> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.beatapi.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.beatapi.io/_mcp/server.

# How can I turn WeChat search results into useful café-opening references?

> Define the reference task, separate articles from videos, preserve search positions and turn relevant results into a sourced reading list.

To collect café-opening references from WeChat, search “咖啡店开业” (coffee shop opening) with `wechat_search.v2.fetch_search` for articles and accounts, and with `wechat_search.v2.fetch_search_videos` for short videos. Use articles to see how cafés present the location and the offer terms, and use videos to see how cafés show the visit. If you already know which format you need, search that `business_type` directly. Deliver a reading list with the original links, the reason each example made the list and what still needs checking.

## 1. Define the reference task

Write down the café concept, the questions and the search scope. For example: inspect up to two pages from the past week, and look for clear address and opening-hours information, offer conditions and visit experiences.

A one-week window shows recent examples. If the one-week results are sparse, run a separate search with a broader time window or another keyword, and record the change. If the brief needs older cases, choose the longer window from the start.

## 2. Discover material, then choose a content type

A universal search (`"business_type": "all"`) finds several content types in one request:

```json
{
  "action": "wechat_search.v2.fetch_search",
  "params": {
    "keyword": "咖啡店开业",
    "business_type": "all",
    "sort": "latest",
    "publish_time": "week",
    "offset": 0,
    "raw": false
  }
}
```

Send the request body to `POST https://api.beatapi.io/v1/social-data/call` with standard BeatAPI Bearer authentication. Save the request, the retrieval time, the original response and the request ID, then check which material types and sources the response actually contains.

When the task needs only articles, set `business_type:"article"` directly; an article task does not need a universal search first. Set `video` for videos or `account` to find accounts.

Choose the sort order by purpose: `latest` finds newer publications, `default` finds relevant candidates, and `hot` orders candidates by likes. The time options are `day`, `week`, `half_year` and `all`. Search order does not show whether an opening campaign succeeded as a business.

## 3. Add duration for short-video references

For examples of presenting a café visit in a few minutes, use the dedicated video search `wechat_search.v2.fetch_search_videos`:

```json
{
  "action": "wechat_search.v2.fetch_search_videos",
  "params": {
    "keyword": "咖啡店开业",
    "duration": "short",
    "sort": "default",
    "publish_time": "week",
    "offset": 0,
    "raw": false
  }
}
```

`short` means under five minutes. The other options are `medium` (5–10 minutes), `long` (20 minutes or longer) and `all`. For another range such as 10–20 minutes, use `all` and check the actual durations. The dedicated video search and the universal search keep separate request histories and separate continuation positions.

## 4. Continue without repeating the first page

Start with `offset:0` and no continuation cursor. When more results remain, copy the previous response’s `cursor` unchanged into `params.cursor` of the next request to the same action. Keep the keyword, content type, sort and time unchanged; for video search, also keep the duration. Use the returned `continue_flag` to decide whether to continue.

Increasing `offset` alone repeats the first page. Stop when you reach the planned page limit, when no results remain or when the cursor stops advancing, and record the stop reason. To resume, start from the saved position of the same query; a changed keyword or filter starts a separate query.

Saved article-only responses can go through the [article search archive workflow](/faq/wechat-search-articles); the workflow's local processor keeps the page, position and source references. Save universal and video responses separately, because the article-specific processor only accepts article search input.

## 5. Turn result cards into a reference list

The table below lists the columns of your working table. Fill the columns from the actual material. When a link, time or author identifier is missing, keep the item as open and traceable to the original response.

| Column                             | Use in this task                                                            |
| ---------------------------------- | --------------------------------------------------------------------------- |
| Material type and source           | Article, video or account; confirmed link/ID and original-response position |
| Title or caption, account and time | Preserve available information and check namesakes or republication         |
| Selection reason                   | For example, “offer conditions look clear; read the full explanation”       |
| Material actually read             | Search card, article body or other material obtained                        |
| Next step                          | Read the article, inspect video material, keep as a reference or exclude    |

An article may show how to structure information, and a video may show how to present a visit; give a separate reason for each. If you read only a search caption, record what the caption says, for example “the caption mentions a soft opening.” To discuss shots and the actual visit, inspect the video itself.

