---
name: Lumen Doctor
slug: lumen-doctor
category: AI Engineering
description: Lumen Doctor runs a health check on the bundled Lumen semantic search setup for the current project. It verifies backend reachability, checks index freshness, and summarizes storage and remediation steps.
github: "https://github.com/ory/lumen/tree/main/skills/doctor"
language: Go
stars: 254
forks: 31
install: "npx degit https://github.com/ory/lumen/tree/main/skills/doctor ~/.claude/skills/doctor"
installs_to: ~/.claude/skills/doctor
source_path: skills/doctor/SKILL.md
collection_size: 3
category_size: 2631
collection_url: "https://dirskills.com/collections/ory/lumen"
added: 2026-09-02T05:20:41.468Z
last_synced: 2026-09-02T05:20:41.468Z
canonical_url: "https://dirskills.com/skills/lumen-doctor"
---

# Lumen Doctor

Lumen Doctor runs a health check on the bundled Lumen semantic search setup for the current project. It verifies backend reachability, checks index freshness, and summarizes storage and remediation steps.

**Install:**

```bash
npx degit https://github.com/ory/lumen/tree/main/skills/doctor ~/.claude/skills/doctor
```

## README

# Lumen Doctor

Run a health check on the bundled Lumen semantic search setup for the current
project.

## Steps

1. Call the Lumen `health_check` tool to verify the embedding service is
   reachable.
2. Call the Lumen `index_status` tool with `path` or `cwd` set to the current
   working directory to check index freshness.
3. Report a concise summary:
   - Embedding service status, backend, host, and model
   - Index totals: files, chunks, last indexed time, stale or fresh
   - Storage totals: unique vectors, shared references, deduplication ratio,
     vector precision, database bytes, and reclaimable bytes
   - Any MCP or plugin setup issue that blocks the tools
4. If no index exists yet, explain that the Lumen `semantic_search` tool seeds
   the index on first use.
5. If the user wants eager indexing instead of waiting for the next search,
   suggest running `lumen index .` in the repository root.
6. Interpret storage totals as collection-wide values shared by compatible Git
   worktrees; file, chunk, freshness, and indexing values are project-local.
   A high deduplication ratio is expected and is not an index-health problem.