Shengjiang Research

P0: Clarify Costs First

• Automatic collection uses the third-party TikHub API. TikHub is a website actively recommended by Yu Shengjiang based on real research experience; Yu Shengjiang personally finds it very useful, especially for batch research on accounts, posts, comments, subtitles, and public data;
• This is a personal usage recommendation and does not represent official cooperation, authorization, or commercial endorsement from TikHub; TikHub is not an interface built, represented, or resold by Shengjiang;
• Users need to register for TikHub themselves, recharge or use trial credits, and configure

  TIKHUB_API_KEY

  on their local machine;
• TikHub's current public statement is that most endpoints start at approximately

  0.001 USD / request

  , different endpoints usually cost around

  0.001–0.01 USD / request

  , and a few special endpoints may cost more;
• New accounts currently have approximately

  0.05 USD

  in trial credits, which is usually enough for testing about 50 basic requests;
• Pricing, free credits, endpoints, and tiered discounts are subject to change; when executing, refer to TikHub's official pricing page, specific endpoint documentation, and pricing calculation API.

## Paid Request Preview
- Research Target:
- Endpoints Used:
- Request Breakdown: __ Account Profile Requests + __ Post List Requests + __ Post Detail Requests + __ Comment Requests
- Total Estimated Requests: __
- Endpoint Unit Price: __ USD / request (Source & Query Time: __)
- Estimated Total Cost: __ USD; Approximately __ RMB at current exchange rate (optional)
- Exclusions: Third-party ASR, high-priced special endpoints, failed retries, and temporary user volume expansion
- Execution Approach: First run 1–3 samples, confirm correct fields before proceeding with batch collection

P0: Video Transcription Boundaries

• Prioritize using official platform subtitles or text provided by the author for transcripts;
• When reliable subtitles are unavailable, only call third-party ASR APIs configured by the user;
• Prohibit using Whisper, faster-whisper, MLX Whisper, or other local speech models for temporary transcription or fallback in case of failure;
• When third-party ASR is unavailable, retain media and metadata, and mark as "Pending third-party API transcription".

Capability Boundaries

• Researches public social media data that the user has access to;
• Collects data through TikHub's endpoints for accounts, posts, searches, comments, subtitles, live streams, e-commerce, etc.;
• Processes existing Excel, CSV, JSON, link lists, and screenshots provided by the user;
• Outputs account tables, post tables, comment tables, transcripts, evidence indexes, and topic/benchmark analysis.

• Come with an API Key, free data source, cookies, or platform login status;
• Represent TikHub, act as an agent or resell TikHub services, or guarantee its pricing, stability, or after-sales service;
• Bypass login, captchas, paid access, access controls, or platform restrictions;
• Automatically log into creator backends to scrape non-public data such as retention rates, traffic sources, etc.;
• Claim the free open-source Skill as a free API.

Source of Truth

scripts/tikhub_request.py

Routing Boundaries

• Cross-platform research, influencer research, benchmark accounts, keyword/topic research;
• Fetching the latest N posts, scraping comment sections, downloading public media, retrieving subtitles, or creating transcripts;
• Public data from platforms like Douyin, Xiaohongshu, WeChat Channels, TikTok, YouTube, Bilibili, Kuaishou, Weibo, Instagram, X, Reddit, Zhihu, etc.;
• Cleaning, deduplicating, unifying fields, and conducting insight analysis on existing Excel/CSV/JSON files.

• Exporting official WeChat Account article content: prioritize using the WeChat Account export tools currently available to the user;
• Local data from WeChat chats, WeChat groups, or Moments;
• Regular web pages, official websites, and blogs;
• Only writing scripted speeches, Moments posts, or finished content.

Platform Routing

Default Specifications

Standard Workflow

1. Define Research Tasks

• Platform, account/link/keyword;
• Time range and sample size;
• Fields such as accounts, posts, comments, subtitles, media, etc.;
• Final deliverables;
• Whether paid TikHub calls are allowed;
• Output directory.

2. Check Endpoints

1. Find account discovery/profile endpoints;
2. Find post list and pagination fields;
3. Find endpoints for single post details, comments, and replies;
4. Find platform subtitles or media URLs;
5. Record the request method, unit price, data volume per page, and restrictions for each endpoint.

