Playground User Guide

Playground User Guide

The Playground is Simba Intelligence’s interactive interface for querying your data using natural language. Instead of writing SQL queries, you simply ask questions in plain English and receive instant answers with explanations, visualizations, and raw data.

Overview

What is the Playground?

Think of the Playground as having a conversation with an expert data analyst who has immediate access to all your databases. You ask questions like “What were our top-selling products last quarter?” and get back not just the data, but explanations of what the data means and how it was found.

Key Capabilities

🗣️ Natural Language Querying

• Ask questions in plain English about your data
• No SQL knowledge required
• Context-aware follow-up questions

⚡ Real-Time Results

• Streaming responses as queries execute
• Live progress updates during processing
• Immediate feedback on query success or issues

� Conversational Interface

• Chat-like experience that maintains context
• Ask follow-up questions naturally
• Multi-turn conversations about your data
• Context preserved while browser session is active

📊 Data-Backed Answers

• Natural language explanations of your insights
• Access to underlying data used for each answer
• Interactive controls for feedback

🎯 Smart Suggestions

• AI-generated question suggestions based on your data
• Context-aware recommendations
• Learn what’s possible with your data sources

---

Prerequisites

Before using the Playground, ensure you have:

1. Data Sources Available

• At least one data source configured and accessible
• Created via Data Agent or manual configuration
• Proper permissions to query the data

2. AI Provider Configuration

• LLM provider configured (Google Vertex AI, OpenAI, AWS Bedrock)
• Sufficient API quota/credits for query processing
• Network connectivity to AI service endpoints

3. User Permissions

• Basic user access to Simba Intelligence (all users can access Playground)
• Query permissions for your target data sources
• Network access to the Playground interface

💡 Getting Started: If you don’t have data sources yet, visit the Data Agent to create one using AI assistance.

---

Accessing the Playground

Navigation

1. Log into Simba Intelligence
2. Click “Playground” in the main navigation menu
3. Or visit directly: http://your-domain/playground

Initial Setup

When you first access the Playground:

• The system loads all available data sources you can query
• AI provider connectivity is verified
• Query suggestions are generated for available data

---

Selecting Data Sources

Data Source Dropdown

The Playground header contains a data source selector that shows:

• All data sources you have permission to query
• Source descriptions and metadata
• Connection status indicators

Choosing the Right Data Source

For specific datasets:

• Select the data source that contains the information you need
• Each data source typically represents a specific business area or database

For exploring:

• Start with data sources that have rich suggestions
• Look for sources with clear, descriptive names
• Try sources created recently or specifically for your use case

URL State Management

The Playground maintains your data source selection in the URL:

• http://your-domain/playground?sourceId=12345
• Bookmarkable URLs for specific data source contexts
• Shareable links with colleagues (permissions permitting)

---

Writing Effective Natural Language Queries

Query Input Interface

Chat Input Features:

• Conversational interface that maintains context throughout your session
• Auto-expanding text area adapts to message length
• Enter key submission (Shift+Enter for new lines)
• Follow-up questions that reference previous conversation context
• Session-based history (context preserved while browser is open)

Note: Chat sessions are not saved or shareable. Your conversation context is maintained during your active browser session but will reset when you close or refresh the page.

Query Types and Examples

Basic Data Exploration

What data is available in this source?
Show me the first 10 rows of data
What are the column names and types?
How many records are in this dataset?

Analytical Questions

What are our top 5 best-selling products this year?
Show me sales trends by month for the last 6 months
Which regions have the highest customer satisfaction scores?
Compare revenue between Q3 and Q4 of this year

Comparative Analysis

How does this year's performance compare to last year?
What's the difference in sales between our top and bottom performing stores?
Which product categories grew the most year-over-year?
Compare customer acquisition costs across marketing channels

Filtered and Conditional Queries

Show me all customers with orders over $1000 in the last 30 days
What are the sales figures for the electronics category in California?
Find all orders that were delivered late in Q4
Show me employees hired after January 2023 with salaries above $75k

Writing Tips for Better Results

✅ Do:

• Be specific about what you want to see
• Include time ranges when relevant (“last quarter”, “this year”)
• Mention specific values or thresholds (“over $1000”, “top 10”)
• Use business terminology familiar to your domain

❌ Avoid:

• Overly complex questions combining multiple unrelated analyses
• Ambiguous time references (“recently”, “a while ago”)
• Technical database terminology unless necessary
• Questions that require data not in the selected source

🔍 Example Progression:

1. Start broad: “What sales data do we have?”
2. Get specific: “Show me monthly sales totals for 2024”
3. Dive deeper: “Which months had sales below the average?”
4. Analyze patterns: “What factors contributed to low sales in those months?”

