Web Search Skill

Here’s the full, polished Web Search Skill Documentation for the Raia platform, incorporating your updated explanation of how it works:


🌐 Raia Web Search Skill Documentation

The Web Search Skill enables AI agents on the Raia platform to access real-time information from the internet, allowing them to provide up-to-date, relevant, and context-aware answers. This dynamic capability helps bridge gaps in static training data and ensures more responsive and informed conversations across a variety of use cases.


🧠 Purpose

Use the Web Search Skill when agents need to:

  • Retrieve live or recent data not found in internal sources

  • Fact-check statements or verify details

  • Answer time-sensitive questions (e.g. news, pricing, events)

  • Reference public information to supplement domain-specific knowledge


🔍 Key Capabilities

Feature
Description

Live Web Lookup

Real-time querying via OpenAI’s Web Search integration

Contextual Inference

Automatically determines if a search is necessary based on the prompt

Integrated Response Generation

Blends search results with model knowledge for seamless, natural answers

Trusted Source Ranking

Prioritizes reliable and high-authority sources


⚙️ How It Works

The Web Search Skill uses OpenAI’s Web Search tool, which is intelligently invoked by the AI agent when needed—no manual trigger required.

  1. Autonomous Triggering When a user asks a question, the agent evaluates whether it can confidently answer using its internal training or previously provided documents. If not, it autonomously performs a web search.

  2. Query Execution The agent sends the query through OpenAI’s Web Search tool, retrieving real-time results from the public internet.

  3. Integrated Reasoning Instead of quoting web results directly, the AI blends them into its working context, alongside its trained knowledge and your custom system instructions. This creates a holistic understanding that powers a highly accurate and fluent answer.

  4. Final Output The agent returns a response that is informative, timely, and conversational, seamlessly enriched with insights from the web—without disrupting the tone or logic of the overall interaction.

✅ Users don’t need to know when a search happens—it’s fully automated and invisible unless explicitly requested.


✅ Best Practices

  • Use cases that benefit from Web Search:

    • Current events, financial updates, market trends

    • Business operating hours, availability, or weather

    • Product reviews, Reddit discussions, social buzz

    • Verifying claims or identifying breaking changes

  • When NOT to rely on it:

    • Secure/internal-only knowledge

    • Complex topics requiring proprietary documents

    • Questions where exact citations are mandatory (use scraping or static sources instead)


🧪 Example Use Cases

User Prompt
What the Agent Does

“What’s the latest on the AI copyright lawsuit?”

Searches news sites for updates and summarizes current events

“Are there any delays at TPA airport today?”

Checks public travel alerts and news feeds for live status

“What are people saying about Apple’s new headset?”

Aggregates sentiment and highlights key Reddit or blog discussions

“Is Google Cloud still offering their AI free tier?”

Looks up the most recent product/pricing page from official sources


🧰 Integration Notes (for Developers)

  • Web Search is baked into the Raia agent runtime and does not require custom API setup.

  • It plays well with other skills like Scraping, Document Retrieval, and Memory.

  • Developers can encourage or suppress Web Search through agent instructions if finer control is needed.

Last updated