Search Indexes

Configure search indexes that become agent tools for knowledge retrieval.

Overview

Search Indexes connect your Azure AI Search indexes to ExecAssist, making them available as tools that agents can use. When you configure a search index, ExecAssist automatically creates a corresponding search tool that agents can call to find relevant information.

How Search Indexes Work

From Index to Tool

Azure AI Search Index
    ↓
Configure in ExecAssist
    ↓
Search Tool Created Automatically
    ↓
Assign Tool to Agents
    ↓
Agent Searches When Needed

Example Flow

1. You have a "Product Documentation" index in Azure AI Search
2. Configure it in ExecAssist as "product_docs"
3. ExecAssist creates search_product_docs tool
4. Assign the tool to your Support Agent
5. When customers ask product questions, the agent searches the docs

Viewing Search Indexes

Navigate to Capabilities > Tool Management > Search Tools to see configured indexes.

Index List

Column	Description
Name	Your configured name
Index	Azure AI Search index name
Provider	Which search provider
Tool Name	Generated tool name
Status	Active or inactive

Adding a Search Index

Prerequisites

• Search provider configured
• Azure AI Search index exists with content (or you'll create one)
• Understanding of what the index will contain

Quick Start Templates

The wizard offers templates to help you get started quickly:

Template	Use Case	Index Fields
Email Knowledge Base	Scanned email content	Subject, Sender, Summary, Keywords, Body
Custom	Build your own schema	You define the fields

Three-Step Wizard

Adding a search index uses a guided wizard with three steps:

Step 1: Basic Information

Field	Description	Example
Provider	Select search provider	"Production Search"
Index Name	Azure index name	"email-knowledge"
Display Name	Human-readable label	"Email Knowledge Base"

Step 2: Search Configuration

Configure how the index will work:

Query Settings:

Setting	Description	Recommendation
Query Type	Simple, full, or semantic	"semantic" for best results
Search Mode	Any or All terms match	"any" for broader results
Result Limit	Max results per query	3-5 results
Semantic Config	Named configuration in Azure	Required for semantic search

Schema Settings:

Define the index schema by specifying fields and their types:

• Use the template to pre-populate fields
• View sample schemas for reference
• Add, remove, or modify fields as needed

Field Configuration:

Field Property	Description
Name	Field identifier
Type	String, Int32, Double, Boolean, etc.
Searchable	Include in full-text search
Filterable	Allow filtering
Sortable	Allow sorting
Facetable	Enable faceting
Key	Primary key field

Tip: Click Copy Schema to copy a JSON schema you can use when creating the index in Azure Portal.

After configuring the schema, click Verify Index to confirm the index exists in Azure with the expected configuration.

Step 3: Tool Configuration

Configure the search tool that agents will use:

Field	Description	Example
Tool Name	Identifier for the tool	"search_email_knowledge"
Description	What the tool searches	"Search scanned email content for relevant information"
Active	Enable the tool	Yes

Index Verification

Between steps 2 and 3, the wizard verifies your index:

• Confirms the index exists in Azure
• Validates the schema matches your configuration
• Reports any discrepancies

If verification fails:

1. Create the index in Azure Portal first
2. Copy the schema from the wizard
3. Return and verify again

Semantic Search Settings

If your Azure AI Search tier supports semantic search:

Setting	Description
Enable Semantic	Use AI-enhanced search
Semantic Config	Named configuration in Azure
Query Type	Set to "semantic"

Note: Semantic search requires Standard tier (S1) or higher in Azure AI Search.

Managing Indexes

Editing an Index

1. Click on the index
2. Update configuration
3. Save changes

Note: Changing the tool name affects which agents have access.

Deactivating an Index

1. Edit the index
2. Set status to Inactive
3. Save

Inactive indexes won't be searchable, but configuration is preserved.

Deleting an Index

1. Remove the tool from all agents first
2. Click delete icon
3. Confirm deletion

Tool Generation

Automatic Tool Creation

When you add a search index, ExecAssist:

1. Creates a tool with the specified name
2. Sets up the search parameters
3. Makes the tool available for assignment

Tool Naming

Tool names are generated from your configuration:

• Input: "Product Documentation"
• Output: search_product_documentation

Or use a custom tool name if specified.

