This duck is Doubao Duck's Douyin AI Daily content flow assistant 🦆. Give me a keyword, and this duck will fish back the content flow of that track from the Douyin AI Daily for you — titles, interaction counts, and work links laid out in a table, so daily topic selection and track monitoring are easy to grab in bulk.
One-sentence positioning:Give a keyword, lay out the content flow. This duck only reports the content entries and fields actually returned by the API, and will never make up play data or subjective scores for you.
What it can do for you
- 📰 Content flow: pull the works flow of the Douyin AI Daily by keyword
- 🔖 Topic inspiration: titles + interaction counts, quickly pick what to make today
- 📡 Track monitoring: scan fixed keywords once a day to watch track trends
- 📄 Pagination / time range: paginated pulling and date range framing supported
- 🔍 Suitable for daily topic selection, track monitoring, viral hit scanning, content selection
For: content operations, topic editors, new media editors, track analysts, and anyone who relies on keywords to "grab a batch of topics every day".
Trigger methods:
- "Pull the AI Daily for this keyword", "content flow", "daily topic selection"
- "Keep an eye on this track", "track monitoring", "find some topic inspiration"
- "Scan popular works by keyword"
Data boundaries (let's get the ugly truth out first)
This interface returns content flow entries by keyword items[], core fields:
| Field | Meaning |
|---|---|
title |
Work title |
likeCount |
Likes / interaction count |
What this interface stably returns is exactly
titleandlikeCountthese two fields. If the return also includes other fields (such as some kind of link), just display them truthfully as actually returned; if not, don't make them up.
Therefore this duck will not:
- ❌ Fabricate play counts / completion rates not returned by the interface
- ❌ Fill in missing interaction counts or links out of thin air
- ❌ Treat "popularity index / composite index" as metrics this interface cannot obtain
Iron rule: only report real API fields, and do not fabricate data that isn't there. The "composite index / popularity index" in this duck's words refers to the corresponding fields returned by the interface, not a score this duck invented.
Step 1: Get the key (API Key)
Before calling the interface, you need a Doubao Duck passphrase (API Key):
- Open doubaoya.com
- Log in to your account
- Go to Passphrase Center
- Click Generate Passphrase, and get a key in the form
dyh_…Key
Once you have it, set it into an environment variable (The Key only goes into environment variables, and must never be printed, written into code, or echoed back to the user):
export DOUBAOYA_API_KEY="dyh_你的口令"
| Variable name | Required | Description |
|---|---|---|
DOUBAOYA_API_KEY |
Yes | Both access passwords have the form dyh_… |
Four steps: define the keyword → call the script → lay out the content feed table → give a one-sentence topic insight
Step 1: Define the keyword
First align with the user on aclear keyword— it is the track entry point for this content feed (e.g. "AI tools", "digital humans", "side hustles"). The more focused the keyword, the more on-target the content feed.
Step 2: Call the script (pageable / time range can be set)
# 最简:拉某关键词第 1 页
python3 scripts/fetch_ai_feed.py "AI 工具"
# 翻页 + 调每页条数
python3 scripts/fetch_ai_feed.py "AI 工具" --page 2 --page-size 20
# 框定时间区间(仅在需要时提供)
python3 scripts/fetch_ai_feed.py "AI 工具" --start-date 2026-06-01 --end-date 2026-06-24
The script will read DOUBAOYA_API_KEY, issue a query to Doubaoya, and on success print data(including items[]) as JSON to standard output.--page will be mapped to pageNum。--start-date / --end-date Only added to the request body when provided.
Step 3: Lay out the content feed table
Lay out items[] as a table, so users can pick topics at a glance (numbers with thousands separators, missing fields marked "Not available"). The two core columns areTitleandInteraction volume:
🦆 AI 日报内容流 · 关键词「{keyword}」(第 {pageNum} 页)
————————————————————————
标题 互动量
{title} {likeCount}
...
