Syngnat/GoNavi

GoNavi – A lightweight, AI‑enhanced database client

What it is – GoNavi is a cross‑platform desktop application (and optional headless CLI) for connecting to a huge variety of data sources: classic relational databases (MySQL, PostgreSQL, Oracle, …), caches (Redis), vector stores (Milvus, Qdrant, Chroma), message queues (Kafka, RabbitMQ, MQTT), time‑series engines, search engines, and many more. It is built with Go + Wails for the native backend and React + TypeScript for the UI, so it runs in a system WebView instead of an Electron shell, keeping the installer under 30 MB and the runtime memory lower than typical Electron‑based tools.

Why it matters for AI/agents – The client ships a first‑class MCP (Multi‑Channel Provider) AI assistant that can:

• Attach live table schemas to prompts so LLMs can generate, explain, or optimise SQL that actually matches the connected database.
• Run slash‑style commands (e.g., /explain SELECT …) that invoke OpenAI, Gemini, Claude, or any OpenAI‑compatible endpoint.
• Hand the generated queries to the built‑in executor while keeping credentials on the host, enabling safe “agent‑ready” workflows.

Key capabilities

Feature	What you get
AI‑aware query writing	Schema‑aware completions, one‑click generation of SELECT/INSERT/UPDATE statements, and automatic EXPLAIN/OPTIMIZE suggestions.
Virtualised DataGrid	Handles millions of rows without freezing; supports in‑place edit, batch CRUD, transaction commit/rollback, and export to CSV/Excel/JSON/Markdown.
Rich connectivity	Built‑in drivers for >20 sources; optional driver agents add many more (ClickHouse, Elasticsearch, MongoDB, etc.). Supports SSH tunnels, proxies, and DSN‑based custom drivers.
Audit & observability	Execution logs with timing, redacted audit centre, and optional web‑server mode for remote access.
Packaging	Desktop installers (~20‑26 MB) for Windows/macOS/Linux; Docker/K8s/Helm images; a separate CLI (gonavi) for scripting.
Extensible AI stack	Plug‑in‑style support for OpenAI, Gemini, Claude, DeepSeek, Kimi, Grok, and any OpenAI‑compatible API (e.g., APISmart, HuaLongAI).

How you use it

1. Install – download a binary from the GitHub releases page (Windows AMD64/ARM64, macOS AMD64/ARM64, Linux WebKitGTK 4.0/4.1). The installer is ~20 MB; the running UI typically uses ~400‑800 MB of RAM on Linux.
2. Add a connection – pick a data source, fill the DSN (or use the built‑in SSH tunnel UI), and test the connection.
3. Open the workbench – the left pane shows connections, the central pane is a Monaco‑based SQL editor with AI‑assistant side‑panel, and the bottom pane displays results in a virtualised grid.
4. Ask the AI – type a natural‑language request or a slash command; the assistant injects the current schema, calls the selected LLM, and returns a ready‑to‑run SQL snippet.
5. Run / edit – execute the query, edit rows directly in the grid, or export the result set.
6. CLI automation – use the gonavi binary for headless operations (gonavi query …, gonavi export …, etc.) in CI pipelines or containers.

Who it’s for – developers, data engineers, and DBA‑type users who need a fast, low‑overhead GUI that can also act as a front‑end for LLM‑driven data‑access agents. Because credentials never leave the host, it is suitable for environments where security policies forbid exposing passwords to external services.

Where to learn more –

• Website: https://gonavi.org
• GitHub repo: https://github.com/Syngnat/GoNavi (releases, issue tracker, docs)
• Chinese README: README.zh-CN.md
• Sponsor links: APISmart and HuaLongAI (provide the AI back‑ends used by the assistant).

---

All information above is taken directly from the project’s README; no additional features have been inferred.