---

Understanding Responses

The Playground provides conversational responses with full context awareness and data transparency:

Chat-Based Responses

What you receive:

• AI-generated explanations in natural language
• Key insights and patterns identified in your data
• Business context and interpretation
• Formatted text with emphasis on important findings
• Ability to ask follow-up questions that reference previous context

Conversational flow:

The chat interface maintains context throughout your session, enabling natural follow-up questions:

You: "What were our top-selling products last quarter?"

Assistant: Based on your sales data, here are the top performers...
[Product A led with $2.4M revenue (32% of total)...]

You: "How does Product A compare to last year?"

Assistant: Product A showed strong year-over-year growth...
[Uses context from previous question about Product A]

You: "What factors contributed to that growth?"

Assistant: Analyzing the factors behind Product A's growth...
[Continues building on the conversation context]

Data-Backed Answers

Interactive data controls:

When the assistant provides an answer based on actual data from your sources, you’ll see a small menu with these options:

• 👍 Upvote: Mark the answer as helpful and accurate
• 👎 Downvote: Indicate the answer was incorrect or unhelpful
• 📋 Copy: Copy the response text to your clipboard
• 📊 View Data: See the underlying data that was used to generate the answer

Accessing backing data:

Click the “View Data” option to see:

• Raw query results used to formulate the answer
• Structured data in JSON format
• Actual values and calculations referenced in the response
• Verification of the assistant’s interpretation
• Data can be copied and pasted for use in other tools

When to check backing data:

• Validating AI interpretations against raw data
• Verifying specific numbers or calculations
• Understanding data quality or completeness
• Copying data for further analysis in other tools

---

Query Suggestions

Smart Recommendations

The Playground generates relevant question suggestions based on:

• Data source content: Fields, relationships, and data patterns
• Common analysis patterns: Typical business questions for your data type
• Usage context: Questions similar to what others ask about this data
• AI insights: Interesting patterns the AI identifies in your data

When Suggestions Appear

Initial load:

• When you first open the Playground, suggested questions are displayed
• Click any suggestion to immediately start a conversation
• Suggestions are tailored to your selected data source
• Questions are organized by relevance and complexity

During conversation:

• After you start chatting, suggestions are no longer automatically displayed
• You can request suggestions at any time by asking: “What questions can I ask?” or “Give me some suggested questions”
• The assistant will provide relevant suggestions based on your data source and current conversation context

Using Suggestions Effectively

Suggestion types:

📊 Exploratory Questions:

• “What data is available in this source?”
• “Show me a sample of the data”
• “What are the key metrics I can analyze?”

📈 Analytical Questions:

• “What are the trends over time?”
• “Which categories perform best?”
• “How does performance vary by region?”

🔍 Specific Deep Dives:

• Custom questions based on your specific data schema
• Business-relevant queries for your industry
• Complex analytical patterns found in your data

Learning from Suggestions

Use suggestions to:

• Discover capabilities: Learn what questions are possible
• Improve query writing: See examples of effective natural language queries
• Find patterns: Identify analytical approaches you hadn’t considered
• Get unstuck: Find inspiration when you’re not sure what to ask

---

Interactive Features

Conversational Context

Natural follow-up questions:

The chat interface maintains full context during your browser session:

• Reference previous questions and answers naturally
• Drill down deeper without repeating context
• Ask clarifying questions about earlier results
• Build complex analysis through conversation

Example conversation flow:

"What were our top 5 products last quarter?"
→ "Show me more details about the first one"
→ "How does it compare to the same quarter last year?"
→ "What regions performed best for this product?"

