Providers

What are providers?

Providers are the AI platforms that OneGlanse monitors for brand mentions. Each provider represents a different AI service that users interact with to get information, recommendations, and answers. OneGlanse supports five major AI providers:

ChatGPT

OpenAI’s conversational AI assistant

Claude

Anthropic’s AI assistant focused on helpful, honest, and harmless interactions

Perplexity

AI-powered search engine with real-time information

Gemini

Google’s multimodal AI model and assistant

AI Overview

Google’s AI-generated summaries in search results

Why providers matter

Different AI providers have different: Knowledge sources - Each provider trains on different datasets and accesses different information Response styles - Providers format answers differently (lists, paragraphs, comparative tables) Citation practices - Some providers extensively cite sources, others don’t Update frequencies - Providers refresh their knowledge at different intervals User bases - Each platform serves different user demographics and use cases Monitoring multiple providers gives you comprehensive coverage of how AI represents your brand across the ecosystem.

How providers work

Provider architecture

OneGlanse uses a provider config system that defines all provider-specific behavior in one place:

interface ProviderConfig {
  url: string;                // Landing URL for the provider
  warmupDelayMs: number;      // Initial wait after page load
  label: string;              // Short identifier (e.g., "ChatGPT")
  displayName: string;        // UI-friendly name
  skip?: boolean;             // Disable this provider
  requiresWarmup: boolean;    // Pre-prompt editor preparation
  
  // Core operations
  waitForResponse: (page: Page) => Promise<void>;
  extractResponse: (page: Page) => Promise<string>;
  extractSources: (page: Page) => Promise<Source[]>;
  
  // Optional hooks
  beforeRetryHook?: (page: Page) => Promise<void>;
  betweenPromptsHook?: (page: Page) => Promise<void>;
  checkSubmitSuccess?: (page: Page, preSubmitUrl: string) => Promise<boolean | undefined>;
  preNavigationHook?: (page: Page) => Promise<void>;
  postNavigationHook?: (page: Page) => Promise<void>;
}

This architecture makes adding new providers straightforward - create a config file and register it. No other code needs to change.

Provider registry

All providers are registered in a central configuration:

import { chatgptConfig } from "./chatgpt";
import { claudeConfig } from "./claude";
import { perplexityConfig } from "./perplexity";
import { geminiConfig } from "./gemini";
import { aiOverviewConfig } from "./ai-overview";

export const PROVIDER_CONFIGS: Record<Provider, ProviderConfig> = {
  "chatgpt": chatgptConfig,
  "claude": claudeConfig,
  "perplexity": perplexityConfig,
  "gemini": geminiConfig,
  "ai-overview": aiOverviewConfig,
};

This single source of truth ensures consistent provider handling throughout the system.

Provider configurations

ChatGPT

ChatGPT uses a conversational interface with streaming responses and optional source citations.

export const chatgptConfig: ProviderConfig = {
  url: "https://chatgpt.com/",
  warmupDelayMs: 5000,
  label: "ChatGPT",
  displayName: "ChatGPT",
  requiresWarmup: true,
  
  waitForResponse: (page) => waitForAssistantToFinish(page, "chatgpt"),
  extractResponse: (page) => extractAssistantMarkdown(page, "chatgpt"),
  
  extractSources: async (page) => {
    const btn = await findSourcesButton(page);
    if (!btn) return [];
    await openSourcesPanel(page, btn);
    return extractSourcesFromChatgpt(page, btn);
  },
};

Key characteristics:

Requires editor warm-up before first prompt
Streaming text generation
Sources available via expandable panel
5-second warmup delay

Claude

Claude provides conversational AI with a clean interface and streaming responses.

export const claudeConfig: ProviderConfig = {
  url: "https://claude.ai/new",
  warmupDelayMs: 5000,
  skip: true,  // Currently disabled
  label: "Claude",
  displayName: "Claude",
  requiresWarmup: true,
  
  waitForResponse: (page) => waitForAssistantToFinish(page, "claude"),
  extractResponse: (page) => extractAssistantMarkdown(page, "claude"),
  extractSources: async (_page) => [],  // No source citations
};

Key characteristics:

Starts on new conversation page
No source citations (returns empty array)
Currently disabled in default configuration

Perplexity

Perplexity is an AI search engine that extensively cites sources for its answers.