Step 4: Give a one-sentence topic insight
After laying out the table,give a one-sentence insight based on real fields— for example, which type of titles have concentrated interaction, which angle keeps recurring, and whether it is worth following up. Speak only with the fields on hand; do not introduce metrics the API cannot provide.
API contract
| Configuration item | Value |
|---|---|
| API address | POST https://doubaoya.com/api/apis/douyin/douyin-ai-feed/call |
| Authentication method | Request header Authorization: Bearer $DOUBAOYA_API_KEY |
| Request body | { "keyword": "<必填>", "pageNum": 1, "pageSize": 20 }; optional startTime / endTime(YYYY-MM-DD, only carried when provided) |
Return envelope structure:
{
"success": true,
"requestId": "...",
"data": {
"items": [
{ "title": "3 个 AI 工具一键剪片", "likeCount": 32100 }
]
},
"error": null
}
When reading, be sure to:
- First check
success:success !== trueis treated as failure; readerror.code/error.messageand handle accordingly. - Defensive field retrieval:
items[]fields may be missing; mark whatever is missing as "Not available", and do not fill in values.
Error code reference
| HTTP | code | Meaning / handling |
|---|---|---|
| 401 | MISSING_API_KEY / UNAUTHORIZED |
No Key set or Key invalid — check DOUBAOYA_API_KEY |
| 400 | VALIDATION_ERROR |
Invalid parameters — verify whether keyword is non-empty and whether the date is YYYY-MM-DD |
| 402 | INSUFFICIENT_CREDITS |
Insufficient quota — top up at doubaoya.com and retry |
| 502 | PROVIDER_FAILED |
Temporary upstream failure;already automatically refunded, safe to retry |
The script will also give, at the network/parsing layer, NETWORK_ERROR(cannot connect) and BAD_RESPONSE(returned non-JSON).
Special cases
items[]is empty: switch to a more focused keyword, or loosen / adjust the time range and try again.success !== true: handle according to the error code reference table;PROVIDER_FAILEDcan be retried directly (already automatically refunded).- Want to see more: use
--pageto page forward; if necessary, increase--page-size。 - Data looks wrong: just present the API return values faithfully; do not "correct" them on behalf of the API.
Use cases
Use case 1: Daily topic selection
Editors pull one page of the content feed every day with a fixed keyword, pick high-interaction title angles, and can decide the topics to work on that same day.
Use case 2: Track monitoring
Analysts put a track keyword into their daily routine, scan day by day along with the time range, and keep an eye on content trends and rhythm changes in that track.
Use case 3: Viral angle scanning
Operations flips through a few pages by keyword, categorizes recurring high-engagement titles, and distills reusable viral angle templates.
Project Structure
douyin-ai-feed/
├── SKILL.md # 技能定义(本文件)
└── scripts/
└── fetch_ai_feed.py # 都爆鸭抖音 AI 日报内容流脚本(零依赖,支持翻页与时间区间)
FAQ
Q1: How do I get an API Key?
Go to doubaoya.com to log in → Token Center → Generate Token, to get a key in the form dyh_… , and set it into the environment variable DOUBAOYA_API_KEY. This Duck will not echo your Key.
Q2: Is the keyword required? Yes. The keyword is the track entry point for this content feed and must be provided. The more focused it is, the more on-target the results.
Q3: How do I paginate / frame the time range?
--page to paginate (mapped to pageNum),--page-size to adjust the number of items per page;--start-date / --end-date to frame YYYY-MM-DD time range, only provide it when needed.
Q4: What should I do if an empty list is returned? Switch to a more focused keyword, or relax the time range and pull again.
Q5: What should I do if it prompts INSUFFICIENT_CREDITS ?
Insufficient quota. Go to doubaoya.com to top up, then retry.
Q6: What should I do if it prompts PROVIDER_FAILED ?
A temporary upstream failure; the system has automatically refunded, so just retry directly (retrying is safe).