dev:sidecar

Using Hephaes as a Python Package

pip install hephaes

If you want the Python CLI or library instead of the desktop app, start with hephaes/README.md. It covers CLI workflows, the Workspace API, and direct conversion examples.

For product and usage documentation, see docs.hephaes.ai.

Enterprise Features

We are also building features for enterprise that sit on top of the local OSS core.

Planned enterprise features include:

• cloud ingestion from buckets, remote URLs, and managed connectors
• multi-user authentication, organizations, workspaces, roles, and ownership
• shared catalogs with team-wide browsing and admin views
• saved searches, shared presets, and richer metadata search
• managed conversion jobs with retries, scheduling, and distributed execution
• first-class named datasets with versioning, sharing, approvals, and publishing
• dataset lineage with hashes, creators, schema governance, and audit history
• remote replay and visualization with access control and collaboration
• team workflows for outputs, approvals, integrations, and downstream compute actions

If you are interested in being a design partner, please reach out to hello@hephaes.ai

Repository Layout

• frontend/: React + Vite UI with a Tauri desktop shell
• backend/: FastAPI service
• hephaes/: Python package for log inspection and dataset conversion
• docs/: project documentation site (Nextra)

Python Setup

pip install -r requirements.txt

Or individually:

pip install -e "./hephaes[dev]"
pip install -e "./backend[dev]"