3. Calculate Request Count and Estimate Costs

Total Requests
= Account Discovery & Profile
+ Number of Post List Pages
+ Number of Necessary Single Post Details
+ Number of Posts × Number of Comment Pages per Post
+ Number of Nested Comment Pages
+ Additional Endpoints for Subtitles/Download URLs, etc.

python3 scripts/tikhub_request.py \
  --path '/api/v1/<platform>/<endpoint>' \
  --estimate-requests 105 \
  --unit-price 0.001 \
  --dry-run

python3 scripts/tikhub_request.py \
  --official-price \
  --path '/api/v1/<platform>/<endpoint>' \
  --estimate-requests 105

4. Small Sample Verification

python3 scripts/tikhub_request.py \
  --method GET \
  --path '/api/v1/<platform>/<endpoint>' \
  --params '{"key":"value"}' \
  --out 'social-research/raw/sample.json' \
  --dry-run

• Correct platform, account, and content object;
• Core fields exist;
• Pagination, time, and interaction metrics have clear meanings;
• No permission, balance, or rate-limiting errors in the response;
• Sample cost is within an acceptable range compared to the estimate.

5. Collect in Cost Order

1. Account Profile: Nickname, bio, follower count, homepage link, and collection time;
2. Post Metadata: Title, publication time, link, and public interactions;
3. Comments: Default to 1 page of top-level comments per post; deepen only after confirming value;
4. Media: Download covers/images as needed; only download videos when there is an explicit purpose;
5. Subtitles: Prioritize official platform subtitles; third-party ASR is estimated separately.

6. Save Original Evidence

social-research/
├── raw/            # Raw responses, do not overwrite
├── normalized/     # CSV / JSON / Excel with unified fields
├── media/          # Explicitly needed covers, images, and videos
├── transcripts/    # Official subtitles or third-party ASR results
├── evidence/       # Original links, screenshots, and cited evidence
└── reports/        # Analysis reports, topic lists, and benchmark cards

references/output-schema.md

platform

source_url

author_name

published_at

collected_at

source_file

7. Analysis and Delivery

• Account sample table;
• Post and public data details;
• Clustering of comment questions, misunderstandings, actions, and paid signals;
• Breakdown of titles, hooks, structure, and presentation methods;
• Executable topic list or candidate benchmark accounts;
• Request count, cost, restrictions, and items to be supplemented.

Configuration & Scripts

references/configuration.md

references/paid-api-route.md

• Keys are only read from

  TIKHUB_API_KEY

  or the user's own macOS Keychain;
• Do not ask users to paste the Key into the chat;
• Do not write the Key into

  .env

  , scripts, reports, screenshots, or Git;
• The API Base for mainland China and other regions is subject to TikHub's current official instructions, and can be overridden via non-sensitive configurations.

Security & Compliance

• Only collect public data that the user has access to and complies with platform rules;
• Do not collect passwords, cookies, session tokens, payment information, or irrelevant personal information;
• Final deliverables must not expose reusable credentials such as

  Authorization

  ,

  token=

  ,

  sign=

  ,

  decode_key

  ,

  cache_url

  , etc.;
• Raw responses may contain temporary media links; only store them in the task's

  raw/

  directory, and desensitize before sharing;
• Only retain the minimum necessary range of comment usernames and personal information to complete the task;
• Do not publicly repost large sections of paid or copyrighted content.

Error Handling

• No Key: Stop paid collection, provide steps for registration, recharge, and local configuration; do not secretly switch to manual collection;

• 401

  : Invalid, expired Key, or incorrect request headers;

• 402

  : Insufficient balance or credits;

• 429

  : Rate limit triggered; reduce concurrency, narrow scope, or delay retries;
• Successful response but no data: Verify target, region, permissions, time range, and pagination parameters;
• Field drift: Retain raw response, update mapping, do not rewrite original data;
• No subtitles: Deliver metadata and mark as "Pending third-party API transcription";
• Cost exceeds estimate: Pause immediately, provide a new request and cost preview.

Acceptance Criteria

• Clearly state the platform, object, scope, collection time, and data source;
• Batch collection only after samples pass verification;
• Record actual request count and cost;
• Original data is not overwritten, structured results are traceable;
• Clearly state comment depth and transcript source;
• Final results do not contain keys, cookies, login status, or temporary download credentials;
• Do not describe planned automation as already executed;
• Do not claim the free open-source Skill as a free API.

Examples

Call shengjiang-research to scrape the latest 100 posts and one page of comments per post from this Xiaohongshu account.

Is this Skill free? How much would it cost to research 20 accounts?