Indexing Overview

​

Introduction

Indexing turns your external content (repos, wikis, issues, files, designs, tests) into searchable knowledge that Agents and Pipelines, as well as LLMs from Conversations can use. Instead of manually browsing systems, create indexes once and then search or ask questions with natural language. The Indexes section provides a comprehensive, user-friendly way to create, manage, and search indexes directly within Toolkit Configuration, featuring visual status indicators, real-time progress monitoring, and dedicated index pages.

​

Purpose of Indexing

• Centralize knowledge from multiple tools into a consistent, searchable store.
• Improve retrieval quality for Agents and LLMs with chunking and metadata.
• Keep results fresh by re-running indexing when content changes.
• Replace legacy datasets with standardized tools across many toolkits.

​

Indexing Tools

ELITEA provides six standardized tools available across supported toolkits:

• Index Data — create or update an index for a chosen scope.
• Search Index — run a search query against existing indexes.
• Stepback Search Index — advanced search using stepback context for improved relevance.
• Stepback Summary Index — stepback search with on-the-fly summary generation.
• Remove Index — delete an existing index.
• List Indexes — list available indexes for the toolkit.

​

Toolkits That Support Indexes

The following toolkits support indexing:

​

Toolkit Configuration for Indexing

Before creating indexes, the following must be in place at the project and toolkit level:

• Credentials — a valid credential assigned to the toolkit for the target system (not required for Artifact).
• Vector Storage (PgVector) — configured at project level to store indexed data.
• Embedding Model — selected at project level to generate vector embeddings.
• Index Data tool turned on — required for the Indexes section to appear on the toolkit.
• Toolkit-specific fields — source parameters such as repository, branch, space key, or project ID depending on the toolkit type.

For the full setup walkthrough including prerequisites and toolkit configuration, see Create & Use Indexes. For source-specific configuration, use the relevant guide:

​

Indexes Features

The Indexes section in Toolkit Configuration is the primary interface for all index management tasks. The capabilities listed here are available from the index pages within each supported toolkit.

​

Create Index

A new index collects content from the connected source, chunks and embeds it, and stores the result in the vector database. Configure the index name, chunking strategy, and toolkit-specific parameters (branch, filters, space keys, and so on) at creation time. After the run starts, the index page tracks progress in real time. When complete, the index enters Completed or Partially Indexed state and becomes available for search, scheduling, reindexing, and history review. For step-by-step instructions and index status details, see Create & Use Indexes.

​

Search

Completed and Partially Indexed indexes support three search tools accessible directly from the index page:

• Search Index — direct semantic search; returns matching chunks ranked by similarity.
• Stepback Search Index — rewrites or simplifies the query before searching; improves retrieval for complex or multi-part questions.
• Stepback Summary Index — performs a stepback search and generates a summarized answer from the retrieved chunks.

For full parameter reference and usage details, see Indexing Tools.

​

Reindex

Reindexing refreshes an existing index with the latest source content using the stored configuration. Files whose content or source timestamp has changed are updated; unchanged files are not duplicated. Before confirming a reindex, you can adjust the configuration fields (all except the index name). For details on reindexing behavior and skipped-file categories, see Create & Use Indexes – Reindex.

​

Schedule

Each user can configure an independent automated reindexing schedule per index using a cron expression. Schedules can be turned on or off at any time without losing the configured expression. The minimum supported schedule frequency is once per day. For cron expression reference, schedule management, troubleshooting, and best practices, see Schedule Indexing.

​

Index History

The index page records all past indexing operations in chronological order — initial indexing, manual reindexes, scheduled reindexes, and any failures or stops. The history includes the event type, timestamp, indexed and skipped file counts. For details on history entries and available columns, see Create & Use Indexes – History.

​

Delete Index

An index and all its associated data can be permanently deleted from the index page. Deletion requires confirmation by re-entering the index name. The Delete action is only available when the Remove Index tool is turned on in the toolkit. For the delete workflow, see Create & Use Indexes.

​

Trigger From Chat

Indexing and search can also be triggered directly from Chat by referencing an Agent or Toolkit that has the relevant tools turned on. This is useful for one-off operations without going to the Toolkit configuration. For general conversation setup, see Chat. For the available tool parameters, see Indexing Tools.