Use the Microsoft Web IQ SDKs for Python and TypeScript

Official SDKs for the Microsoft Web IQ API, available in Python and TypeScript. These SDKs give your applications real-time access to web search, videos search, browse, news search, and images search — designed to ground LLM responses with accurate, up-to-date information.

Python TypeScript
Package pip install webiq npm install @microsoft/webiq
Runtime Python 3.11+ Node.js 22+
License MIT MIT

Installation

Python

pip install webiq

TypeScript

npm install @microsoft/webiq

Quickstart

Python Guide

import os
from webiq import WebIQClient
from webiq.types import BrowseContentFormat, ContentFormat

with WebIQClient(api_key=os.environ["WEBIQ_API_KEY"]) as client:
    # Web search
    web = client.web.search(
        "Python programming",
        max_results=5,
        content_format=ContentFormat.passage,
    )
    for r in web.webResults or []:
        print(r.title, r.url)

    # Videos search
    videos = client.videos.search("machine learning tutorial", max_results=5)
    for v in videos.videoResults or []:
        print(v.title, v.length)

    # Browse a URL
    page = client.browse.fetch("https://www.microsoft.com", content_format=BrowseContentFormat.markdown)
    print(page.content)

    # News search
    news = client.news.search("technology", max_results=5)
    for item in news.newsResults or []:
        print(item.title, "-", item.source)

    # Image search
    images = client.images.search("golden gate bridge", max_results=5)
    for img in images.imageResults or []:
        print(img.title, img.url)

TypeScript Guide

import { WebIQClient, BrowseContentFormat, ContentFormat } from '@microsoft/webiq';

async function main() {
  const client = new WebIQClient({ apiKey: process.env.WEBIQ_API_KEY });

  // Web search
  const web = await client.web.search('TypeScript programming', {
    maxResults: 5,
    contentFormat: ContentFormat.PASSAGE,
  });
  for (const r of web.webResults ?? []) console.log(r.title, r.url);

  // Videos search
  const videos = await client.videos.search('machine learning tutorial');
  for (const v of videos.videoResults ?? []) console.log(v.title, v.length);

  // Browse a URL
  const page = await client.browse.fetch('https://www.microsoft.com', {
    contentFormat: BrowseContentFormat.MARKDOWN,
  });
  console.log(page.content);

  // News search
  const news = await client.news.search('technology', { maxResults: 5 });
  for (const item of news.newsResults ?? []) console.log(item.title, '-', item.source);

  // Image search
  const images = await client.images.search('golden gate bridge', { maxResults: 5 });
  for (const img of images.imageResults ?? []) console.log(img.title, img.url);

  await client.close();
}

main().catch(console.error);

Authentication

API Key

# Python
import os
from webiq import WebIQClient

client = WebIQClient(api_key=os.environ["WEBIQ_API_KEY"])

// TypeScript
import { WebIQClient } from '@microsoft/webiq';

const client = new WebIQClient({ apiKey: process.env.WEBIQ_API_KEY });

Microsoft Entra ID

Pass any Azure Identity TokenCredential.

# Python — uses `azure-identity`
from azure.identity import DefaultAzureCredential
from webiq import WebIQClient

client = WebIQClient(credential=DefaultAzureCredential())

// TypeScript — uses `@azure/identity`
import { DefaultAzureCredential } from '@azure/identity';
import { WebIQClient } from '@microsoft/webiq';

const client = new WebIQClient({
  credential: new DefaultAzureCredential(),
});

SDK resources and methods

The client exposes five resource namespaces: web, videos, browse, news, and images.

client.web.search() — Search the web and get page content for grounding.

response = client.web.search(
    query="search terms",          # Required, 1–1000 chars
    max_results=10,                # 1–50 (default 10)
    language="en",                 # ISO 639-1 code
    region="US",                   # Country/region code
    location="lat:40.7;long:-74.0",  # Optional
    content_format=ContentFormat.passage,   # passage (default) | text | html | markdown (full-page content)
    max_length=3000,               # Max content chars (default 3000)
)
# response.webResults → list of { title, url, content, lastUpdatedAt, crawledAt, language, isAdult }

client.videos.search() — Find relevant video content with optional playlist support.

