Index ADO Wiki Data

Overview

ADO Wiki indexing turns your Azure DevOps wiki content into a searchable knowledge base inside ELITEA. You can index wiki pages—including project wikis and code wikis—with their hierarchical path structure, optionally describe embedded images with an LLM and optionally index referenced attachments as their own searchable documents.

Knowledge Search

Find documentation, procedures and technical guidance across wiki pages using natural-language queries.

Context-Aware Chat

Get AI-generated answers from your wiki content with citations to specific pages.

Faster Onboarding

Help new team members learn processes and standards from existing wiki documentation.

Documentation Audits

Identify gaps, overlap and outdated information in your wiki documentation.

Prerequisites

Before indexing ADO Wiki data, ensure you have:

ADO Credential: An Azure DevOps personal access token configured as an ADO credential in ELITEA
Vector Storage: PgVector selected in Settings → AI Providers
Embedding Model: Selected in Settings → AI Providers
ADO Wiki Toolkit: Configured with your Azure DevOps project details and credential, with the Index Data tool enabled

The Indexes section requires:

PgVector and an Embedding Model configured at the project level
The Index Data tool enabled in your toolkit configuration

Complete both project-level setup and toolkit configuration to access indexing functionality.

Required Permissions

Your ADO personal access token must have

Wiki → Read scope to index wiki content. Use Wiki → Read & write instead if the toolkit also modifies pages.
Access to the specific project and wikis you want to index.

Set Up the ADO Wiki Toolkit

Generate a Personal Access Token in your Azure DevOps account (User Settings → Personal Access Tokens → + New Token) with Wiki scope
Create a Credential in ELITEA: Go to Credentials → + Create → Ado, enter the Organization Url and Token, then save
Create the Toolkit: Go to Toolkits & Indexes → + Create → Azure Wiki (ADO Wiki)
Enable Required Tools: Select Index Data, List Indexes, Search Index, Stepback Search Index, Stepback Summary Index and Remove Index

The Index Data tool must be enabled for the Indexes section to appear on the toolkit detail page.

Tool Overview

Tool	Purpose
Index Data	Creates or refreshes a searchable index from ADO wiki pages.
List Indexes	Lists the indexes available for this toolkit.
Search Index	Performs semantic search across indexed content.
Stepback Search Index	Rewrites or simplifies the query before searching, using conversation context.
Stepback Summary Index	Performs a stepback search and generates a summarized answer.
Remove Index	Deletes an existing index.

For complete credential and toolkit setup details, see:

The following sections describe how to create, verify, reindex, schedule and delete an ADO Wiki index.

Access the Interface

Go to Toolkits & Indexes in the main navigation.
Select your configured ADO Wiki toolkit to open its detail page.
The toolkit detail page shows a two-panel layout: the left panel has the toolkit configuration form, and the right panel has the Indexes section listing indexes for this toolkit.

If the Index Data tool is not enabled, the Indexes section displays “Indexing is not available for now”. Enable Index Data in the toolkit configuration before using indexing features.

Create a New Index

In the Indexes section, select + Index.
A dedicated Create Index page opens with the Index Data configuration form.
Enter the required and optional values for your ADO Wiki index.

