Playground
What is Playground?
Playground is a specialized agent designed for users to interact with their data stack using natural language. It is an easy-to-use solution for all your questions across Snowflake, DBT, Tableau, Databricks, Airflow, GitHub and more.
Key Capabilities
Playground empowers users to:
- Analyze performance of existing workloads across your data infrastructure
- dbt Project analysis of entire dbt project structure, models, tests, dependencies etc.
- Lineage exploration to explore the lineage of dbt models and answer related questions
- Documentation search allows to search through dbt models documentation as well as add missing documentation
- Perform root cause analysis to quickly identify and resolve issues
- Conduct cross-tool analysis spanning multiple platforms in a single conversation
- Generate detailed optimization plans with quantified outcomes and projected savings
- Best practices review to get recommendations on dbt models best practices implementation
Why Do We Need Playground?
Traditional AI approaches have critical limitations. These limitations are well taken care of through Playground:
Without Playground
- User asks: "Why is my dbt model running slow?"
- Agent searches basic data
- Finds execution time metrics
- Provides incomplete analysis with:
- Missing: Root cause of performance issue
- Missing: Downstream impact on dependent models
- Missing: Historical performance trends
Result: Incomplete, Siloed AI Responses
With Playground
- User asks: "Why is my dbt model running slow?"
- Multi-agentic framework assembles rich context
- Provides comprehensive analysis:
- Root cause identified (e.g., inefficient CTEs, missing indexes)
- Downstream impact on dependent models assessed
- Actionable optimization recommendations with code examples
Result: Context-Rich, Memory-Enhanced Intelligence
Extending Context
Beyond the integrations already connected to your SaaS instance, Playground allows you to enrich your queries with additional context:
| Option | Description |
|---|---|
| Datamates | Select any Datamate to widen the context based on tools |
| Knowledge Bases | Connect organizational knowledge repositories from Knowledge Hub for deeper context |
| File Attachments | Upload documents, queries, or data files directly into your conversation |
This flexibility ensures you can ask questions across a broader tool set while providing the agent with the context it needs to deliver accurate, actionable insights.
How Playground Works
1. Providing Input
When starting a conversation, users have multiple ways to provide context:
| Option | Description |
|---|---|
| Ask a question | Type your query directly in the chat input |
| Add Datamate | Select a Datamate to access any tool specific information |
| Attach files | Upload files for analysis (via "+" menu) |
| Knowledge Base | Access organizational knowledge repositories from Knowledge Hub (via "+" menu) |
| View Prompt Library | Browse pre-built prompts for common tasks (via 'Select Prompt Library') |
2. Results Delivery
Once the search is complete, you receive a comprehensive analysis with:
- Executive Summary - High-level findings
- Key Metrics - Data tables with critical information
- Detailed Analysis - In-depth analysis (e.g., query-by-query breakdown)
- Key Findings - Insights with evidence and impact
- Root Cause Analysis - Why issues occurred
- Optimization Recommendations - Prioritized action items
- Next Steps - Implementation roadmap
3. Output Features
Analysis results include interactive options:
| Feature | Description |
|---|---|
| View detailed steps | Expand to see the full execution trace |
| Task Progress | Visual checklist of completed subtasks |
| Download | Export results for offline use or reporting |
4. Confidence Indicators
Playground provides transparency about result quality:
| Indicator | Meaning |
|---|---|
| Low Confidence (50%) | Consider refining your query for better results |
| High Confidence | Results are reliable and well-supported |
Components of Playground
Two important components of Playground are Chat History and Prompt Library:
Chat History
Chat History is a persistent record of all your previous conversations with Playground. Located in the left sidebar, it provides quick access to past interactions, allowing you to revisit analyses, continue previous work, or reference earlier insights.
Key Features
| Feature | Description |
|---|---|
| Conversation List | All past chats are displayed chronologically with preview titles |
| Status Indicators | Each conversation shows its current status (e.g., "Completed") |
| Timestamps | Relative timestamps (e.g., "8h ago", "11h ago") help you locate recent work |
| Search | Quickly find specific conversations using the search bar |
| New Chat | Start a fresh conversation at any time with the "+ New Chat" button |
How It Works
- Automatic Saving - Every conversation is automatically saved to your Chat History
- Title Generation - Conversations are titled based on your initial query (e.g., "Help me create a Jira ticket with the below details...", "Analyse the query")
- Quick Resume - Click any conversation to instantly resume where you left off
Use Cases
- Continuing Analysis - Pick up where you left off on a complex investigation
- Reference Past Insights - Look back at previous cost analyses or query optimizations
- Audit Trail - Track what questions you've asked and what answers you received
- Knowledge Building - Build on previous conversations rather than starting from scratch
Prompt Library
Prompt Library is a shared repository of pre-built, reusable prompts that help you get started quickly with common data analysis tasks. It serves as a knowledge base of effective queries created by you, your team and the organization.
Key Features
| Feature | Description |
|---|---|
| Pre-built Prompts | Ready-to-use prompts for common analytical tasks |
| Tagging System | Organize prompts with tags like "query", "summary", "test" for easy discovery |
| Ownership Tracking | See who created each prompt (individual or "Altimate AI") |
| Team Sharing | Share prompts with "All org" or specific teams |
| Search & Filter | Find prompts by type, team, owner, or tags |
| Custom Prompts | Create and save your own prompts with "+ New Saved Prompt" |
How to Use
- Access the Library - Click "View Prompt Library" from the Suggested Prompts link on the main Playground screen
- Browse or Search - Use filters (Type, Teams, Owner, Tags) or search to find relevant prompts
- Select a Prompt - Click the send icon to use a prompt directly
- Customize - Many prompts have placeholders (e.g.,
<enter no. of days>) that you can fill in - Save Your Own - Click "+ New Saved Prompt" to add your custom prompts to the library
Filters Available
| Filter | Purpose |
|---|---|
| Type | Filter by prompt category |
| Teams | Show prompts shared with specific teams |
| Owner | Filter by prompt creator |
| Tags | Filter by semantic tags (query, summary, test, etc.) |
| Sort | Order by "Recently Added" or other criteria |
Benefits
- Faster Start - Don't write prompts from scratch; use proven templates
- Best Practices - Leverage prompts crafted by experts
- Consistency - Teams use standardized prompts for common analyses
- Knowledge Sharing - Share effective prompts across the organization or with specific teams
- Parameterized Templates - Prompts with placeholders adapt to different scenarios
Playground is currently in Beta. We're continuously improving the platform based on user feedback.
For more information and access to Playground, visit app.myaltimate.com