response = client.videos.search(
    query="search terms",          # Required, 1–1000 chars
    max_results=30,                # 1–30 (default 30)
    language="en",
    region="US",
    enable_playlist=True,          # Include playlist results
    freshness="month",             # week | month | year
)
# response.videoResults → list of { title, url, description, length, viewCount, moments[], thumbnailUrl, ... }
# response.playlists   → list of { title, videos[] }

Browse

client.browse.fetch() — Extract content from any URL. Supports live crawling and JavaScript rendering.

response = client.browse.fetch(
    url="https://www.microsoft.com",   # Required
    max_length=10000,                  # Max content chars, 1–500000
    live_crawl="fallback",             # "none" (cached only, default) | "fallback" (crawl if not cached) | "force" (always crawl, bypass cache)
    include_web_links=True,            # Include outbound hyperlinks
    include_image_links=True,          # Include image URLs
    render_dynamic_pages=False,        # JS rendering (requires live_crawl="fallback")
    content_format=BrowseContentFormat.markdown,   # text | html | markdown
)
# response → { url, title, content, isAdult, lastUpdatedAt, crawledAt, retryAfter, traceId }

client.news.search() — Get the latest news articles with the passages most relevant to the query.

response = client.news.search(
    query="search terms",          # Required, 1–1000 chars
    max_results=10,                # 1–20 (default 10)
    language="en",
    region="US",
    location="lat:40.7;long:-74.0",  # Optional
    content_format=ContentFormat.passage,
    max_length=3000,
)
# response.newsResults → list of { title, url, content, source, lastUpdatedAt, thumbnail, isAdult }

client.images.search() — Find images with filtering by aspect ratio, size, and safe search.

from webiq.types import ImageAspectRatio, ImageSize, SafeSearch

response = client.images.search(
    query="search terms",          # Required, 1–1000 chars
    max_results=30,                # 1–30 (default 30)
    language="en",
    region="US",
    aspect_ratio=ImageAspectRatio.wide,   # square | wide | tall
    image_size=ImageSize.large,           # small | medium | large | extraLarge
    safe_search=SafeSearch.strict,        # off | strict
    watermark_free=True,
)
# response.imageResults → list of { title, url, hostPageUrl, caption, width, height, thumbnailUrl, ... }

Note

TypeScript methods use the same parameters in camelCase (e.g., maxResults, contentFormat, liveCrawl, aspectRatio). See the TypeScript Quick Start above for the calling convention.

Async Support

Python Support

import asyncio
from webiq import WebIQAsyncClient

async def main():
    async with WebIQAsyncClient(api_key="your-api-key") as client:
        web, news = await asyncio.gather(
            client.web.search("async Python"),
            client.news.search("programming"),
        )
        for r in web.webResults or []:
            print(r.title)

asyncio.run(main())

TypeScript Support

TypeScript is natively async. Use Promise.all for concurrency:

const [web, news, videos] = await Promise.all([
  client.web.search('TypeScript'),
  client.news.search('programming'),
  client.videos.search('tutorial'),
]);

Configuration

Timeout and Retry

The client accepts timeout and retry as top-level options. Per-call timeout on each resource method overrides the client-level default. The SDK auto-retries 408, 500, 502, 503, 504 with exponential backoff and jitter; if the response body includes a retryAfter field (e.g. "30s"), the SDK waits for that hint instead of the backoff delay. 429 and 430 are never auto-retried — they surface immediately as RateLimitError so your application can apply its own back-off policy. Client-side timeouts are also not retried (they raise APIConnectionError).

# Python
from webiq import WebIQClient, RetryPolicy

client = WebIQClient(
    api_key="your-api-key",
    timeout=10.0,                  # seconds (default: 10.0)
    retry=RetryPolicy(
        max_retries=2,             # defaults shown
        base_delay_s=0.25,
        max_delay_s=4.0,
    ),
)

# Per-call override
response = client.web.search("query", language="de", region="DE", timeout=30.0)

// TypeScript
import { WebIQClient, RetryPolicy } from '@microsoft/webiq';

const client = new WebIQClient({
  apiKey: 'your-api-key',
  timeout: 10000, // milliseconds (default: 10000)
  retry: new RetryPolicy({
    maxRetries: 2, // defaults shown
    baseDelaySeconds: 0.25,
    maxDelaySeconds: 4.0,
  }),
});

Customized HTTP client (proxy, TLS, ...)