ADO Wiki Index Parameters

Parameter	Required	Default	Validation	Description
Index Name	Yes	None	1-32 characters	Unique name for the index (used as the collection identifier). Read-only once the index is created.
Clean Index	No	`FALSE`	boolean	Rebuild the index by re-processing every page from the source instead of skipping unchanged ones. Previously indexed pages the source no longer returns may remain searchable.
Progress Step	No	`10`	0-100	Step size for progress reporting during indexing.
Chunking Tool	No	`markdown`	`markdown` or empty	Chunking method applied to page content.
Chunking Config	No	`{}`	JSON object	Configuration for the chunking tool.
Wiki Identifier	No	None	string	Wiki to index, for example `ProjectAlpha.wiki`. If omitted, the toolkit’s Default Wiki Identifier is used; when a default is configured, this field’s helper text shows it (`Default wiki: <value>`). If neither is provided, indexing fails with an error. Applied the same way for manual index creation, manual reindexing and scheduled reindexing.
Path Contains	No	None	string	Case-insensitive substring filter matched against the full wiki page path. Hyphens and spaces are treated as equivalent (`Feature-Analysis` matches `Feature Analysis`), and matching a parent folder name also includes its descendant pages. Leave empty to index all pages.
Process Images	No	`FALSE`	boolean	Replace inline image markdown in page content with LLM-generated image descriptions so image content becomes searchable text. Only affects images embedded in page content, not attachments.
Image Description Prompt	No	None	string	Custom prompt used when generating image descriptions (shared by Process Images and Include Attachments).
Include Attachments	No	`FALSE`	boolean	Index files referenced under `/.attachments/` in page content as their own dependent documents. Include/Skip Extensions (below) filter only these attachments—wiki pages themselves are never filtered by extension.
Include Extensions	No	`[]`	array of extensions	Attachment file extensions to include. If empty, all attachments are processed except those in Skip Extensions.
Skip Extensions	No	`[]`	array of extensions	Attachment file extensions to exclude.
Workers	No	`1` (serial)	1-10	Maximum number of attachments/images processed concurrently.

Toolkit-Specific SettingsIndex parameters vary by toolkit type. For a full comparison across all supported toolkits, see Indexing Tools.

The Index button remains disabled until all required fields are valid.

Select Index to start indexing.
The page navigates to the dedicated index detail page, which opens on the Activity tab and shows live progress updates.
When indexing completes, the index enters Completed or Partially Indexed state and becomes available for search, scheduling, reindexing and history review.

Verify Index Creation

Once indexing completes:

The Configuration tab shows the saved index parameters (all editable except Index Name).
The General section shows the created date and the Files indexed / Files skipped counts.
The Indexes section shows an index card with the index name, created date, file counts and status icon. Hover over a card to reveal quick actions: Open, Reindex, Delete.

Reindex an Index

Use Reindex on the index detail page to refresh the index with the latest wiki content.

Open the Configuration tab and adjust any fields if needed (all fields are editable except Index Name).
Select Reindex (shown as Save & Reindex if changes are pending) and confirm.

ADO Wiki Reindex Behavior

Pages are deduplicated by hashing the raw page markdown, so re-running indexing does not re-embed unchanged pages—including when only Process Images output would differ, since that hash is computed before image descriptions are generated.
If you change Include Attachments, Include Extensions or Skip Extensions between runs, affected pages are automatically reprocessed even though their page content hash is unchanged, because the toolkit also compares the expected set of attachment documents for each page.
If an attachment fails to download after retries, its parent page is automatically marked for reprocessing on the next reindex, so the failed attachment is retried rather than silently skipped forever.
Pages removed from the source wiki are not automatically removed from the index during reindexing.

Schedule Reindexing

Use the Schedule section on the index detail page to automate reindexing on a cron schedule. Scheduling is available only for indexes in Completed or Partially Indexed state.

Scheduled reindexing applies the Wiki Identifier fallback the same way as a manual reindex: if the field is left empty, each scheduled run resolves it from the toolkit’s current Default Wiki Identifier at run time, with no need to specify it in the schedule itself. If you update the Default Wiki Identifier on the toolkit later, subsequent scheduled runs pick up the new value automatically.

For cron expression syntax, credential selection and schedule management, see Schedule Indexing.

Delete an Index

Open the index, select Delete, and enter the index name to confirm. Deleting an index permanently removes its indexed data, run history, and schedule configuration. The Delete action is only available when the Remove Index tool is enabled for the toolkit.

Use Search Tools

Once your ADO Wiki index is Completed or Partially Indexed, open its index detail page and select Search to run any of the search tools enabled for the toolkit:

Tool	Description
Search Index	Basic semantic search across indexed content.
Stepback Search Index	Advanced search that rewrites the query using conversation context.
Stepback Summary Index	Stepback search with an AI-generated summary answer.

For the full parameter reference for each search tool, see Indexing Tools.

