Skip to main content

Advanced search tips

Get more from Synjar's search with these advanced techniques.

Query syntax​

Exact phrases​

Wrap phrases in quotes for literal matches:

"annual subscription refund policy"

Tag filtering​

Search within specific tags:

tag:support-docs how to reset password

Combining filters​

Use multiple filters together:

tag:engineering "database migration"

Optimizing results​

Relevance scoring​

Results are scored 0-1 based on semantic similarity:

  • 0.9+ - Excellent match
  • 0.7-0.9 - Good match
  • 0.5-0.7 - Partial match
  • <0.5 - Weak match

Improving relevance​

  1. Use domain vocabulary - Match terms used in your documents
  2. Be specific - "How do I configure SMTP for password reset emails?" > "email setup"
  3. Add context - Include relevant product names, versions, departments

Search strategies​

When you're not sure what to ask:

  1. Start broad: "onboarding process"
  2. Review results for terminology
  3. Refine with specific terms found

When you know what you're looking for:

  1. Use exact phrases: "error code 403"
  2. Include unique identifiers
  3. Filter by relevant tags

Finding differences or comparisons:

  • "What's the difference between Team and Business plans?"
  • "How does the API differ between v1 and v2?"

Working with large knowledge bases​

Use tags strategically​

  • Tag documents by topic, department, or project
  • Filter searches to relevant subsets
  • Create search links with tag scopes

For specific use cases, create dedicated search links:

  • Public FAQ search (tag: public)
  • Engineering docs (tag: engineering)
  • HR policies (tag: hr-policies)

For developers, the API offers additional control:

const results = await search({
  query: "password reset",
  limit: 10,           // Number of results
  threshold: 0.7,      // Minimum relevance
  tags: ["support"],   // Filter by tags
});

See API reference for full documentation.

Performance tips​

Query length​

  • Optimal: 5-15 words
  • Very long queries may dilute relevance
  • Very short queries may be too ambiguous

Batch vs real-time​

  • Interactive search: Real-time semantic search
  • Large-scale retrieval: Use API with pagination

See also​