Getting Started - VVG-web/aurora-studio GitHub Wiki

Getting started

Wiki home Β· next: The cycle

What you need

  • Python 3.11+ (3.14 recommended; 3.10 works with a warning, 3.9 is not supported) and git. Runtime dependencies are listed in requirements.txt; the optional ones are installed from the "Install" page of the panel.
  • Optional: pandoc (export to docx/pdf), beautifulsoup4 + markdownify (Confluence sync), markitdown, openpyxl, pypdf (office documents). The panel's Install section shows what is missing and which commands it affects.
  • Optional: an OpenAI-compatible LLM gateway for the built-in agent. Everything that does not need a model works without it.

Install in three commands

git clone https://github.com/<org>/aurora-studio.git
cd aurora-studio
python3 aurora.py new /path/to/your-project     # deploy into a new or existing folder; answer the questions
python3 aurora.py cockpit                       # open the local control panel

No terminal? Double-click start-aurora.command (macOS) or start-aurora.bat (Windows): it checks Python and git, offers to install what is missing and opens the panel. Without a terminal the setup questions are skipped and the project is deployed with defaults β€” finish the settings later with python3 aurora.py setup <project> or in the panel.

What you get

A project folder with a fixed layout (the same in every Aurora project): Sources/ (mirrors, written only by sync), Raw/ (immutable evidence), AuroraKnowledgeDB/ (the cards), Artifacts/, Deliverables/, Workspaces/, and .aurora/ with a copy of the engine. Details: INSTALL.

A teammate joins a project that already exists

Clone the project and open it in the panel β€” no terminal. .aurora/ (the engine copy) does not travel in git, so a clone arrives without it; when you pick the project the panel completes it from the kit by itself and says what it created (kit:update --engine-only: only the paths under .aurora/, project files and git stay untouched). The rest of the update β€” AGENTS.md, templates, the .gitignore rules β€” changes git, so it waits for Version β†’ Update preview β†’ Apply. The state of the engine (ok, missing, incomplete, stale) is visible in Version and in kit:doctor.

The first fifteen minutes

  1. Check readiness. In the panel, kit:doctor (or python3 aurora.py doctor <project>): it names what is missing β€” a git repository, the config, an AGENTS.md, a secret that must not be in git.
  2. Fill in the settings. aurora.config.yaml holds URLs, spaces, project keys and JQL (it is committed); tokens go to .env.aurora.local (it is git-ignored). Copy aurora.env.local.example to start. See INSTALL Β§3–4.
  3. Open the project in the panel and press "Update the base". It syncs the mirrors, builds cards in batches, writes theses, links the cards, computes trust and commits every lap to git. You may stop it at any time: what is done stays valid. β†’ The cycle
  4. Look at Health. Share of knowledge, lint errors, what is left for a human (ops:todo).
  5. Ask something. The Ask section answers from cards only and puts a link on every claim. β†’ Asking the base

Where to go next

If you are… Read
an analyst working daily Daily work, then Practice
setting up the team Team rules, Looking after the base
connecting Confluence and Jira Sources and mirrors
connecting an LLM The agent

This wiki is generated from the wiki/ folder of the repository β€” edit the pages there, changes made here are overwritten. Β· RU-Home

⚠️ **GitHub.com Fallback** ⚠️