Search Modes

OmniSearches offers five distinct search modes, each optimized for different types of queries and use cases. Each mode uses Google’s Gemini 2.0 Flash model with customized parameters to deliver the best results.

Available Modes

Concise Mode

Quick, direct answers with minimal detail

Default Mode

Balanced responses with comprehensive information

Exhaustive Mode

Deep, thorough research with extensive detail

Search Mode

URL-focused results for finding specific resources

Reasoning Mode

Advanced analysis with Deepseek reasoning integration

Mode Specifications

Concise
Default
Exhaustive
Search
Reasoning

Concise Mode

Perfect for quick answers when you need information fast without extensive details.Model Configuration:

{
  model: "gemini-2.0-flash",
  temperature: 0.1,      // Very low for consistent, focused responses
  topP: 1,               // Full probability mass
  topK: 1,               // Single most likely token
  maxOutputTokens: 150   // Limited output for brevity
}

Best For:

Quick facts and definitions
Simple yes/no questions
Brief explanations
Fast information lookup

Example Queries:

“What is the capital of France?”
“When was Python created?”
“Netflix login”

Concise mode prioritizes speed and brevity over comprehensive detail. Responses are limited to 150 tokens.

Default Mode

The balanced option that provides comprehensive information with good depth while maintaining readability.Model Configuration:

{
  model: "gemini-2.0-flash",
  temperature: 1.2,      // Higher for creative, diverse responses
  topP: 0.95,            // Top 95% probability mass
  topK: 40,              // Consider top 40 tokens
  maxOutputTokens: 65536 // Full output capacity
}

Capabilities:

Search Mode: For keyword-based queries finding specific websites (≥5 searches)
Information Assistant Mode: For complex queries requiring explanations
Mixed Mode: Automatic fallback for ambiguous queries
Image Gallery: Includes relevant images from Wikimedia Commons
Related Questions: Suggests 3 follow-up questions

Best For:

General information requests
Comparing products or concepts
Understanding complex topics
Exploring new subjects

Example Queries:

“What are the benefits of meditation?”
“Compare iPhone 14 and Samsung Galaxy S23”
“Find kid-friendly activities in Sydney”

Default mode is recommended for most searches as it balances depth with readability.

Exhaustive Mode

Maximum depth and detail for comprehensive research and thorough understanding.Model Configuration:

{
  model: "gemini-2.0-flash",
  temperature: 0.8,      // Moderate creativity
  topP: 0.95,            // Top 95% probability mass
  topK: 40,              // Consider top 40 tokens
  maxOutputTokens: 65536 // Full output capacity
}

Enhanced Features:

Performs ≥10 searches (vs 5 in default mode)
Extended image galleries (10+ images)
Comprehensive source citations
Detailed analysis and explanations
Multiple perspectives on topics

Best For:

Academic research
In-depth topic exploration
Professional analysis
Comprehensive reports

Example Queries:

“Explain quantum computing in detail”
“Comprehensive analysis of climate change impacts”
“History of artificial intelligence”

Exhaustive mode may take longer to respond due to extensive research and content generation.

Search Mode

Optimized for finding and listing relevant URLs with descriptions.Model Configuration:

{
  model: "gemini-2.0-flash",
  temperature: 0.4,      // Lower for focused results
  topP: 1,               // Full probability mass
  topK: 1,               // Single most likely token
  maxOutputTokens: 1024  // Moderate output
}

Output Format:

[URL] - [Concise description of website content]
[URL] - [Concise description of website content]
...

RELATED_QUESTIONS:
1. [Question related to the search query]
2. [Question related to the search query]
3. [Question related to the search query]

Best For:

Finding specific websites
Resource discovery
Curated link lists
Website recommendations

Example Queries:

“Best online learning platforms”
“JavaScript documentation sites”
“Design inspiration websites”

Search mode returns structured lists of URLs with descriptions, perfect for resource discovery.

