Analyze
AI chat
Chat with your datasets in plain English. Ask for totals, comparisons, outliers, trends — the AI reads your schema, records, and statistics to answer in real time.
Starting a conversation
Open any dataset and click the Chat tab. A conversation is scoped to that dataset — its schema, records, and computed statistics become the context for every message.
What you can ask
Anything that could be answered from the data. A few examples:
- “What was our total revenue last month?”
- “Show top 5 products by unit sales this quarter.”
- “Which invoices are more than 30 days overdue?”
- “Compare average order value between Lagos and Abuja stores.”
- “Find anomalies in the expense column — anything > 3 standard deviations.”
How answers are generated
The backend (backend/app/services/chat_service.py) composes a prompt from:
- The dataset schema (column names, types, and descriptions).
- Pre-computed statistics (counts, sums, aggregations).
- A sample of recent records.
- The rolling conversation history.
Responses stream back token-by-token so you see the answer forming in real time. For larger datasets, the AI will ask clarifying questions rather than guessing.
Multiple conversations
Each dataset can hold many conversations. Switch between them from the chat sidebar, archive ones you no longer need, or rename a conversation to track a specific thread of analysis.
Limitations
- The AI answers based on the current dataset state. If you change filters in the table, re-ask in chat to get fresh answers.
- Chat does not mutate your data — it cannot insert, update, or delete rows. Use the table for that.
- Very large datasets are summarized statistically rather than read row-by-row.
Privacy & AI data handling
Prompts and dataset excerpts are sent to our AI provider (OpenAI by default) via TLS. We have data-processing agreements that prohibit use of your content for model training. See Security for details.
Plan access
AI chat is included on Pro and Enterprise plans.