Talk to your Data
The AI Chat (powered by Gemini) is the fastest way to build your dashboard. Instead of drag-and-dropping nodes manually, you can just ask for the insights you need.
The AI will interpret this query, select the Channel Subscribers Temporal Node, set the time range to 90 days, enable cumulative tracking, and automatically create the nodes on your canvas.
@ Video Mentions
You can quickly isolate a specific video by typing @ followed by its name (e.g.,@My Video Title). This allows the AI to target this video for analysis.
Example Queries to Try:
- Performance: "How is my watch time trending this month?"
- Audience: "Show me a breakdown of where my viewers are from."
- Discovery: "What are the top search terms for my channel?"
- Comparison: "Show me subs gained vs lost for the last year."
- Intelligent Analysis: "Why did my views drop on my latest video?" or "What should I focus on for my next upload?"
Intelligence Features
Incremental Updates: Tubient is state-aware. If you ask to "change the date to 1 year" or "add a watch hours node", the AI refines your existing canvas and links the new nodes automatically rather than starting over.
Context Caching: We cache your large analytics instructions in the browser, providing faster responses and 90% lower usage costs.
Pointer System: To ensure high performance, the AI uses an on-demand data retrieval system. It only fetches heavy metric arrays when you ask specific analytical questions, keeping the orchestration layer light and fast.
Strict Output Modes: The AI adheres to operational guardrails: it only modifies the graph when action verbs are used, and focuses on text analysis when you ask questions about your existing data.
Intelligent Analysis Mode: Triggered by interrogatives (Who, What, When, Where, Why, How) or analytical phrases like "Tell me about." This mode prioritizes deep narrative analysis.
- Queries with time-based terms like "over time", "trend", or "growth" default to Temporal Nodes.
- Queries starting with "Compare", "Rank", or "Which" default to Comparison Nodes.
- Queries asking for counts or calculations (e.g., "total views", "how many", "average") default to Numeric Nodes.
- Queries naming a breakdown dimension (e.g., "traffic sources", "by country", "devices") default to Numeric Nodes.
- Queries targeting advanced metrics (e.g., "retention curve", "CTR", "outliers") default to Premium Nodes.
Answers Stay Tied to Their Dates
Every answer that quotes figures shows the dates they cover, for example Measured 19 Aug to 17 Sep. If you later change the date range, earlier answers are marked Superseded rather than left looking current, and the assistant treats their figures as history in follow-ups. See Visualizing Data for the details, including what +1 window means.
Freshness-Aware Date Ranges
The assistant knows two dates when it builds your canvas: today's date, which it uses for wording like "this month", and the latest day YouTube has data for. YouTube Analytics typically runs 2-3 days behind, so a request like "since 27 August" is measured up to that latest available day rather than up to today.
This keeps windows the length you asked for and prevents the assistant from presenting the most recent days as complete when YouTube has not reported them yet. If the latest data date can't be retrieved, the assistant still builds your canvas; it just works without that hint.
Canvas-Aware Help
You can ask the AI questions about the system itself and it will tailor the answer to what's on your canvas:
- Describe something broken — "Why isn't my comparison chart showing anything?" gets a setup-specific response based on what nodes you currently have.
- Ask what's possible — "What can I do next?" surfaces suggestions based on your current pipeline, not a generic feature list.
- Learn a concept — "What does a filter node do?" explains it in context of what you've already built.