Search_Mode_Guide

Purpose

Use this guide to choose search_mode (keyword vs semantic) for GovTribe Search_* requests. Use it when:

• a request needs exact lookup by identifiers, codes, titles, or quoted phrases.

• a request is conceptual and needs meaning-based retrieval.

• the user asks for aggregations, counts, top-N lists, distributions, or trends.

• a request mixes strict identifiers with exploratory context and needs a tie-break decision.

Decision Checklist

Parameter Contract

• Omit search_mode to use the default keyword behavior.

• Always include query.

• Keep mode and query style aligned:

  • keyword: exact tokens, identifiers, and literal overlap.

  • semantic: plain-language intent and conceptual similarity.

Aggregations Support

• Aggregations and rollups should use the default keyword behavior.

• Aggregation signals include totals, counts, top-N lists, breakdowns by field, distributions, and time-series trends.

• Set search_mode: "semantic" only when you intentionally want semantic retrieval; it should not be the default for aggregation-driven asks.

Keyword Mode

• What it does:

  • prioritizes lexical and phrase overlap with typo tolerance.

• Strengths:

  • best for exact identifiers, codes, titles, quoted phrases, and short specific lookups.

  • supports aggregation workflows.

• Limitations:

  • weaker for broad conceptual asks where synonym coverage drives relevance.

• Choose keyword when:

  • a query includes identifier-like tokens (for example solicitation IDs, UEI/CAGE, NAICS/PSC).

  • a query includes quoted text or exact title/name intent.

  • the request is a direct lookup or navigation to a specific item.

  • the query is short and specific (roughly 1-3 salient tokens).

Semantic Mode

• What it does:

  • prioritizes conceptual similarity and meaning over strict token overlap.

• Strengths:

  • best for exploratory requests such as “similar to,” “alternatives,” and “ways to.”

  • handles synonym-heavy and wording-variable topics.

• Limitations:

  • does not reliably support aggregation-heavy workflows.

  • can underweight strict identifier precision.

• Choose semantic when:

  • the request is conceptual, exploratory, or intent-driven.

  • the query reads like a question or guidance ask.

  • the topic is broad/ambiguous and depends on synonym matching.

One-Screen Decision Checklist

• Choose keyword if any are true:

  • the query includes quoted text.

  • the query includes identifier-like tokens or strict codes.

  • the intent is direct lookup.

  • the user asks for aggregations or rollups.

• Choose semantic if any are true:

  • the request asks for similar, related, or alternative results.

  • the query is conceptual and meaning-driven.

  • the query has no unique identifiers and would benefit from synonym matching.

• Tie-breakers:

  • unique identifier plus conceptual context: choose keyword and keep the ID explicit.

  • no unique tokens and broad concept: choose semantic.

  • strict phrase plus “similar to”: choose semantic unless a unique ID is present.

Examples

Default Keyword-Behavior Examples

Exact solicitation lookup: Tool: Search_Federal_Contract_Opportunities

{
  "query": "\"W912HQ-24-R-0123\""
}

Quoted phrase and entity lookup: Tool: Search_Federal_Contract_Opportunities

{
  "query": "\"CISA\" \"endpoint detection\" RFI"
}

Aggregation-driven request with structured filters: Tool: Search_Federal_Contract_Awards

{
  "query": "",
  "aggregations": ["top_awardees_by_dollars_obligated"],
  "naics_category_ids": ["541512"],
  "naics_category_ids_operator": "in"
}

Semantic Mode Examples

Conceptual similarity request: Tool: Search_Federal_Contract_Opportunities

{
  "query": "find opportunities similar to FAA SWIM support for aviation data integration",
  "search_mode": "semantic"
}

Alternatives-oriented discovery request: Tool: Search_Federal_Contract_Opportunities

{
  "query": "alternatives to legacy ERP modernization support contracts for defense logistics organizations",
  "search_mode": "semantic"
}

Broad intent without unique identifiers: Tool: Search_Federal_Contract_Opportunities

{
  "query": "ways to find recompete opportunities for small business cloud modernization work",
  "search_mode": "semantic"
}