DAGA

Digital Agency Growth Academy

AI operator training, hosted products, and community

TextPreview access

Install qmd and connect your vault

Install qmd on the VPS, index your vault, and run your first search queries.

Time 15 minModule qmd retrievalCourse progress 0%

Lesson outcome

You will have qmd installed on the VPS with your vault indexed, and be able to search for notes using keyword and semantic queries.

Why this matters in an agency

When your vault is small, you can find things by browsing. When it grows past 50 notes, browsing breaks down. You forget what you named things. You do not remember which folder holds the pricing logic versus the service descriptions. qmd solves this by letting you search in three ways: exact keywords (like searching email), meaning-based queries (like asking a question), and hypothetical answers (describing what the answer should look like). Together, these find the right note even when you do not remember the exact title or wording.

Inputs, tools, and prerequisites

Claude Code running on the VPS. Your vault at /root/vault/ with 20+ notes from Module 10.

Step-by-step walkthrough

Install qmd

SSH into the VPS and start Claude Code:

```
ssh vps
claude
```

Ask Claude Code to install qmd:

```
Install qmd on this server. It is a local markdown search engine. Check if it is available via npm or pip, install it, and verify it works.
```

Claude Code will find the installation method and run it. After installation, verify:

```
qmd --version
```

Or whatever the version check command is. The tool should be installed and accessible.

Index your vault

qmd needs to build an index of your vault before it can search. Ask Claude Code:

```
Configure qmd to index the vault at /root/vault/. Run the initial indexing.
```

Claude Code will run the appropriate qmd command to scan and index all .md files in the vault. This creates embeddings (mathematical representations of your note content) that enable semantic search. The initial indexing takes a few seconds for a small vault.

Run your first keyword search

Keyword search (lexical search) finds exact words and phrases. This is the fastest search type — use it when you know what words appear in the note.

Ask Claude Code:

```
Use qmd to search for "onboarding" in my vault. Use keyword/lexical search.
```

This should return notes that contain the word "onboarding" — your Client Onboarding SOP, possibly other notes that mention it.

Run a semantic search

Semantic search finds notes based on meaning, not exact words. This is useful when you know the concept but not the exact terminology used in the note.

```
Use qmd to do a semantic search for "how we set pricing for our services." Intent: understand the logic behind service pricing decisions.
```

This should return your Pricing Logic note, even though the search query does not use the exact same words as the note title.

Run a hypothetical-answer search

The most powerful search type: describe what the answer should look like, and qmd finds notes whose content matches that description.

```
Use qmd to search using a hypothetical answer: "A note that describes the step-by-step process for bringing on a new client, including contract signing, tool access, kickoff call, and first deliverable setup."
```

This should return your Client Onboarding SOP or the onboarding checklist from Module 13, because the content of those notes matches the description you provided.

Compare the three search types

Try the same question with all three search types:

```
Search my vault for information about "client reporting" using all three search modes: keyword, semantic, and hypothetical. Compare the results.
```

Notice what each type finds:

  • Keyword finds notes containing the exact phrase "client reporting"
  • Semantic finds notes about the concept of client reporting, even if they use different words like "performance reports" or "monthly updates"
  • Hypothetical finds notes whose content matches a description of what client reporting documentation looks like

Each type has strengths. Use keyword for known terms, semantic for concepts, and hypothetical when you know what the answer looks like but not where it lives.

Failure modes and verification checks

The main failure is searching with queries that are too vague. "Find my stuff" will not return useful results. Be specific about what you are looking for, even if you do not know the exact note.

Another failure is not re-indexing after adding new notes. If you add notes to the vault and qmd does not find them, run the indexing command again.

Verification: run a search for a note you know exists. Does qmd find it? Then run a search using different words for the same concept. Does semantic search find it? If both work, qmd is properly configured.

Implementation checklist

  • Install qmd on the VPS.
  • Index the vault.
  • Run a keyword search and verify results.
  • Run a semantic search and verify results.
  • Run a hypothetical-answer search and verify results.
  • Compare results across all three search types.

Immediate next action

Move to the next lesson and practice using qmd for real business questions.

Exercise

Think of three real questions about your business that you should be able to answer from your vault:

  1. A factual question (keyword search): "What tools do we use?"
  2. A conceptual question (semantic search): "How do we handle client complaints?"
  3. A descriptive question (hypothetical): "Where is the note that explains our competitive advantage?"

Run all three searches. For each one, verify that qmd returns the right note. If it does not, the issue is either the search query (refine it) or the note quality (improve the title and opening paragraph). Fix the root cause.