Lantide Data
This translation is available, but has not yet been checked against the latest Traditional Chinese source.

Quick Start

2. Quick Start

The walkthrough below uses Agent project analysis (order funnel example). For manual SQL only, see §2.8.

2.1 Launch and First Setup

Open Lantide Data; the backend starts automatically and you land on the main screen. On first launch a default workspace is created.

First-launch experience:

  1. The AI Chat panel opens automatically (later launches default to collapsed).
  2. A welcome dialog appears with a product intro video (placeholder for now); click Get Started to close.
  3. After closing the dialog, the WelcomePanel in AI Chat guides you to set an API Key—click Open AI Settings to open settings in one step.
  4. In settings, add or edit a Profile (OpenAI / Ollama / Custom), enter API Key and model, then save. Full steps: §12.3.
  5. (Optional) After LLM setup, fill in Agent Preferences (reply language, style, etc.) written to User Knowledge. See §15.3.1.

[Image] Initial screen after launch

2.2 Load Data

  1. Expand the Data section in the sidebar.
  2. If no folder is bound, you see No local folder — click to choose: click, confirm, then pick a folder in the file picker.
  3. After you choose a folder (e.g. containing orders.csv), Lantide Data scans supported files and lists them under Data.
  4. To change folders, right-click the folder name row → Change data folder. More options (Reveal in Finder, list sort): §3.2, §4.1.

[Image] Sidebar file list after choosing a folder

2.3 Create and Focus a Project

  1. In sidebar Projects, click ＋, enter a project name (e.g. "Order Funnel"), confirm.
  2. Right-click the project name → New File, create a Plan file.
  3. Click the eye icon on the project row to focus—the icon turns green and the row highlights. When focused, the Agent enters project analysis mode and no longer asks quick vs. project. See §11.1, §11.4.

[Image] New project created and focused

2.4 Describe the Analysis in Natural Language

In the AI panel, enter a request such as:

Analyze the order funnel and find the stage with the most drop-off. Denominator starts at "order created"; time range last 90 days.

The Agent may confirm goal, denominator, or time range (through interactive question blocks in chat). The AI status indicator shows Project Focused or Planning. The Agent explores the schema of tables such as orders.csv before writing the Plan—it does not guess column names.

For simple questions like "What columns does this table have?", the Agent answers directly without a full Plan flow.

2.5 Review the Plan: Metrics and Checkpoints

After drafting the Plan, the Agent asks you to review. A solid funnel Plan usually includes:

  • Analysis goal — e.g. find the stage with the most drop-off.
  • Stage definitions — the filter condition for each funnel stage (e.g. "paid", "shipped").
  • Denominator — how each stage's conversion rate is defined (previous-stage count vs. a fixed start).
  • Grain — per order vs. per user.
  • Validation checkpoints — e.g. confirm the max drop-off stage, check NULL or abnormal states.

To change a metric, in Plan Visual mode select text → click the annotation icon → write your change. The Agent patches only the annotated section, not the whole document. Annotations: §11.7.

Do not press Approve & Execute at this stage—first confirm the Plan's metrics and steps match your intent. The Agent will not pretend you have approved and start fetching data on its own.

2.6 Run the Analysis

When satisfied, click Approve & Execute on the Plan toolbar. The system switches the Plan to Executing and automatically sends a start-execution instruction to the Agent—you do not need to type "start execution" in the AI panel. This button approves the current Plan's lifecycle and execution scope; if you use an external Agent, it does not raise the connection's Observe / Execute / Admin access mode.

During execution you will see:

  1. The Agent reads the Plan, then asks for your preferred analysis execution mode (foreground auto analysis / hybrid). Modes: §12.5; you choose only once in the same conversation during Executing (separate from Quick Analysis background mode).
  2. The Progress button on the Plan toolbar shows a Todo list and query steps (a red dot appears when updated). See §11.6.
  3. Before each major SQL step the Agent validates syntax, then fetches data; exploratory trial and formal pulls are separated to avoid pointless retry loops.
  4. If SQL is too complex (e.g. 3+ CTEs), the Agent is required to split it into smaller steps so you can review incrementally.

On a dual or wide monitor, open a persistent SQL tab or the Plan in Compare view for reference while you keep editing or running the next step in the main window. See §11.10.

If execution is interrupted (user stop, network issue, app close, etc.), the Plan shows a Continue button; click it to resume from the interruption without starting over.

2.7 Read the Report and Next Steps

When analysis completes, the Agent adds a Report in the project. A solid funnel Report should include:

  • Stage definitions and denominator explanations
  • The max drop-off stage with concrete counts / rates (e.g. "Approved → handoff to logistics: 1,623 orders, 1.63%")
  • Limitations — questions the data cannot answer, metric assumptions, correlation vs. causation, etc.

You can keep annotating the Report to have the Agent revise wording, or use the toolbar HTML Report to export a standalone HTML file to share in a browser. See §10.

When accepting the Report, open the project's Plan in Compare view alongside the Report in the main window to check metrics and numbers (Compare view is read-only). See §11.10.

If the analysis naturally leads to a follow-up question (e.g. drilling down by product category), create a new Plan in the same project and keep iterating.

[Image] Completed Plan and Report side by side

2.8 Appendix: Manual SQL Quick Query

If you do not need Agent collaboration, here is the shortest manual query path:

  1. (Optional) Create a workspace — In the sidebar Workspace section, click ＋, enter a name (1–15 characters), confirm. See §3.1.
  2. Load data — Same as §2.2.
  3. New SQL tab — Click ＋ on the tab bar, enter SQL in the editor, e.g.:
 SELECT * FROM "orders.csv" LIMIT 10
  1. Run — Click Run or press Shift + Enter.
  2. View results — Results appear below the editor; you can paginate (20–100 rows per page), with an on-screen preview of up to 500 rows. Use export for the full dataset.

Tip: A table name is the file name, wrapped in double quotes. Double-click a file name in the sidebar to insert the table name automatically. More editor features: §5, §6.

[Image / GIF] Writing and running SQL in the editor

← Previous: Welcome to Lantide DataNext: Workspace Management →