export const perplexityConfig: ProviderConfig = {
  url: "https://www.perplexity.ai/",
  warmupDelayMs: 5000,
  label: "Perplexity",
  displayName: "Perplexity",
  requiresWarmup: true,
  
  waitForResponse: (page) => waitForAssistantToFinish(page, "perplexity"),
  extractResponse: (page) => extractAssistantMarkdown(page, "perplexity"),
  
  postNavigationHook: async (page) => {
    // Anti-bot detection: randomized delays
    const randomDelay = 2000 + Math.floor(Math.random() * 3000);
    await page.waitForTimeout(randomDelay);
    await page.waitForTimeout(1000 + Math.floor(Math.random() * 1000));
  },
  
  extractSources: async (page) => {
    const btn = await findSourcesButton(page);
    if (!btn) return [];
    await openSourcesPanel(page, btn);
    return extractSourcesFromPerplexity(page);
  },
};

Key characteristics:

Real-time web search integration
Extensive source citations with numbered references
Randomized delays to avoid bot detection
Rich source metadata extraction

Gemini

Google’s Gemini provides multimodal AI capabilities with search integration.

export const geminiConfig: ProviderConfig = {
  url: "https://gemini.google.com/",
  warmupDelayMs: 5000,
  label: "Gemini",
  displayName: "Gemini",
  requiresWarmup: true,
  
  waitForResponse: (page) => waitForAssistantToFinish(page, "gemini"),
  extractResponse: (page) => extractAssistantMarkdown(page, "gemini"),
  
  extractSources: async (page) => {
    const btn = await findSourcesButton(page);
    if (!btn) return [];
    await openSourcesPanel(page, btn);
    return extractSourcesFromGemini(page, btn);
  },
};

Key characteristics:

Google Search integration
Source citations available
Multimodal response support (text focus for OneGlanse)

AI Overview

Google’s AI Overview appears in search results, providing AI-generated summaries.

export const aiOverviewConfig: ProviderConfig = {
  url: "https://www.google.com/?hl=en&pws=0",
  warmupDelayMs: 0,
  label: "AI Overview",
  displayName: "AI Overview",
  requiresWarmup: false,
  
  waitForResponse: async (page) => {
    await page
      .locator(RESPONSE_CONTAINER_SELECTORS)
      .first()
      .waitFor({ state: "visible", timeout: 30000 })
      .catch(() => {
        throw new ExternalServiceError(
          "ai-overview",
          "AI Overview container not visible"
        );
      });
  },
  
  extractResponse: async (page) => {
    const html = await extractAIOverviewResponse(page);
    return turndown.turndown(html);
  },
  
  betweenPromptsHook: async (page) => {
    const input = await findActiveEditor(page, "ai-overview").catch(() => null);
    const cleared = input
      ? await clearEditorInput(page, input, { dismissWithEscape: true })
      : false;
    
    if (!cleared) {
      await navigateWithRetry(page, BASE_URL, {
        waitUntil: "domcontentloaded",
        timeout: 30000,
      });
    }
  },
  
  checkSubmitSuccess: async (page, preSubmitUrl) => {
    const currentUrl = page.url();
    if (currentUrl !== preSubmitUrl) {
      const parsed = new URL(currentUrl);
      if (parsed.searchParams.get("q")?.trim()) return true;
    }
    return undefined;
  },
  
  postNavigationHook: async (page) => {
    // Dismiss consent dialog
    await page
      .locator('button:has-text("Accept all"), button#L2AGLb')
      .first()
      .click({ timeout: 3000 })
      .catch(() => null);
  },
  
  extractSources: (page) => extractAIOverviewSources(page),
};

Key characteristics:

Search-based interface (not chat)
No warmup required
Navigation-based submission (URL changes with ?q= parameter)
Automatic consent dialog handling
Inline source citations within the AI Overview block

Browser automation

OneGlanse uses Playwright to automate real browser sessions for each provider. This approach ensures: Authentic responses - Providers see real user interactions, not API calls Complete feature support - Access to all provider capabilities, including experimental features Source extraction - Ability to extract citation links and metadata Anti-detection - Mimics human behavior with randomized delays and real browser fingerprints

Execution flow

For each prompt, OneGlanse:

Launch browser

Start a Playwright browser session with the provider’s URL:

const config = PROVIDER_CONFIGS[provider];
await page.goto(config.url);
await page.waitForTimeout(config.warmupDelayMs);

