TL;DR: QueryVeil's AI now remembers your analysis patterns across conversations. It learns which queries you run frequently, what insights you've discovered, and how your data relates. All stored locally — nothing leaves your browser.
The problem with stateless AI
Every time you start a new conversation with ChatGPT, it forgets everything. You've analyzed the same sales dataset 20 times, but each session starts from scratch: "What columns does this table have?" "What does mrr stand for?"
This is especially painful for data work, where context compounds. A good analyst doesn't just know the schema — they know that region_code maps to five regions, that Q4 always has a spike, and that the status column has a typo in 3% of rows.
How workspace memory works
QueryVeil now maintains a workspace memory — a persistent context layer that accumulates knowledge as you work.
What it remembers
Frequent query patterns: When you ask "Show me monthly revenue by region" and the AI generates a SQL query that works, that pattern is recorded. The next time you ask a similar question, the AI has a reference point.
Cross-conversation insights: Interesting findings from previous analyses are stored as insights. If you discovered that 15% of your orders have null shipping dates, that context carries forward.
Schema familiarity: The AI builds up knowledge of which columns are useful, which ones are categorical vs. numeric, and how tables relate.
Where it's stored
Everything is in IndexedDB — your browser's local database. Workspace memory is keyed to your workspace, so different projects have separate contexts. Nothing syncs to any server.
How it's used
The workspace memory is injected into the AI's system prompt. When you ask a question, the AI sees:
- Your table schema (column names and types)
- Semantic metadata you've added (descriptions, relationships)
- Workspace memory (frequent patterns + accumulated insights)
This means the AI's responses improve over time. It stops asking you to clarify things it's seen before.
Workspaces: isolated analysis environments
Workspace memory is part of a bigger feature — workspaces. Each workspace is an isolated environment with its own:
- Loaded data tables
- Chat conversations
- Semantic layer metadata
- Query history
- Bookmarks
- AI memory
You can switch between workspaces instantly. Each one has its own IndexedDB store, so nothing bleeds between projects.
This is designed for consultants and analysts who work on multiple clients or datasets. Your client A analysis is completely separate from client B — different data, different conversations, different AI context.
The resume flow
When you return to QueryVeil with existing data and past conversations, a welcome back banner appears with context: "3 tables loaded, 7 past conversations." One click resumes your last conversation.
This sounds simple, but it matters for workflow. Data analysis is rarely a single session. You explore on Monday, come back Tuesday with a new question, and continue Thursday with updated data. The app should feel like picking up a notebook you left on your desk.
Privacy by architecture
The interesting thing about all of this is that it's more private than cloud-based memory features (like ChatGPT's memory). Those systems store your preferences on their servers. QueryVeil stores everything in your browser.
If you clear your browser data, the memory is gone. If you export your workspace as a backup, the memory comes with it. You own the data — including the AI's knowledge about your data.
Workspace memory is available now. Just use QueryVeil normally — it learns as you go.
Read more about privacy data analytics and how QueryVeil keeps your data local.