search-strategy

name	search-strategy
description	Query decomposition and multi-source search orchestration. Breaks natural language questions into targeted searches per household source, translates queries into source-specific syntax, ranks results by relevance, and handles ambiguity and fallback strategies.

Search Strategy

The core intelligence behind household search. Transforms a single natural language question into parallel, source-specific searches and produces ranked, deduplicated results.

The Goal

Turn this:

"When does the property tax payment come due for Maple Street?"

Into targeted searches across every connected source:

google-workspace:  "property tax Maple Street due" (email search)
paperless:         "property tax" (document search, tagged with property)
estate-planning:   get_upcoming_dates (compliance/deadline lookup)
actual-budget:     analytics(operation="spending_by_category") for property tax category

Then synthesize the results into a single coherent answer.

Query Decomposition

Step 1: Identify Query Type

Classify the user's question to determine search strategy:

Query Type	Example	Strategy
Document	"Where's the trust agreement?"	Prioritize paperless, Drive, estate-planning
Financial	"How much did we spend on groceries?"	Prioritize actual-budget, then receipts in paperless
Deadline	"When is the next filing due?"	Prioritize estate-planning dates, calendar, memos
Person	"What did Jim send us?"	Search email, paperless by correspondent
Asset/Entity	"What entities own the rental?"	Prioritize estate-planning, then supporting docs
Household	"Do we have olive oil?"	Prioritize grocy, then mealie recipes
Exploratory	"What do we know about the Greenwood application?"	Broad search across all sources

Step 2: Extract Search Components

From the query, extract:

• Keywords: Core terms that must appear in results
• Entities: People, properties, accounts, trusts, assets (use memory system if available)
• Intent signals: Deadline words, financial words, document words
• Constraints: Time ranges, source hints, category filters
• Negations: Things to exclude

Step 3: Generate Sub-Queries Per Source

For each available source, create one or more targeted queries:

Prefer semantic/keyword search for:

• Conceptual questions ("What do we know about...")
• Questions where exact keywords are unknown
• Exploratory queries

Prefer structured queries for:

• Financial data (specific categories, date ranges, accounts)
• Entity/asset lookups (known names, types)
• Inventory checks (specific items)

Generate multiple query variants when the topic might be referred to differently:

User: "Maple Street property"
Queries: "Maple Street", "123 Maple", "maple reno", "rental property"

Source-Specific Query Translation

Google Workspace (email & docs)

Email search (google-workspace.search_gmail_messages):

query: "property tax Maple Street"
query: "from:jim trust amendment"

Drive search (google-workspace.search_drive_files):

query: "estate plan 2025"

Filter mapping:

User filter	Google Workspace parameter
from:jim	sender filter
after:2025-01-01	date range
type:pdf	file type filter

Paperless (documents)

Document search (paperless.search_documents):

query: "property tax bill"
correspondent: "County Assessor"
document_type: "Tax Document"

Good for: receipts, contracts, tax forms, insurance policies, legal documents, scanned mail.

Actual Budget (finances)

Financial queries (actual-budget.analytics):

operation: "spending_by_category"
operation: "monthly_summary"
operation: "balance_history"

Filter mapping:

User filter	Actual Budget parameter
Category	category filter on analytics
Date range	start_date / end_date
Account	account filter

Estate Planning (entities/assets/dates)

Entity/asset lookup:

estate-planning.list_assets → all assets with metadata
estate-planning.list_entities → trusts, LLCs, etc.
estate-planning.get_upcoming_dates → compliance deadlines

Memos (notes)

Note search (memos.search_memos):

query: "decision about school enrollment"

Good for: meeting notes, decisions, reminders, quick references.

Homebox (inventory)

Item search (homebox.list_items):

query: "warranty dishwasher"

Good for: warranties, manuals, serial numbers, purchase history.

Mealie & Grocy (kitchen)

Recipe search: mealie.get_recipes, mealie.get_recipe_detailed Pantry search: grocy.get_stock_overview, grocy.get_expiring_products

Result Ranking

Relevance Scoring

Score each result on these factors (weighted by query type):

Factor	Weight (Document)	Weight (Financial)	Weight (Deadline)	Weight (Exploratory)
Keyword match	0.4	0.2	0.2	0.3
Freshness	0.2	0.3	0.4	0.2
Authority	0.3	0.3	0.3	0.2
Completeness	0.1	0.2	0.1	0.3

Authority Hierarchy

Depends on query type:

For document/legal questions:

Paperless (OCR'd originals) > Drive docs > Email attachments > Memos

For financial questions:

Actual Budget (structured data) > Paperless receipts > Email confirmations > Memos

For deadline/compliance questions:

Estate Planning dates > Calendar events > Email reminders > Memos

Handling Ambiguity

When a query is ambiguous, prefer asking one focused clarifying question over guessing:

Ambiguous: "search for the application"
→ "I found references to a few applications. Are you looking for:
   1. The Greenwood Academy enrollment application
   2. The building permit application for Maple Street
   3. Something else?"

Only ask for clarification when:

• There are genuinely distinct interpretations that would produce very different results
• The ambiguity would significantly affect which sources to search

Do NOT ask for clarification when:

• The query is clear enough to produce useful results
• Minor ambiguity can be resolved by returning results from multiple interpretations

Fallback Strategies

When a source is unavailable or returns no results:

1. Source unavailable: Skip it, search remaining sources, note the gap
2. No results from a source: Try broader query terms, remove date filters, try alternate keywords
3. All sources return nothing: Suggest query modifications to the user
4. Rate limited: Note the limitation, return results from other sources, suggest retrying later

Query Broadening

If initial queries return too few results:

Original: "Maple Street property tax Q1 payment receipt"
Broader:  "property tax Maple"
Broader:  "property tax"
Broadest: "tax payment"

Remove constraints in this order:

1. Date filters (search all time)
2. Source/location filters
3. Less important keywords
4. Keep only core entity/topic terms

Parallel Execution

Always execute searches across sources in parallel, never sequentially. The total search time should be roughly equal to the slowest single source, not the sum of all sources.

[User query]
     ↓ decompose
[email query] [paperless query] [budget query] [estate query] [memos query]
     ↓            ↓            ↓              ↓            ↓
  (parallel execution)
     ↓
[Merge + Rank + Deduplicate]
     ↓
[Synthesized answer]