Reasoning Mode

Advanced mode that combines Deepseek’s reasoning capabilities with Gemini’s search tools.Model Configuration:

{
  model: "gemini-2.0-flash",
  temperature: 1.0,      // Balanced creativity
  topP: 0.95,            // Top 95% probability mass
  topK: 40,              // Consider top 40 tokens
  maxOutputTokens: 65536 // Full output capacity
}

Two-Stage Process:

Reasoning Stage: Deepseek analyzes the query and creates a search strategy
Search Stage: Gemini executes the strategy and provides results

Reasoning Model (via OpenAI-compatible API):

Uses Deepseek Reasoner (configurable)
Streams reasoning process to client
Defines information needs and explanation strategy

Best For:

Complex analytical queries
Multi-step problems
Strategic research
Academic analysis

Example Queries:

“How would implementing universal basic income affect the economy?”
“Analyze the trade-offs between different database architectures”
“What are the ethical implications of AI in healthcare?”

Reasoning mode shows you the AI’s thought process before searching, providing transparency in how it approaches your query.

See Reasoning Mode for detailed information.

Switching Between Modes

You can easily switch modes using the mode selector in the search interface:

// Mode selector component implementation
import { SearchControls } from '@/components/SearchControls'

<SearchControls
  searchMode="default"
  onSearchModeChange={(mode) => setSearchMode(mode)}
  direction="right"
/>

View SearchControls Component

The SearchControls component provides a dropdown interface with:

Visual mode indicators: Each mode has a unique icon
Mode descriptions: Hover to see what each mode does
Keyboard navigation: Use arrow keys to navigate modes
Persistent selection: Your choice is remembered during the session

Available icons:

Concise: MessageSquare
Default: MessageSquareMore
Exhaustive: Telescope
Search: Search
Reasoning: Sparkle

Common System Instructions

All modes (except Concise) follow this core behavior:

Tool Usage

Always use the google_search tool to find information. Never rely solely on internal knowledge.

Mode Selection

Automatically determine whether to use:

Search Mode: For keyword queries seeking specific websites
Information Assistant Mode: For complex questions requiring explanation
Mixed Mode: When query intent is ambiguous

Response Formatting

Structure responses with proper markdown, including:

Headers and sections
Bullet points and lists
Source citations
Related questions

API Usage

To use different modes programmatically:

curl "https://api.omnisearches.com/api/search?q=your+query&mode=default"

Performance Comparison

Response times and token usage vary based on query complexity and mode selection.

Mode	Avg Response Time	Max Tokens	Searches	Images
Concise	< 2s	150	Auto	No
Default	3-5s	65,536	≥5	4-8
Exhaustive	5-10s	65,536	≥10	10+
Search	2-4s	1,024	Auto	No
Reasoning	8-15s	65,536	≥10	4-8

Image Search

Learn how image galleries work in Default, Exhaustive, and Reasoning modes

Reasoning Mode

Deep dive into the two-stage reasoning process

Get Started

Features

Configuration

Development

Available Modes

Concise Mode

Default Mode

Exhaustive Mode

Search Mode

Reasoning Mode

Mode Specifications

Concise Mode

Default Mode

Exhaustive Mode

Search Mode

Reasoning Mode

Switching Between Modes

Common System Instructions

API Usage

Performance Comparison

Image Search

Related Questions

Reasoning Mode

Build docs developers (and LLMs) love

Get Started

Features

Configuration

Development

​Available Modes

Concise Mode

Default Mode

Exhaustive Mode

Search Mode

Reasoning Mode

​Mode Specifications

​Concise Mode

​Default Mode

​Exhaustive Mode

​Search Mode

​Reasoning Mode

​Switching Between Modes

​Common System Instructions

​API Usage

​Performance Comparison

​Related Features

Image Search

Related Questions

Reasoning Mode

Build docs developers (and LLMs) love

Available Modes

Mode Specifications

Concise Mode

Default Mode

Exhaustive Mode

Search Mode

Reasoning Mode

Switching Between Modes

Common System Instructions

API Usage

Performance Comparison

Related Features