shirube

You hand-write less SQL than ever.
You still own the schema.
_{標べ (shirube) — Japanese for a signpost. It reads your database as a map.}

🚧 Status: Beta. Both the explorer and the AI navigator work today. shirube is pre-1.0, so things may still change.

Ask a question — see the answer on the map.

💡 Why shirube

You hand-write less SQL than you used to — tables become models and methods in your code, and in the AI-coding era much of even that code is written for you. But the schema is still there underneath, and it is still yours to understand: to judge whether the code an AI wrote is right — which tables it reads, which columns it touches, which relationships it follows — and to ask the tools for the right thing in the first place.

A plain list of tables is a poor place to see that. Open them one at a time and you get columns, not connections — not what a table references, what references it, or which hops lead to the data you are after. Relationships are the very thing you need to follow, and a list is where they hide, so shirube lets you read a database as a map instead — just as useful when you drop into a project with hundreds of undocumented tables and need your footing fast.

AI changed who writes the SQL. shirube changes how we understand databases.

🎯 What shirube does

shirube turns a database into a map you can read. It opens on an interactive ER diagram and gives you two ways to find your footing: steer it yourself — search for a table, focus on it, follow its relationships outward — or just ask, and the AI navigator answers on the map. Either way you see how everything connects without reading DDL or writing a single query.

It is not another SQL IDE or database administration console — there is no query editor, and nothing that ever writes. shirube is a tool for understanding a database.

✨ Features

Everything below works today, in the beta:

• 🤖 AI navigator — ask in plain language and get an answer on the map, lighting up the tables involved. Bring your own Claude or OpenAI key, or point it at a local model.
• 🗺️ ER diagram home — auto-generated and centred on the most-connected table; see a table and its immediate neighbours, and travel outward one hop at a time.
• 📋 Table detail — columns with their types, primary keys and nullability, plus relationships split into references and referenced by.
• 🔗 Relationship navigation — click a related table to glide the map over to it.
• ✏️ Draw your own relationships — when the database declares no foreign key, shirube can't see that two tables are related; draw the missing link yourself, column to column, as a dotted edge, saved per connection and never touching your database.
• 🌳 Browse the schema as a tree — a collapsible tree of the whole database (schemas → tables and views → columns); click any object to travel the map to it.
• 👁️ Data preview — read a table or view's actual rows in a drawer, with click-to-sort columns, simple filters and paging.
• ⚡ Instant search — jump straight to any table or column.
• 🔐 Saved connections — passwords kept in your operating system's keychain, never in a config file.
• 🌗 Light and dark themes.

🤖 The AI navigator

Ask in plain language and get your answer on the map: the navigator reads your schema metadata to guide you and lights up the tables involved, with every object name a link that flies the map to it. It is a navigator, not a SQL generator — it never writes or runs a query.

Use your own Claude or OpenAI key, or a local model such as Ollama, configured in Settings → AI navigator. Whichever you pick, the request goes straight from your machine to that model — never through a shirube server (see Safe by design below). The navigator is entirely optional; the explorer works fully without it.

🛡️ Safe by design

• 🔒 Read-only — every connection is opened read-only with a statement timeout. shirube cannot modify your database — no writes, no schema changes, ever.
• 💻 Local-first — it runs on your machine and binds to 127.0.0.1 only. Your database credentials and data never leave your computer.
• 📝 Metadata-only logging — the local log records errors and request timings, but never the values in your data.
• 🤖 Your data, your provider — the AI navigator talks straight from your machine to the model you chose; only question-relevant schema metadata is sent, and a local model sends nothing off your machine at all.

🚀 Getting started

You won't need a database of your own to start — shirube ships with a small sample you can open straight away. To explore your own data, have a PostgreSQL or SQLite database ready. Either way you'll want a way to run a Python application without installing it permanently.

pipx run shirube   # with pipx
# or
uvx shirube        # with uv

💡 pipx and uvx run a Python command-line app in its own throwaway environment, so shirube is never installed permanently (uvx ships with uv).

Either command starts a small local server and opens shirube in your browser. A Sample database (Chinook) is already in your saved connections — open it to explore straight away — or add your own connection (PostgreSQL or SQLite), and shirube inspects the schema and opens the ER diagram.

💡 shirube connects with whatever credentials you give it. A read-only role with CONNECT and SELECT is all it needs — and all it should have.

🤖 To use the AI navigator, add a provider in Settings → AI navigator — your own Claude or OpenAI key, or a local model. It's optional; the explorer works fully without it.

📚 Learn more

• Full README and roadmap — https://github.com/shou-taro/shirube
• Contributing — https://github.com/shou-taro/shirube/blob/main/CONTRIBUTING.md
• Licence — GNU AGPL-3.0