Collections & Connectors

What's a collection?

A collection is a named group of indexed documents from one source. indexed index create … creates a collection. Each has:

• A name (for example eng-tickets, team-wiki, project-docs)
• A connector type (files, jira, or confluence)
• A source (folder path, Jira base URL, or Confluence URL — depending on the connector)
• On-disk metadata plus a FAISS vector index derived from chunked text

Collections are independent: you create, update, search, and remove them separately. With no collection filter, search can run across all of them. How parsing, chunking, and embedding fit together is summarized in Indexing overview.

What's stored

Each collection directory is ~/.indexed/data/collections/<name>/ by default, or .indexed/data/collections/<name>/ in the current directory when you use --local on create (see Index commands). The four files match the configuration guide:

Connectors

A connector reads and normalizes documents from a source. Indexed ships three:

Files

Indexes files under a local path. Parsing uses the stack described under parsing architecture: rich documents via Docling, code via tree-sitter where applicable.

indexed index create files -c my-docs -p ./docs

The files connector supports incremental updates — see Keeping the index fresh.

Best for: project docs, READMEs, exports, codebases.

Jira

Fetches issues with JQL and indexes summary, description, and comments.

indexed index create jira -c tickets -u https://company.atlassian.net -q "project = ENG"

Best for: tickets, bug history, decision context.

Confluence

Fetches pages with CQL and indexes page bodies and comments.

indexed index create confluence -c wiki -u https://company.atlassian.net/wiki -q "space = ENG"

Best for: wikis, runbooks, onboarding.

Managing collections

See Index commands for flags (including --local and --simple-output).

What's next