Tool Description

The description helps agents understand when to use the tool:

Good descriptions:

• "Search product documentation for technical specifications, user guides, and troubleshooting"
• "Find HR policies including PTO, benefits, and company procedures"

Poor descriptions:

• "Search stuff"
• "Knowledge base"

Assigning Tools to Agents

Adding to an Agent

1. Edit the agent
2. In Tools section, select the search tool
3. Save the agent

Multiple Search Tools

Agents can have multiple search tools:

• search_product_docs - Product information
• search_hr_policies - HR questions
• search_technical_kb - Technical support

Guiding Tool Usage

Add instructions to help agents choose the right tool:

When searching for information:
- For product questions, use search_product_docs
- For HR or policy questions, use search_hr_policies
- For technical issues, use search_technical_kb

Search before responding to ensure accuracy.

Search Behavior

Query Formulation

Agents formulate queries based on:

• The original email content
• Their understanding of what's needed
• Previous search results (if refining)

Result Processing

When results return, agents:

1. Review the results
2. Extract relevant information
3. Synthesize into their response
4. Cite sources when appropriate

No Results Handling

Instruct agents on what to do when search returns empty:

If search returns no results:
1. Try rephrasing the query
2. If still no results, acknowledge the limitation
3. Escalate if the information should exist

Search Quality

Improving Relevance

If searches aren't returning good results:

1. Check index content - Is the needed information there?
2. Review field configuration - Are the right fields being searched?
3. Test queries - Try searches directly in Azure portal
4. Enable semantic search - For better understanding of intent

Common Issues

Issue	Cause	Solution
Irrelevant results	Wrong fields searched	Update search fields
Missing results	Content not indexed	Re-index content
Too many results	Query too broad	Reduce result limit
Slow searches	Large index, complex queries	Optimize index

Index Content Management

What to Index

Good candidates for search indexes:

• Product documentation
• Policy documents
• Knowledge base articles
• FAQ content
• Technical specifications

Content Updates

When source content changes:

1. Update the Azure AI Search index
2. Re-index if necessary
3. No changes needed in ExecAssist

Index Health

Monitor your indexes:

• Document count
• Last update time
• Query performance
• Storage usage

Best Practices

Index Design

1. Focused indexes - One topic per index
2. Good descriptions - Help agents understand what's there
3. Appropriate limits - Don't return too many results
4. Regular updates - Keep content fresh

Agent Instructions

1. Tell agents when to search
2. Explain what each index contains
3. Guide query formulation
4. Handle no-results gracefully

Monitoring

1. Review how agents use search tools
2. Check for repeated searches (might indicate poor results)
3. Monitor search latency
4. Track search costs

Troubleshooting

Tool Not Appearing

Symptoms: Created index but tool not available for agents

Solutions:

1. Verify index is active
2. Check tool name was generated
3. Refresh the agent tools list
4. Verify provider is healthy

Search Returning Errors

Symptoms: Agent gets errors when searching

Solutions:

1. Test search directly in Azure portal
2. Check provider connection
3. Verify index exists and has content
4. Review error details in execution logs

Poor Search Quality

Symptoms: Results don't match what's expected

Solutions:

1. Test queries in Azure Search Explorer
2. Review field mappings
3. Check if semantic search would help
4. Verify content is properly indexed

Slow Searches

Symptoms: Search tool takes too long

Solutions:

1. Check Azure service performance
2. Reduce result limit
3. Simplify search fields
4. Consider index optimization

Advanced Configuration

Custom Query Templates

For advanced scenarios, customize how queries are built:

• Add filters (e.g., by date or category)
• Boost certain fields
• Apply scoring profiles

Faceted Search

If your index supports facets:

• Filter by category
• Narrow by date range
• Focus on specific sources

Hybrid Search

Combine text and vector search:

• Better relevance for natural language queries
• Semantic understanding of intent
• Improved results for ambiguous queries

Related Topics

• Search Providers
• Tools
• Agents
• Reading a run