Product Journal
Documenting our fight against over-engineered BI.
v0.7.0 - Semantic & Connect
"Bridge the gap between raw data and business logic. Connection is everywhere."
🧠 Business Semantic Layer
AI can now map obscure column names to meaningful business terms like "Revenue" or "Gross Margin", eliminating interpretation errors.
🔌 Database Connectors
Break the file limit. Directly connect to MySQL and PostgreSQL databases for local streaming sync of millions of records.
🏎️ Unified Ingestion Pipeline
Re-engineered the ingestion flow with concurrent "shopping cart" style imports and industry-standard Parquet support.
v0.6.1 - v0.6.5 (Hotfixes)
- Self-Fix Enhancement: Improved auto-repair logic for filters with automatic parameter backfilling.
- Pro Excel Export: Multi-column layout support. Exported sheets now include chart snapshots and AI insights.
- UI State Persistence: Sidebar modes and view states are now saved locally and restored on restart.
- Deep Validation: Sampling range for type pre-checks during ingestion increased to 50,000 rows.
v0.6.0 - Insight, Power & Reach
"Bridging the gap between data storytelling and professional control. v0.6.6 makes AI-driven insights interactive and expert-level SQL accessible."
🧠 Business Insight & Visual Anchoring
AI now interprets aggregated chart data to generate narrative conclusions. Visual Anchoring allows you to hover over a finding and see related chart elements highlight in real-time. Included a WYSIWYG editor for refining these AI narratives.
💻 Monaco-Powered SQL Lab
Professional SQL editing is here. We've integrated the Monaco Editor (the engine behind VS Code) with full Schema Intelligence. Features smart autocomplete for tables/columns, auto-formatting, and a workflow to send query results directly to the analysis stream.
🌐 Flow Reports & Web Export
A new Flow Report Mode allows for seamless Markdown content and KPI tile interaction. Export your work as a standalone .html file via Hydration Pack, bundling a data snapshot for interactive sharing. Note: Exports currently reference core libraries via CDN.
- Type Enforcement: Prevents data type errors for long IDs at the source.
- Privacy Guard: Insights only process aggregated results; raw rows never leave the engine.
- Airy Theme V2: Refined visualization with soft shadows and borderless cards.
v0.5.0 - The Structure Update
"Note: We skipped v0.4.0. In Chinese culture, '4' is considered unlucky. More importantly, this architectural leap from WASM to Native was so massive that v0.4.0 simply couldn't contain it."
🏎️ Native DuckDB Migration
We’ve swapped the heart of Wansan. By migrating from WASM to Native DuckDB, we’ve eliminated browser sandbox limits. Enjoy true desktop performance with faster I/O, better RAM management, and support for massive datasets that previously crashed the browser.
🏗️ Project Bundle Architecture (.wansan)
Your analysis is now a physical asset. Each project is a self-contained .wansan folder on your disk, containing an isolated native database and semantic context. Move it, back it up, or share it—your data is no longer trapped in a browser cache.
🧙♂️ Data Ingestion Wizard
Professional data handling is here. Our new multi-step wizard allows for Type Enforcement (preventing lost zeros in IDs), Smart Append with conflict resolution (Pro), and a 300% faster Excel parsing engine.
🧠 Project-Level Business Memory
Business logic is now decentralized. Define fiscal years or custom metrics within a specific project bundle. The AI now merges global preferences with project-specific context, ensuring zero context pollution across different workspaces.
- Swiss Style UI: Refined flat sidebar dock and high-contrast typography.
- Native Menu Bar: Full integration with macOS/Windows system menus.
- Relation Decentralization: Manage table links directly inside the Schema Editor.
v0.3.0 - The Analyst Update
"Performance is the foundation, and precision is the goal. We are transforming Wansan from a query tool into a high-performance virtual analyst."
🚀 Data Engine Re-architecture
We completely rewrote the ingestion pipeline. Our new **Streaming Ingestion** (via ExcelJS) now supports million-row files with a minimal memory footprint. Combined with DuckDB's zero-copy disk reading, local data analysis is now possible at a massive scale.
🛡️ Smart Filters: Ask, Don't Guess
AI hallucination ends here. When the AI is unsure about specific values (like SKU or City names), it no longer guesses. It triggers a local **Smart Filter** powered by DuckDB's fuzzy search. This ensures 100% precision, zero latency, and absolute privacy—your data fingerprints never leave your device.
📐 Foundation for a Semantic Layer
Laying the bricks for future intelligence. Define complex metrics like "Gross Margin" using plain English. Wansan infers the logic and creates **AI-powered calculated columns**, building the domain context that will power our future semantic engine.
- Interactive Drill-down: Click chart elements to dive deeper into specific dimensions.
- Sandwich Defense: Hardened AI stability by decoupling user chat from execution protocols.
- Swiss Style UI: Standardized input fields and refined brand-colored focus rings.
v0.2.0 - The Workbench
From a simple chatbot to a professional workspace. We've introduced the Multi-Session Sidebar to keep your mental flow organized.
- Data Hot-Reload: Replace source files without breaking SQL logic.
- AI Web Export: Share interactive dashboards with a Cyberpunk vibe.
- White-Box SQL: Full transparency for every DuckDB query.
v0.1.0 - The Genesis
Initial Public Beta. Proving that high-performance analytics can happen locally on your desktop.