Skip to Content
AgentOverview

Agent

The Agent (formerly Copilot) is Vendo’s command center — the first item in the sidebar. It connects to your BigQuery datasets and your full Vendo workspace, helping you explore data, run queries, build Data Studio, monitor sources, and manage integrations through natural language conversation.


What It Does

CapabilityDescription
Explore DatasetsBrowse your BigQuery datasets, list tables, inspect schemas, and preview sample data
Run QueriesExecute SQL queries and see results in interactive, sortable tables
Build TransformationsGenerate SQL, Python, and notebook transformations with AI assistance, preview results, and save to your pipeline
Dashboard CanvasRender multi-chart dashboards, data tables, and analysis on the right panel
Session HistoryResume previous conversations from the sidebar
Monitor PipelineCheck source status, view recent jobs, find errors, and inspect integrations
Take ActionTrigger syncs, pause or resume sources — all through conversation
Analyze PerformanceAsk questions about your data in plain English and get answers grounded in your workspace

How It Works

The Agent runs as a split-screen interface:

  • Left panel — Chat conversation where you describe what you need
  • Right panel — Dynamic canvas that displays query results, model configurations, or an empty state

When you ask a question, the Agent autonomously:

  1. Determines whether you need data analysis, model building, or pipeline operations
  2. Uses the appropriate tools (BigQuery for data, the Vendo platform for pipeline status)
  3. Writes and executes queries, generates models, or performs actions
  4. Displays results in the right panel or summarizes in chat

Available Tools

The Agent has access to a large toolset spanning several categories (the exact set grows over time):

Data & Warehouse

  • Browse datasets, list tables, inspect schemas, and preview sample data
  • runQuery — Execute read-only SQL queries with results in sortable tables
  • generateChart / generateDashboard — Visualize results as single charts or multi-chart dashboards
  • presentAnalysis — Render a concise analysis summary in the right panel

Transformations

  • generateModelConfig — Generate a Python model configuration
  • runPreview — Test a model against a live 100-row sample before saving

Platform (Read)

  • List sources, integrations, app connections, models, metrics, and recent jobs with status and error details

Platform (Actions)

  • Trigger source syncs, pause or resume sources and integrations, create or update models and metrics

Assistant

  • Memory — Save and search durable facts about your business and preferences
  • Scheduled analyses — Create, list, and disable recurring analyses that run on a cron schedule
  • Plans & navigation — Show a step-by-step checklist for multi-step tasks and offer in-app navigation links

Access Controls

  • The Agent can only access datasets belonging to your account
  • Only SELECT queries are allowed — no data mutations via SQL
  • Platform actions (pause, resume, sync) affect only your account’s resources
  • All interactions are authenticated and scoped to your account membership
  • Tool calls are limited to 16 steps per interaction

AI Models

The Agent supports multiple AI models. GPT-5.4 Mini is the default.

ModelProviderNotes
GPT-5.4 MiniOpenAIDefault — balanced speed and quality
GPT-5.5OpenAIComplex reasoning
GPT-5.4 NanoOpenAIFastest, most affordable OpenAI option
Gemini 2.5 FlashGoogleAvailable on Vendo’s platform key — no setup needed
Claude Opus 4.7AnthropicDeepest reasoning
Claude Sonnet 4.6AnthropicNuanced analysis
Claude Haiku 4.5AnthropicFast, affordable
OpenRouter modelsOpenRouterClaude Sonnet 4.5, Llama 3.3 70B, DeepSeek V3, Grok 4, plus several free-tier models

Gemini 2.5 Flash is available by default using the platform key. OpenAI and Anthropic models require your own API key (BYOK — Bring Your Own Key) in Settings > AI Models. OpenRouter models work with your own key or via Vendo’s platform key with a small markup. Account operators can also disable individual models.

Select your preferred model from the dropdown in the Agent header. Token usage is tracked per model and visible in Usage Metrics.


Next Steps

Last updated on