Skip to main content

Create Your First Index

Once you've installed TeaRAGs and connected to an agent, indexing a codebase is one command.

Index Your Codebase​

If you installed the Claude Code skills plugin (tea-rags@tea-rags), open Claude Code in your project directory and invoke the indexing skill:

/tea-rags:index

First run does a full index. Every subsequent call does an incremental reindex — only changed files are re-processed.

Without the skills plugin (Cursor, Roo Code, etc.)

Ask your agent directly — it will call the index_codebase MCP tool:

Index this codebase with tea-rags

What Happens During Indexing​

  1. File discovery — scans the project, respects .gitignore and .contextignore
  2. AST-aware chunking — parses code into semantic chunks (functions, classes, methods) via tree-sitter
  3. Trajectory enrichment — attaches git + static signals per chunk (age, churn, bug-fix rate, imports, …)
  4. Embedding — converts chunks to vectors using your configured provider
  5. Storage — upserts vectors + payload into Qdrant
  6. Snapshot — saves file hashes for future incremental updates

Update After Changes​

Re-run the same skill after making changes or pulling new commits — it detects modified / added / deleted files automatically:

/tea-rags:index

Only changed files are re-processed; unchanged files are skipped.

Full Re-Index (zero downtime)​

For large rewrites, branch switches, or when payload schema changes, use:

/tea-rags:force-reindex

A new versioned collection is built in the background while search keeps serving the current one. The alias switches atomically when the new build finishes. Requires explicit user confirmation — never invoked automatically.

Check Index Status​

Show me stats for the current index

tip

You can check index status from a parallel agent tab while indexing is still running.