web-research

name: web-research category: information description: Search the web for facts and current information. Brave Search (requires API key), Wikipedia and DuckDuckGo (free). Tools: http. test_prompt: What is artificial intelligence? permissions: - android.permission.INTERNET credentials: - id: brave_search_api_key label: Brave Search API Key type: api_key

Web Research Skill

Search the web, look up facts, read web pages, and synthesize information from multiple sources.

Tools

Source 1: Brave Search API (full web search – requires API key)

Brave Search provides comprehensive web results with snippets. Free tier: 2000 queries/month.

Always check the API key first:

{
  "tool": "check_credential",
  "args": { "credential_id": "brave_search_api_key" }
}

If configured: false → fall back to Wikipedia/DuckDuckGo (see below). Do NOT attempt the Brave API call.

Web search

{
  "method": "GET",
  "url": "https://api.search.brave.com/res/v1/web/search?q={QUERY}&count=5",
  "auth_provider": "brave_search_api_key",
  "auth_header": "X-Subscription-Token"
}

Response fields:

• web.results[] – array of search results
  • .title – page title
  • .url – page URL
  • .description – snippet/summary (HTML stripped)
  • .age – how old the result is (e.g. "2 hours ago")
• query.original – the original search query

News search

{
  "method": "GET",
  "url": "https://api.search.brave.com/res/v1/news/search?q={QUERY}&count=5",
  "auth_provider": "brave_search_api_key",
  "auth_header": "X-Subscription-Token"
}

Response fields:

• results[] – array of news articles
  • .title – article title
  • .url – article URL
  • .description – summary snippet
  • .age – publication time (e.g. "3 hours ago")
  • .meta_url.hostname – source domain (e.g. "bbc.com")

Source 2: Wikipedia REST API (free, no API key)

Perfect for factual lookups, definitions, biographies, historical events, and encyclopedic queries.

Search Wikipedia

{
  "method": "GET",
  "url": "https://en.wikipedia.org/api/rest_v1/page/summary/{TOPIC}",
  "headers": [{ "key": "User-Agent", "value": "SannaBot/1.0" }]
}

Response fields:

• title – article title
• extract – plain text summary (first paragraph)
• description – short description
• content_urls.desktop.page – full article URL

Language support: Replace en with the target language code:

• German: https://de.wikipedia.org/api/rest_v1/page/summary/{TOPIC}
• French: https://fr.wikipedia.org/api/rest_v1/page/summary/{TOPIC}
• Spanish: https://es.wikipedia.org/api/rest_v1/page/summary/{TOPIC}
• Italian: https://it.wikipedia.org/api/rest_v1/page/summary/{TOPIC}

Wikipedia search (when exact title is unknown)

{
  "method": "GET",
  "url": "https://en.wikipedia.org/w/api.php?action=query&list=search&srsearch={QUERY}&utf8=&format=json&srlimit=5",
  "headers": [{ "key": "User-Agent", "value": "SannaBot/1.0" }]
}

Response fields:

• query.search[] – array of results
  • .title – article title (use this with the summary endpoint)
  • .snippet – text excerpt (contains <span> tags – strip HTML)

Workflow: Search → pick best matching title → fetch summary via REST API.

Source 3: DuckDuckGo Instant Answer API (free, no API key)

Provides quick factual answers, definitions, and related topics. Best for simple "what is X?" queries.

{
  "method": "GET",
  "url": "https://api.duckduckgo.com/?q={QUERY}&format=json&no_html=1&skip_disambig=1"
}

Response fields:

• AbstractText – short answer/summary (often from Wikipedia)
• AbstractSource – source name
• AbstractURL – source URL
• Answer – direct answer (e.g. for calculations, conversions)
• RelatedTopics[] – list of related topics
  • .Text – topic description
  • .FirstURL – link to more info

Note: If both AbstractText and Answer are empty, DuckDuckGo has no instant answer → fall back to Brave Search or Wikipedia.