Merge repeated hits for the same article, video or account into one candidate, and keep the queries and pages where the candidate appeared. When several accounts republish the same event, keep each source relationship; do not count repeated descriptions of one event as independent business cases.

## 6. Continue with the useful material

Read selected articles with [WeChat article reading](/guides/read-wechat-article). For ongoing research on one account, pass a confirmed account identifier to the [account article list](/guides/wechat-account-articles). For a known Channels account, continue with [search within a channel](/guides/wechat-channels-channel-search). A display name helps identify a source, but each action needs its own identifier, taken from the material that action accepts.

Deliver the search scope, the pages actually saved, the candidate list, the source responses and three to five specific content recommendations. Attach evidence to each recommendation and name the next material worth reading.

To hand the search to an assistant connected to BeatAPI, use this instruction:

> Find café-opening references from the past week, up to two pages. Look at address, hours and offer wording in articles and visit-related clues in videos. Preserve sources, selection reasons and material read. Retrieve more for the most relevant objects, then suggest concrete ideas for preparing our content.

Estimate costs with current [pricing](https://beatapi.io/pricing). Choose content types and page counts for the task, count any additional article reading, and reuse saved responses instead of searching again.

## How can I pair an existing WeChat Index series with search?

If your team has saved a WeChat Index series for café openings, record the term, the source dates, the time zone and the values. Compare equal windows with comparable coverage for the same term, and flag missing dates. For an unusually high day, find related content with the article or video search from steps 2 and 3, then verify publication dates and the event context.

The public BeatAPI WeChat Search actions retrieve content; no public action retrieves the daily WeChat Index. Use search results as research leads, and keep separate evidence for the index change and for the related event.

## Review a set of Official Account keywords each week

For a weekly tender brief, run `wechat_search.v2.fetch_search` separately for each term with `business_type:"article"`, `publish_time:"week"` and `sort:"latest"`. Start each round on the first page, and keep each query’s continuation state and the article links already seen. Merge repeated hits and keep the matching terms. Check the actual deadlines, regions and eligibility, and read the article body when a promising result needs clarification. After reviewing a first handoff, schedule the weekly runs in your own environment.

## Check a named brand across accounts, articles and short videos

If an editor asks which accounts belong to a named café and how the café's opening is being covered, keep three result sets. Start with one page of each set you need. Replace `<brand name>` with the exact brand name; if the name is common, add a branch or city in a separate search.

| What you need        | How to set up the first page                                                              | What to verify                                                                       |
| -------------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ |
| Brand accounts       | `fetch_search` with `business_type: "account"`, `publish_time: "all"`, `sort: "default"`  | Confirm type, source and identifier before deciding whether it belongs to the brand. |
| Recent coverage      | `fetch_search` with `business_type: "article"`, `publish_time: "week"`, `sort: "latest"`  | Check outlet, date, location and terms before selecting bodies to read.              |
| Short video examples | `fetch_search_videos` with `duration: "short"`, `publish_time: "week"`, `sort: "default"` | Keep the video link and ID. Check the source for what is actually shown.             |

Use `offset: 0` and `raw: false` in each first request. Set up the account request like this:

```json
{"action":"wechat_search.v2.fetch_search","params":{"keyword":"<brand name>","business_type":"account","sort":"default","publish_time":"all","offset":0,"raw":false}}
```

Save each request with the matching response and request ID. Each page is one request, so the first page of all three sets takes three calls. Continue a set only when that set's response signals more results and supplies a new cursor. The account set and the article set need separate cursors even though both use `fetch_search`.

Hand over three blocks: confirmed accounts, sourced coverage and videos worth opening. Label each set as reviewed, empty for the current filters or failed. A matching account name is a lead; run the [account checks](/faq/wechat-search) before attaching the account to the brand.