For advanced HTTP settings — proxy, custom TLS, connection pooling, etc. The SDK does not close caller-owned clients.

# Python — pass a pre-configured `httpx.Client` via `http_client`
import httpx
from webiq import WebIQClient, WebIQAsyncClient

# sync
http_client = httpx.Client(
  base_url="https://api.microsoft.ai/v3",
  proxy="http://proxy:8080",
)
client = WebIQClient(api_key="your-api-key", http_client=http_client)

# async
http_client = httpx.AsyncClient(
  base_url="https://api.microsoft.ai/v3",
  proxy="http://proxy:8080",
)
client = WebIQAsyncClient(api_key="your-api-key", http_client=http_client)

// TypeScript — pass a custom `fetchFn` that wraps the global `fetch`
import { ProxyAgent } from 'undici';
import { WebIQClient } from '@microsoft/webiq';

const proxyAgent = new ProxyAgent('http://my-proxy:8080');

const client = new WebIQClient({
  apiKey: 'your-api-key',
  // Spread `init` so headers, body, and the SDK's internal `AbortSignal` are preserved.
  fetchFn: (url, init) => fetch(url, { ...init, dispatcher: proxyAgent }),
});

Cancellation (TypeScript)

Every TypeScript resource method accepts an AbortSignal via options.signal. When the signal aborts, the in-flight request is cancelled, retries stop, and the signal's reason is re-thrown unchanged (no APIConnectionError wrapping).

// Apply a wall-clock budget that includes retries (the `timeout` option is per attempt only)
const result = await client.web.search('latest news', {
  signal: AbortSignal.timeout(5_000),
});

// Combine signals (e.g. an inbound request signal plus a wall-clock cap)
const result2 = await client.web.search('latest news', {
  signal: AbortSignal.any([req.signal, AbortSignal.timeout(5_000)]),
});

Telemetry Hook

Monitor every request/response for logging, metrics, or debugging.

# Python
from webiq import WebIQClient, TelemetryEvent

def on_request(event: TelemetryEvent):
    print(f"{event.method} {event.path} → {event.status_code} ({event.elapsed_ms}ms)")

client = WebIQClient(api_key="your-api-key", telemetry_hook=on_request)

// TypeScript
import { WebIQClient, TelemetryEvent } from '@microsoft/webiq';

const client = new WebIQClient({
  apiKey: 'your-api-key',
  telemetryHook: (event: TelemetryEvent) => {
    console.log(`${event.method} ${event.path} → ${event.statusCode} (${event.elapsedMs}ms)`);
  },
});

The TelemetryEvent fields are:

  • Python (snake_case): method, path, transport, attempt, status_code, request_id, trace_id, retry_after, elapsed_ms, error, error_body
  • TypeScript (camelCase): method, path, transport, attempt, statusCode, requestId, traceId, retryAfter, elapsedMs, error, errorBody

Error Handling

Both SDKs provide typed exceptions for structured error handling. All errors inherit from WebIQError.

Exception HTTP Status Trigger
AuthenticationError 401 Invalid or missing API key
PermissionDeniedError 403 Authenticated but not authorized for the resource
RateLimitError 429, 430 Rate limit / concurrent-request limit exceeded
APIStatusError 400, 404, 500, 503, 504, ... All other HTTP errors
APIConnectionError — Network issues, DNS failures, timeouts
WebIQError — Base class for all SDK errors

PermissionDeniedError and RateLimitError are both subclasses of APIStatusError, so a broader handler on APIStatusError still catches them. The SDK does not automatically retry 429 / 430 responses — the server's back-off hint is surfaced on error.retry_after (Python) / error.retryAfter (TypeScript) for your application to act on.

