albert-ying/agentic-cortex/dendron-query

Dendron Query

Query the Dendron knowledge base using indexes and SQLite for efficient information retrieval.

When to Use

Invoke when the user wants to find information in their vault: people, project status, scientific topics, recent activity, keyword search, or meeting preparation.

What It Does

1. Parses natural language query to determine intent
2. Routes to the appropriate entry point (by-person, by-project, by-topic, by-date, keyword)
3. Traverses note graph to collect related notes
4. Reads relevant notes in context
5. Synthesizes a structured response with source links

Usage

@dendron-query [natural language query]

Examples:

• @dendron-query Find everything about Sarah Kim
• @dendron-query What's the status of the protein design project?
• @dendron-query Show all papers on aging clocks
• @dendron-query What did I discuss in meetings last week?

Query Types & Routing

Person Query -> _index.by-person.md

Indicators: "about [Person]", "related to [Person]", "with [Person]", "who is [Person]"

1. Read _index.by-person.md -> find person's section
2. Follow links to projects, meetings, papers, contact info
3. Read their user.* note for full context
4. Check recent mentions via SQLite:

``sql SELECT fname, title FROM Note n JOIN Vault v ON n.vaultId=v.id WHERE v.fsPath='.' AND raw LIKE '%person-name%' ORDER BY updated DESC LIMIT 10; ``

1. Synthesize overview with links to all related notes

Project Query -> _index.by-project.md

Indicators: "project status", "about [Project]", "progress", "what's happening with"

1. Read _index.by-project.md -> find project section
2. Read main project file: proj.YYYY.project-name.md
3. Extract people, science, and meetings from frontmatter links
4. Check recent mentions via SQLite
5. Return status + context + next steps

Topic Query -> _index.by-topic.md

Indicators: "papers on", "research about", "all notes on", "everything about [concept]"

1. Read _index.by-topic.md -> find topic section
2. Collect all linked notes (projects, science, papers)
3. Fall back to ripgrep content search if index is insufficient
4. Organize by type and return structured summary

Date/Time Query -> SQLite

Indicators: "last week", "recent", "today", "yesterday", "meetings in January"

SELECT fname, title, date(updated/1000,'unixepoch') as modified
FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.'
AND (fname LIKE 'daily.journal.%' OR fname LIKE 'meet.%')
AND date(n.updated/1000,'unixepoch') >= date('now','-7 day')
ORDER BY fname DESC;

Keyword Search -> ripgrep + SQLite

Indicators: "find notes with", "search for", "mentions of", or any unmatched query

SELECT fname, title FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.' AND raw LIKE '%keyword%'
ORDER BY updated DESC LIMIT 20;

Supplement with ripgrep when SQLite raw column is incomplete or for multi-word phrases.

Common SQL Templates

-- Recent activity (last 7 days)
SELECT fname, title, date(updated/1000,'unixepoch') as modified
FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.' AND date(updated/1000,'unixepoch') >= date('now','-7 day')
ORDER BY updated DESC;
 
-- All meetings mentioning a person
SELECT fname, title FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.' AND fname LIKE 'meet.%' AND raw LIKE '%person-name%'
ORDER BY fname DESC;
 
-- Papers on a topic
SELECT fname, title, desc FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.' AND fname LIKE 'user.%'
AND (raw LIKE '%topic%' OR tags LIKE '%topic%');
 
-- Active projects
SELECT fname, title FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.' AND fname LIKE 'proj.%' AND raw LIKE '%status%active%';
 
-- All people (excluding paper references)
SELECT fname, title FROM Note n JOIN Vault v ON n.vaultId=v.id
WHERE v.fsPath='.' AND fname LIKE 'user.%' AND fname NOT LIKE 'user.%[0-9][0-9][0-9][0-9]'
ORDER BY fname;

Response Format

Structure every response with:

1. Summary — 2-3 sentences, most important information first
2. Details — Organized by category or chronology
3. Sources — Dendron links: [[title|note.id]]
4. Next Steps — Follow-up queries or actions (when applicable)

Example:

## Summary
Sarah Kim is the PI of the aging biology lab at Westfield University.
She advises two active projects and was last seen in the March 10 lab meeting.
 
## Details
- **Profile**: Professor, Dept. of Biology, Westfield University
- **Projects**: [[Protein Binder Design|proj.2026.protein-design]], [[Foundation Model|proj.2026.foundation-model]]
- **Recent**: [[Lab meeting Mar 10|meet.2026.03.10]] — binder screening results
 
## Sources
[[user.sarah-kim]], [[proj.2026.protein-design]], [[proj.2026.foundation-model]], [[meet.2026.03.10]]
 
## Next Steps
- `@dendron-query meetings with Sarah Kim in 2026`
- `@dendron-query protein design project status`

Advanced Features

• Multi-Entity: Find connections between two people or topics (shared meetings, overlapping projects)
• Temporal: Compare project state across time by reading notes from different date ranges
• Aggregation: List all active projects with summaries, or all people in a given project
• Meeting Prep: Combine person + project queries to build context before an upcoming meeting