Ask a Firestore question in plain English and inspect the query behind the answer.
Firestore Query Studio is an Electron and React app for exploring Firestore through natural language. An LLM proposes a JSON query plan, Zod validates its structure, and the desktop executor reads from a live Firebase project or emulator. Results include sorting, JSON/CSV export, explanations, optional AI insights, and query history. A separate PWA build uses Firebase Auth and the Web SDK.
From question to query
The planner uses inferred collection schema to turn a question into a structured query plan. Invalid JSON and plans that fail Zod validation are rejected. The Explain view exposes the plan and rationale so the user can inspect what the app is asking Firestore to do.
Bounded Firestore reads
The Firestore executor supports queries and bounded scans, applies the lower of the plan and profile scan caps, and reports truncation. Missing composite-index errors can provide a Firebase Console link. The Firestore query flow is designed for reads rather than database mutation.
Desktop and browser boundaries
Electron runs Firestore reads through the Admin SDK and supports the emulator. The PWA uses Firebase Auth and the Web SDK, so reads depend on Firestore Security Rules. Planning can use an OpenAI-compatible endpoint or desktop CLI backends. AI insights send sampled result rows to the selected model backend, so local storage does not mean inference always stays on the device.
Highlights
- Zod-validated query plans with inspectable explanations
- Live Firestore and emulator profiles on desktop
- Scan caps, truncation warnings, and composite-index links
- Sortable results with JSON/CSV export and query history
- Shared React UI with a separate browser transport