Use Indexed Data in Conversations & Agents

Your ADO Wiki toolkit can be used in two main contexts:

In Conversations: Add the toolkit as a participant to ask questions and search your indexed wiki data
In Agents: Include the toolkit when creating AI agents to give them access to your documentation

Real-Life Example: Index Development Team Documentation Scenario: You have a development team wiki in Azure DevOps containing project documentation, API guides and troubleshooting information. You want to make this documentation searchable for your team. Step 1: Create the index

Navigate to Toolkits & Indexes → select your ADO Wiki toolkit
In the Indexes section, select + Index
Configure parameters:
- Index Name: docs
- Wiki Identifier: WebApp.wiki (or leave empty to use the toolkit’s default wiki)
- Clean Index: ✓ (checked to rebuild the index from the source)
- Path Contains: (leave empty to index all pages)
- Chunking Tool: markdown
Select Index and monitor progress on the Activity tab
Wait for Completed status

Step 2: Add the toolkit to a conversation

Navigate to Chat → + Conversation
Select the + icon in the chat input toolbar
Hover over the Toolkits category and enable your ADO Wiki toolkit

Step 3: Ask questions

User Request: “Find all documentation related to microservices deployment and provide a summary of the deployment process from our wiki.”
ADO Wiki Toolkit: “Based on your indexed wiki pages, here is a comprehensive summary of microservices deployment: Deployment Documentation Found:
1. CI/CD Pipeline Setup: Complete guide for automated deployment workflows
2. Environment Configuration: Development, staging and production environment setup
3. Container Deployment: Docker containerization and Kubernetes deployment procedures
Deployment Process Summary:
- Pre-Deployment: Code review, automated testing, security scanning
- Development: Automatic deployment on code commit
- Staging: Manual approval with integration testing
- Production: Blue-green deployment with rollback capabilities
[Citations: Deployment Guide v2.1, Container Best Practices, Production Runbook]”

Advanced search example:

User: “What are the authentication mechanisms used across our microservices?”
ADO Wiki Toolkit: “According to the indexed documentation, your microservices use a multi-layered authentication approach:
1. JWT Tokens: Primary authentication for API calls
2. OAuth 2.0: Third-party service integration
3. Service-to-Service: Mutual TLS authentication
[Citations: Authentication Architecture, API Security Guide, Service Mesh Configuration]“

Troubleshooting

Indexes Section Not Visible

Verify PgVector and an Embedding Model are configured in Settings → AI Providers.
Ensure the Index Data tool is enabled in your ADO Wiki toolkit configuration.
Refresh the page and try again.

Wiki Identifier Not Provided Error

Problem: Indexing fails with an error such as “Wiki identifier must be provided either as a parameter or configured as default in toolkit settings.”Cause: The Wiki Identifier field was left empty and no Default Wiki Identifier is set in the toolkit configuration.Resolution:

Set the Wiki Identifier field on the index, or
Open the toolkit configuration and set Default Wiki Identifier to your wiki name (for example, ProjectAlpha.wiki)

Connection or Authentication Failures

Verify the ADO personal access token has Wiki: Read (or Read & write) scope and has not expired.
Check the Project field and the credential’s Organization Url for typos.
Confirm you have access to the specified project and wiki.

Index Creation Failed or Stuck In Progress

Verify the wiki exists and contains pages matching your Path Contains filter (if set).
Check for very large pages or a large number of attachments that might extend processing time.
Verify PgVector and the embedding model are reachable.

Attachments Not Searchable

Verify Include Attachments was enabled when the index was created or last reindexed.
Check Include Extensions / Skip Extensions—an attachment excluded by either filter is not indexed.
Attachment downloads are retried automatically; a persistently failing attachment marks its parent page for reprocessing on the next reindex rather than failing silently.

No Search Results Returned

Verify the index status is Completed or Partially Indexed.
Lower the Cut Off score (for example, from 0.5 to 0.35).
Increase Search Top.
Try Stepback Search Index for complex, multi-part questions.
Confirm the Index Name used in the search tool matches the index you created.

Vector Database or Embedding Model Errors

Ensure PgVector is configured correctly in Settings → AI Providers.
Verify an embedding model is selected and available.
Confirm the vector database is reachable.

Related Documentation

Indexing Overview—Complete guide to indexing in ELITEA and its capabilities
Indexing Tools—Detailed reference for all indexing tools and parameters
Create & Use Indexes—Full walkthrough of the Indexes interface, reindexing, history and search

Chats

Indexing

Agents, Pipelines & Skills

Credentials & Toolkits

Entity Management