Find Hubble Query in Obsidian

Search saved queries in ~/Documents/stripe-obsidian/Queries/ using flexible two-stage search strategy.

Search Strategy

Use two-stage progressive search:

•Fast search (metadata only) - Check frontmatter fields (<1 second)
•Deep search (full content) - Search SQL and descriptions (2-5 seconds, only if needed)

Quick Start

Parse User Intent

Extract search criteria from natural language:

Keywords: Main concepts

•"smart retry experiments" → keywords: smart, retry, experiment

Table names: If mentioned explicitly

•"queries using fact.charges" → table: fact.charges
•"payment and dispute tables" → tables: [fact.payments, fact.disputes]

Project: If mentioned

•"fraud queries in risk project" → project: risk

Tags/Type: If mentioned

•"experiment queries" → tag: experiment OR query_type: experiment
•"debugging queries" → query_type: debugging

Author: If mentioned

•"queries by mingc" → author: mingc

Stage 1: Fast Metadata Search

Search YAML frontmatter only using Grep:

bash

grep -r "<keywords>" ~/Documents/stripe-obsidian/Queries/ \
  --include="*.md" \
  -A 20 | grep -E "(^---|title:|tags:|tables:|project:|query_type:)"

Search in:

•title: - Query title
•tags: - Tag array
•tables: - Input tables array
•project: - Project name
•query_type: - Query purpose

Relevance scoring:

•Exact title match: 10 points
•Tag match: 8 points
•Table match: 7 points
•Project match: 5 points (if specified)
•Keyword in title: 6 points

Show top 5-10 results with:

•Title
•Project
•Tags
•Tables used
•File path
•Created date and author

Stage 2: Deep Content Search

Trigger deep search when:

•Fast search returns < 3 results
•User says "search deeper" or "check SQL content"
•User says results don't match

Use Grep in full file content:

bash

grep -r "<keywords>" ~/Documents/stripe-obsidian/Queries/ \
  --include="*.md" \
  -B 5 -A 10

Search in:

•SQL query text (WHERE clauses, column names, JOIN conditions)
•Description section
•Notes section
•Comments in SQL

Also check:

•Similar operations (GROUP BY, aggregations)
•Related table patterns (fact., dim., etc.)

Search Scenarios

Scenario A: Natural Language (No Specifics)

Example: "find queries about payment success rates"

Actions:

•Extract keywords: payment, success, rate
•Fast search in title and tags
•If < 3 results, search table names containing "payment"
•If still insufficient, deep search in SQL for these terms
•Rank by relevance

Scenario B: Table-Specific

Example: "what queries use fact.charges?"

Actions:

•Grep for "fact.charges" in tables: array (frontmatter)
•List all matches grouped by project
•Show most recent 10

Scenario C: Project-Scoped

Example: "find smart retry queries in payments-intelligence"

Actions:

•Limit Grep to ~/Documents/stripe-obsidian/Queries/payments-intelligence/
•Search for "smart retry" OR "smart-retry" in title, tags, description
•Rank by relevance

Scenario D: Multi-Table JOIN

Example: "queries that join charges and disputes"

Actions:

•Grep for files with BOTH "fact.charges" AND "fact.disputes" in tables array
•These are likely JOIN queries
•Show SQL snippet of how they're joined

Scenario E: Similar Query Request

Example: "help me write a query to calculate auth rate, check if we have examples"

Actions:

•Extract intent: authorization rate calculation
•Keywords: auth, rate, authorization
•Fast search for these terms
•
Look for queries with:
- •"rate" in output columns
- •"auth" OR "authorization" in title/tags/SQL
- •Aggregation patterns (COUNT/SUM ratio)
•Show top matches with SQL

Scenario F: Vague Exploratory

Example: "what experiment queries do we have?"

Actions:

•Search for query_type: experiment OR tag experiment
•Group results by project
•Show distribution: "Found 12 experiment queries across 4 projects"
•List most recent 10

Result Format

Present results clearly:

code

Found 5 queries matching "smart retry":

1. **Smart Retry Model Score Distribution** (payments-intelligence)
   - Tables: fact.charges, exp.smart_retry_scores
   - Tags: experiment, smart-retry, ml-model, score-analysis
   - File: smart-retry-model-score-distribution.md
   - Created: 2026-02-09 by mingc

2. **Smart Retry Success Rate Analysis** (payments-intelligence)
   - Tables: fact.charges, exp.smart_retry_decisions
   - Tags: experiment, smart-retry, success-rate
   - File: smart-retry-success-rate-analysis.md
   - Created: 2026-01-15 by jane

[show top 5-10]

Would you like me to:
- Show full SQL for any query?
- Narrow down with filters?
- Search with different terms?

Follow-up Actions

After showing results:

Show full query: Read the complete markdown file

code

Read(file_path="~/Documents/stripe-obsidian/Queries/<project>/<filename>.md")

Compare queries: Read multiple files to show differences

Refine search: Ask for additional criteria

•"Which project should I focus on?"
•"Looking for recent queries only?"
•"Need queries with specific tables?"

Edge Cases

No Results

code

No queries found matching '<criteria>'.

Try:
- Broader search terms (e.g., "payment" instead of "payment_intent_id")
- Different project
- Deep search in SQL content

Too Many Results (>20)

code

Found 45 queries matching '<criteria>'. Showing top 10 most relevant.

To narrow down:
- Specify project: "in payments-intelligence"
- Add table filter: "using fact.charges"
- Add time filter: "from last month"
- Add query type: "experiment queries only"

Ambiguous Intent

Ask clarifying questions:

•"Looking for queries that use these tables, or queries ABOUT these tables?"
•"Want experiment-type queries, or queries about experiments?"
•"Recent queries, or best examples regardless of age?"

Performance Tips

Fast search targets:

•Should complete in < 1 second
•Grep frontmatter only (first ~20 lines)
•No need to read full files

Deep search considerations:

•Takes 2-5 seconds
•Greps entire file content
•Use only when needed

Result ranking:

•Show most relevant first
•Consider recency if relevance tied
•Highlight why each result matches

Integration

After finding query:

•Offer to show full SQL
•Suggest modifications for user's needs
•Reference related queries if found

If query doesn't exist:

•"No saved queries found. Would you like to save one after writing it?"
•Suggest creating new query or using Hubble directly

Example Execution

User: "help me write a query to check model scores, do we have examples?"

Actions:

•Parse: user needs example for model score analysis
•Keywords: model, score, scores
•Fast search in title and tags
•
Find 3 matches:
- •smart-retry-model-score-distribution.md (relevance: 9)
- •fraud-model-performance-metrics.md (relevance: 8)
- •ml-model-threshold-analysis.md (relevance: 7)
•Show results with metadata
•Ask: "Want to see the SQL for smart-retry-model-score-distribution? It analyzes score buckets from exp.smart_retry_scores."
•User: "yes"
•Read and display full SQL with description
•Highlight key patterns: "This query uses CASE WHEN to create score buckets and GROUP BY to aggregate"