Session limitations:

• Context is maintained while your browser session is active
• Closing or refreshing the page resets the conversation
• Chat history cannot be saved or shared

Rating and Feedback

Per-answer interactions:

For any response that includes data from your sources, you’ll see interactive controls:

👍 Upvote: Mark the answer as helpful and accurate

• Helps improve AI performance for similar questions
• Indicates successful query interpretation
• Signals useful analytical approaches

👎 Downvote: Indicate issues with the response

• Results don’t match expectations
• AI misinterpreted your question
• Data appears incorrect or incomplete
• Query was too slow or failed

📋 Copy: Copy the response text

• Share insights with colleagues
• Document findings in reports
• Save important analysis results

📊 View Data: Access underlying data

• See raw data used for the answer
• Verify calculations and interpretations

Working with Backing Data

Data access:

Access backing data through the “View Data” option on any data-backed answer:

• View structured data in JSON format
• Copy and paste data for use in other tools
• Verify and validate AI-generated insights
• Save important query results before session ends by copying the data

---

Advanced Usage Patterns

Complex Query Construction

Breaking down complex questions:

Instead of: “Show me a comprehensive analysis of customer behavior patterns including demographics, purchase history, satisfaction scores, and predictive lifetime value across different segments for the past two years with seasonal adjustments”

Try this approach:

1. “What customer data do we have available?”
2. “Show me customer demographics breakdown”
3. “How do purchase patterns vary by demographic group?”
4. “What are the satisfaction scores for each customer segment?”
5. “How has customer behavior changed over the past two years?”

Working with Multiple Data Sources

Source switching workflow:

1. Start with one data source for core analysis
2. Note insights that might benefit from additional data
3. Switch to complementary data source
4. Cross-reference findings and build comprehensive picture

---

Troubleshooting Common Issues

”No Data Sources Available”

Problem: Dropdown shows no available data sources Solutions:

1. Check that you have appropriate permissions
2. Verify data sources exist and are properly configured
3. Contact administrator if you should have access
4. Create new data sources using the Data Agent

”LLM Configuration Required”

Problem: Queries fail due to missing AI provider setup Solutions:

1. Navigate to LLM Configuration page
2. Configure at least one AI provider (requires admin access)
3. Contact administrator if you don’t have configuration permissions
4. Wait for configuration to be completed by your team

Query Execution Failures

Common causes and solutions:

“Query timeout”:

• Simplify your question to focus on specific data
• Try breaking complex questions into smaller parts
• Contact administrator about database performance

“No results found”:

• Verify your filters and conditions are appropriate
• Check date ranges and spelling of specific values
• Try broader queries to explore available data

“Permission denied”:

• Confirm you have access to the selected data source
• Check with administrator about query permissions
• Verify data source is properly configured

Slow Query Performance

Optimization strategies:

1. Be specific: Use date ranges and filters to limit data scope
2. Focus queries: Ask about specific metrics rather than “everything”
3. Use appropriate data sources: Choose sources optimized for your question type
4. Time of day: Consider database load during peak business hours

Unexpected Results

When results don’t match expectations:

1. Verify data source: Ensure you’re querying the right dataset
2. Clarify your question: Rephrase with more specific terms
3. Use follow-up questions: Ask the AI to explain its interpretation
4. Provide feedback: Use rating buttons to improve future results

---

Best Practices

Query Writing Guidelines

Start simple and iterate:

Step 1: "Show me sales data"
Step 2: "Show me monthly sales totals for 2024"  
Step 3: "Which months had the highest sales?"
Step 4: "What products drove those high sales months?"

Use business terminology:

• Reference metrics as they’re known in your organization
• Use familiar date ranges (“last quarter”, “fiscal year 2024”)
• Include relevant business context in questions

Be explicit about what you want:

• “Show me the top 10…” instead of “Show me the best…”
• “Sales figures for Q4” instead of “recent sales”
• “Revenue by product category” instead of “product performance”

Effective Data Exploration

Understanding your data:

1. Start with exploratory questions to understand data structure
2. Identify key dimensions and metrics available
3. Explore data quality and completeness
4. Build complexity gradually based on initial findings

Validating results:

• Cross-reference AI summaries with raw data in Data tab
• Spot-check calculations and trends
• Use domain knowledge to validate insights
• Question results that seem too good (or bad) to be true

Collaboration and Sharing

Sharing insights:

• Copy responses to share findings with colleagues
• Copy backing data for use in presentations or reports
• Document successful query patterns for team reuse
• Use bookmarkable URLs to share specific data source contexts
• Note: Chat sessions themselves cannot be saved or shared

Building team knowledge:

• Document query examples that work well for your data
• Share data source recommendations with colleagues
• Rate queries to improve AI performance for your team
• Collaborate on complex analysis by sharing intermediate results

---

Ready to start exploring your data with natural language? Head to the Playground and ask your first question. Remember: there’s no such thing as a bad question when you’re learning about your data!