Skip to main content

Documentation Index

Fetch the complete documentation index at: https://basedash.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

AI context allows you to enhance AI understanding by providing additional information about your organization, data sources, schemas, tables, and columns. This helps the AI generate more accurate and relevant responses when creating charts, automations, or answering questions in chat.

​
Automatic AI context

Basedash automatically provides the AI with comprehensive context about your data sources, including:
  • Data source structure: Table names, column names, and data types from all connected databases and warehouses
  • Table and column descriptions: Any existing comments or descriptions from your source databases (PostgreSQL and Snowflake comments, BigQuery descriptions)
  • Semantic layer context: If you use a semantic layer like dbt, Basedash includes any documentation, descriptions, and metadata defined in your dbt models
This automatic context ensures the AI has a foundational understanding of your data structure without requiring any manual setup.

​
How AI interprets your data

Basedash performs automatic pre-processing when you connect your data sources to understand the structure of your schemas, tables, and columns. The AI can analyze data types, relationships, and basic metadata to generate queries and visualizations. However, custom context provides the business intelligence that goes beyond the technical structure. It helps the AI understand:
  • How your organization operates and what metrics matter most
  • What your data actually represents in business terms
  • Internal terminology and KPIs specific to your company
  • The context and meaning behind complex data structures
This additional layer of understanding enables the AI to generate more relevant, business-focused insights rather than just technical queries.

​
How it works

Custom context is automatically considered by the AI whenever you create charts, chat, or automations in Basedash. Global context applies to every conversation within your organization. You can also add group context to control how the AI works for different teams. Data source, schema, table, and column-level context is used when that specific data is referenced by the AI.

​
Accessing custom context

​
Global context

Global context applies to all AI interactions in the organization. You can access global context through the organization dropdown in the top-left corner. Organization dropdown with global context option

​
Skills

Skills are reusable bundles of instructions that you write once and every Basedash AI surface can read on demand. Unlike global context, which is always loaded in full by every agent, skills sit in a catalogue: each agent sees the skill names available and loads the full instructions only when a skill is relevant to the current request. Skills are picked up across chat, chart generation, dashboards, automations, and insights, so you can teach Basedash about a concept once instead of repeating the same context in every prompt.

​
When to use skills

Skills are best for modular playbooks focused on a single concept. Common patterns:
  • Metric definitions — Define how the AI should calculate activation rate, MRR, churn, retention, or any KPI specific to your business
  • Team conventions — Encode the rules a specific team follows (finance uses GAAP revenue; growth uses calendar weeks in UTC)
  • Analytical playbooks — Describe how to write A/B reports, triage support tickets, or summarize cohorts
  • Chart preferences — Specify preferred chart types and breakdowns for a particular topic
Each skill can be up to 50,000 characters, so even nuanced playbooks fit. Skills are the natural home for the kind of guidance you’d give a new analyst joining your team.

​
Managing skills

Skills are managed by organization admins. To create or edit skills:
  1. Open Settings → Organization → AI context
  2. In the Skills section, click New skill (or click an existing skill to edit it)
  3. Give the skill a clear, descriptive name (for example, “Activation rate” or “Support triage”)
  4. In the Instructions field, describe in plain language how the AI should use this skill
  5. Click Save changes
Only admins can create, edit, or delete skills, so the shared context stays consistent for everyone in the organization.

​
Skills vs. global context

Global contextSkills
LoadingAlways loaded by every agentCatalogue always; instructions loaded on demand
ShapeOne block of textMany named bundles, one concept each
Best forShort, durable org-wide factsModular playbooks, metric definitions
Edited byAdmins (and AI in some surfaces)Admins only
For most teams, both work well together: global context for the handful of facts that apply to every conversation, and skills for the richer concepts you want the AI to apply consistently when they come up.

​
Auditing which skills the AI used

When an agent loads a skill, you can see it in the thinking trace. Look for the Reading <skill name> skill step — clicking it expands to show the skill’s full instructions. This makes it easy to verify which skills shaped any given answer.

​
Group context

Group context applies to users in a specific group. This is useful when different teams need the AI to behave differently, such as:
  • Finance: Define strict metric definitions and preferred accounting terminology
  • Support: Explain ticket lifecycle, common issue categories, and internal abbreviations
  • Sales: Define pipeline stages and how to interpret deal status fields
To edit group context:
  1. Open the command menu
  2. Go to Groups
  3. Select a group
  4. Choose Edit AI context

​
Context in chat

You can also add context directly from the chat interface. Look for the “Global context” button (asterisk icon) in the top-left of the chat input. Chat interface with add context button

​
Data source context

You can also add context tied to a specific data source, schema, table, or column.
  1. Navigate to the Data page
  2. Right-click on a data source, schema, table, or column
  3. Select “Edit AI context”
Data page with edit AI context option

​
Best practices

​
Start with global context

We recommend beginning with global context as one of your first setup steps. This provides the AI with fundamental understanding of your business, terminology, and key metrics.

​
Define internal terminology

Global context is a great place to define internal terminology. This helps the AI understand your business, terminology, and key metrics. Use custom context to define:
  • Internal KPIs: Specific metrics that only your organization understands
  • User terminology: How you refer to different user types (e.g., “signups,” “active users,” “premium customers”)
  • Business jargon: Company-specific terms and definitions
  • Metric definitions: Custom calculations or business logic for specific metrics
Once defined, the AI will understand and use your terminology consistently across all interactions.

​
Getting started

We recommend adding custom context as part of your initial Basedash setup. This ensures the AI has the right context from the beginning, leading to more accurate and relevant responses. For step-by-step guidance, see our getting started guide which includes custom context setup as part of the recommended workflow.

​
Examples

​
Organization context

Our company is a SaaS platform for e-commerce businesses.
We refer to our customers as "merchants" and their customers as "shoppers."
"MRR" refers to Monthly Recurring Revenue from subscription plans.
"Churn" means when a merchant cancels their subscription.

​
Schema context

The "analytics" schema contains event tracking data from our platform.
The "billing" schema contains subscription and payment information.
The "support" schema contains customer support ticket data.

​
Group context

This group is the finance team.

Use conservative language in summaries.
If there is ambiguity in a metric, ask a clarifying question instead of assuming.

"Revenue" means GAAP revenue (not bookings).
"Churn" should be computed as logo churn unless explicitly asked for revenue churn.

​
Skill: activation rate

Activation = signed up AND completed onboarding within 7 days.
Exclude trial-only accounts and internal email domains.
Use calendar weeks in UTC; weeks start on Monday.
Always compare cohorts by signup_month, not by event_month.
Prefer a line chart for the trend over time.

​
Skill: support triage

When summarizing support tickets:
- Group tickets by category and priority before summarizing.
- Tickets open longer than 48 hours are SLA risk; call them out.
- Use human-readable status names ("Open", "Waiting on customer"), not raw enum values.
- Roll up volume by week unless the question specifies a different window.

​
Column context

The "metadata" JSON column contains user preferences and settings.
The "status" column uses values: "active", "suspended", "cancelled".
The "created_at" timestamp is in UTC timezone.