Warm up editor

For providers that require it, prepare the input editor:

if (config.requiresWarmup) {
  await warmUpEditor(page);
}

Submit prompt

Type and submit the prompt through the provider’s interface:

await askPrompt(page, prompt, provider);

Wait for completion

Use provider-specific logic to detect when streaming finishes:

await config.waitForResponse(page);

Extract data

Extract the response text and sources:

const response = await config.extractResponse(page);
const sources = await config.extractSources(page);

Source citations

Providers handle source citations differently:

ChatGPT - Optional citations

ChatGPT sometimes includes a “Sources” button that opens a panel with citations:

interface Source {
  title: "OpenAI Documentation";
  cited_text: "OpenAI provides API access to GPT models...";
  url: "https://platform.openai.com/docs";
  domain: "platform.openai.com";
  favicon: "https://platform.openai.com/favicon.ico";
}

Claude - No citations

Claude does not provide source citations. The extractSources function returns an empty array.

Perplexity - Numbered citations

Perplexity extensively cites sources with numbered references:

Example

The best project management tools include Asana[1], Monday.com[2], 
and Jira[3]. Each offers different strengths...

Sources:
[1] https://asana.com/guide - "Asana Features Guide"
[2] https://monday.com/features - "Monday.com Overview"
[3] https://atlassian.com/jira - "Jira for Software Teams"

Gemini - Search citations

Gemini includes sources from Google Search integration, available through an expandable panel.

AI Overview - Inline citations

AI Overview embeds source links directly in the response text:

Example

Project management tools like Asana (asana.com) and Monday.com 
(monday.com) help teams collaborate effectively.

Provider-specific considerations

Rate limiting

Providers may rate-limit automated requests. OneGlanse handles this through:

Randomized delays - Vary timing between requests
Retry policies - Exponential backoff on failures
Proxy rotation - Distribute requests across IP addresses

Response validation

Each response is validated before storage:

// Empty responses trigger retries
if (!response || response.trim().length === 0) {
  throw new ExternalServiceError(provider, "Empty response");
}

// Provider-specific validation rules
const validation = validateResponse(response, provider);
if (!validation.valid) {
  throw new ValidationError(`Invalid response: ${validation.reason}`);
}

Error handling

Providers can fail for various reasons:

Network timeouts
Bot detection
Interface changes
Service outages

OneGlanse implements automatic retry logic with provider-specific recovery strategies defined in beforeRetryHook.

Enabling and disabling providers

You can control which providers run for each workspace:

// Enable all providers (default)
enabledProviders: '["chatgpt","claude","perplexity","gemini","ai-overview"]'

// Enable only search-focused providers
enabledProviders: '["perplexity","ai-overview"]'

// Enable only conversational AI
enabledProviders: '["chatgpt","claude","gemini"]'

Disabling providers in your workspace configuration prevents them from running during scheduled jobs, saving resources and focusing on the platforms most relevant to your audience.

Workspaces

Configure which providers to monitor per workspace

Prompts

Learn how prompts are executed across providers

Adding Providers

Developer guide for adding new AI providers

Agent Worker

Technical details of browser automation architecture

Get Started

Core Concepts

Guides

What are providers?

ChatGPT

Claude

Perplexity

Gemini

AI Overview

Why providers matter

How providers work

Provider architecture

Provider registry

Provider configurations

ChatGPT

Claude

Perplexity

Gemini

AI Overview

Browser automation

Execution flow

Source citations

Provider-specific considerations

Rate limiting

Response validation

Error handling

Enabling and disabling providers

Workspaces

Prompts

Adding Providers

Agent Worker

Get Started

Core Concepts

Guides

Documentation Index

​What are providers?

ChatGPT

Claude

Perplexity

Gemini

AI Overview

​Why providers matter

​How providers work

​Provider architecture

​Provider registry

​Provider configurations

​ChatGPT

​Claude

​Perplexity

​Gemini

​AI Overview

​Browser automation

​Execution flow

​Source citations

​Provider-specific considerations

​Rate limiting

​Response validation

​Error handling

​Enabling and disabling providers

​Related resources

Workspaces

Prompts

Adding Providers

Agent Worker

What are providers?

Why providers matter

How providers work

Provider architecture

Provider registry

Provider configurations

ChatGPT

Claude

Perplexity

Gemini

AI Overview

Browser automation

Execution flow

Source citations

Provider-specific considerations

Rate limiting

Response validation

Error handling

Enabling and disabling providers

Related resources