Source 4: Fetching web pages (reading articles)

Use the http tool with response_format: "text" to fetch and read a specific URL:

{
  "method": "GET",
  "url": "{PAGE_URL}",
  "response_format": "text",
  "headers": [{ "key": "User-Agent", "value": "SannaBot/1.0" }]
}

Important:

• Raw HTML is noisy – the 5000-char truncation means you may only get the header/nav portion
• Prefer structured APIs (Wikipedia, search API snippets) over raw page fetches
• If you must fetch a page, extract key information from the HTML and summarize
• For news articles, prefer the headlines skill (RSS feeds) when available

Decision Matrix: Which source to use

Workflows

"Search the web for [topic]" (general web search)

1. check_credential("brave_search_api_key")
2. If configured:
   • http → GET https://api.search.brave.com/res/v1/web/search?q={TOPIC}&count=5 with auth_provider: "brave_search_api_key", auth_header: "X-Subscription-Token"
   • Parse web.results[]: extract title, description, URL for each result
   • Summarize: "Here's what I found about {TOPIC}: 1. … 2. … 3. …"
3. If NOT configured:
   • http → GET Wikipedia search API for {TOPIC}
   • Pick best result → fetch Wikipedia summary
   • Also try DuckDuckGo Instant Answer API
   • Summarize findings. Mention to the user: "For broader web search results, you can add a Brave Search API key in Settings."

"What is [X]?" / "Who is [person]?" (factual lookup)

1. http → GET https://en.wikipedia.org/api/rest_v1/page/summary/{X} (use de.wikipedia.org for German users, etc.)
2. If Wikipedia returns a result → summarize extract field
3. If Wikipedia returns 404 (no article):
   • Try Wikipedia search API to find the correct title
   • Or fall back to DuckDuckGo Instant Answer API
4. Present the answer naturally, citing the source

"Latest news about [topic]" (news search)

1. check_credential("brave_search_api_key")
2. If configured:
   • http → GET https://api.search.brave.com/res/v1/news/search?q={TOPIC}&count=5 with auth
   • Summarize results with source names and publication times
3. If NOT configured:
   • Suggest using the headlines skill for RSS-based news instead
   • Or try Wikipedia for background on the topic

"Tell me about [X] and [Y]" (comparison / multi-topic)

1. Fetch Wikipedia summaries for both X and Y (parallel http calls)
2. If more detail needed and Brave Search is configured, search for "X vs Y" or "X compared to Y"
3. Synthesize a comparison from the gathered information

"Read this page: [URL]" (direct page fetch)

1. http → GET {URL} with response_format: "text"
2. Extract meaningful content from the HTML (skip navigation, ads, scripts)
3. Summarize the page content for the user
4. Note: content will be truncated at 5000 chars – warn the user if the article appears cut off

Deep research (multi-step)

1. Start with a broad search (Brave or Wikipedia)
2. Identify the most relevant results
3. Optionally fetch 1–2 promising URLs for more detail
4. Synthesize all findings into a coherent answer
5. Cite sources: "According to [Source]: …"

URL encoding rules

• Replace spaces with + or %20: artificial+intelligence
• Encode special chars: & → %26, # → %23, ? → %3F
• Umlauts can be used directly in Wikipedia URLs: Zürich, München
• For Brave Search, URL-encode the query parameter

Output format rules

• Voice (TTS): Summarize in 2–4 sentences. Name the source: "According to Wikipedia…" / "A web search shows…"
• Text response: Present results as a numbered list with source names, brief descriptions, and links
• Always summarize in natural language – never dump raw JSON/HTML
• Always cite sources – mention where the information comes from
• Multiple results: Present the top 3–5 most relevant results, not all
• If all sources return no results, tell the user honestly: "I couldn't find information about {topic}."
• Summarize in the user's configured language, even if the source is in a different language

Notes

• Brave Search API free tier: 2000 queries/month – sufficient for personal use
• Wikipedia is always available as a free, high-quality fallback
• The http tool truncates responses at 5000 characters – this is enough for search results and Wikipedia summaries
• For Brave Search: get a free API key at https://brave.com/search/api/
• DuckDuckGo Instant Answer API often returns empty results for complex queries – use it only for simple factual lookups
• When the user says "Google X" or "search for X", they mean web search – use Brave Search or Wikipedia, not literally Google