# Python
from webiq import (
    WebIQError,
    APIConnectionError,
    APIStatusError,
    AuthenticationError,
    PermissionDeniedError,
    RateLimitError,
)

try:
    response = client.web.search("test")
except PermissionDeniedError as e:
    # 403 — authenticated, but not allowed to call this resource
    print(f"Forbidden (HTTP {e.status_code}): {e}")
except AuthenticationError as e:
    # 401 — invalid or missing API key
    print(f"Auth failed (HTTP {e.status_code}): {e}")
except RateLimitError as e:
    # 429 or 430 — rate limit / concurrent-request limit (never auto-retried)
    print(f"Rate limited. Retry after: {e.retry_after}")
except APIStatusError as e:
    print(f"API error (HTTP {e.status_code}): {e}")
except APIConnectionError as e:
    print(f"Connection failed: {e}")
except WebIQError as e:
    print(f"SDK error: {e}")

// TypeScript
import {
  WebIQError,
  APIConnectionError,
  APIStatusError,
  AuthenticationError,
  PermissionDeniedError,
  RateLimitError,
} from '@microsoft/webiq';

try {
  const results = await client.web.search('test');
} catch (error) {
  if (error instanceof PermissionDeniedError) {
    // 403 — authenticated, but not allowed to call this resource
    console.error(`Forbidden (HTTP ${error.statusCode}):`, error.message);
  } else if (error instanceof AuthenticationError) {
    // 401 — invalid or missing API key
    console.error(`Auth failed (HTTP ${error.statusCode}):`, error.message);
  } else if (error instanceof RateLimitError) {
    // 429 or 430 — rate limit / concurrent-request limit (never auto-retried)
    console.error(`Rate limited. Retry after: ${error.retryAfter}`);
  } else if (error instanceof APIStatusError) {
    console.error(`API error (HTTP ${error.statusCode}):`, error.message);
  } else if (error instanceof APIConnectionError) {
    console.error('Connection failed:', error.message);
  }
}

Inspecting Error Details

All APIStatusError instances (including PermissionDeniedError and RateLimitError) expose the full error response body. Most input-validation problems on the Python SDK (e.g. max_results out of range, empty query) are caught client-side by Pydantic and raise pydantic.ValidationError before any request is sent; the pattern below applies when the server rejects a request. browse.fetch is a convenient way to trigger one — a missing or filtered URL surfaces as a structured error:

# Python
from webiq import WebIQClient
from webiq.errors import APIStatusError

client = WebIQClient(api_key="your-api-key")

try:
    response = client.browse.fetch("https://www.not-microsoft.com")
except APIStatusError as e:
    print(f"Status: {e.status_code}")                          # e.g. 404
    print(f"Message: {e}")                                      # e.g. "No result is found"
    if isinstance(e.body, dict):
        print(f"Error code: {e.body.get('errorCode')}")        # e.g. "BrowseApiDocNotFound"
        print(f"Category: {e.body.get('errorCategory')}")      # e.g. "UserError"
        print(f"Details: {e.body.get('technicalDetails')}")    # e.g. "NotFound"
        print(f"Trace ID: {e.body.get('traceId')}")
        print(f"Retry after: {e.body.get('retryAfter')}")

// TypeScript
import { WebIQClient, APIStatusError } from '@microsoft/webiq';

async function main() {
  const client = new WebIQClient({ apiKey: 'your-api-key' });

  try {
    await client.browse.fetch('https://www.not-microsoft.com');
  } catch (error) {
    if (error instanceof APIStatusError) {
      console.error(`Status: ${error.statusCode}`); // e.g. 404
      console.error(`Message: ${error.message}`); // e.g. "No result is found"

      // Full error body from the API response
      if (error.body && typeof error.body === 'object') {
        const body = error.body as Record<string, unknown>;
        console.error(`Error code: ${body.errorCode}`); // e.g. "BrowseApiDocNotFound"
        console.error(`Category: ${body.errorCategory}`); // e.g. "UserError"
        console.error(`Details: ${body.technicalDetails}`); // e.g. "NotFound"
        console.error(`Trace ID: ${body.traceId}`); // for debugging with support
        console.error(`Retry after: ${body.retryAfter}`); // for retryable errors
      }
    }
  }
}

main().catch(console.error);

Debugging Tips

  • traceId — every response includes a traceId field for end-to-end debugging. Include it when contacting support.
  • Telemetry hook — attach a callback to log every request attempt, status code, and timing.
  • Error body — APIStatusError.body contains the full server response, including errorCode, errorCategory, technicalDetails, and traceId.
  • Last updated on