Zhihu · Comments
Replies to comments
Send comment IDs; get all replies under each comment with text, likes, time, IP location and the replier's profile.
POST/v1/zhihu/replies
$0.00359per record
- 1,000 records
- ≈ $3.59
- minimum per page of 3 records
- $0.00125
Works with trial balancePages with a cursor
Send it
POST with a JSON body, or GET with the same parameters in the query string. Your key goes in the Authorization header.
curl -X POST "https://api.silklinedata.com/v1/zhihu/replies" \
-H "Authorization: Bearer $SILKLINE_KEY" \
-H "Content-Type: application/json" \
-d '{"limit":5,"commentIds":["10655110604"]}'import os, requests
response = requests.post(
"https://api.silklinedata.com/v1/zhihu/replies",
headers={"Authorization": "Bearer " + os.environ["SILKLINE_KEY"]},
json={
"limit": 5,
"commentIds": ["10655110604"],
},
)
answer = response.json()
print(answer["data"], answer["meta"]["charged_usd"])const response = await fetch("https://api.silklinedata.com/v1/zhihu/replies", {
method: "POST",
headers: { Authorization: `Bearer ${process.env.SILKLINE_KEY}`, "Content-Type": "application/json" },
body: JSON.stringify({
"limit": 5,
"commentIds": [
"10655110604"
]
}),
});
const { data, cursor, meta } = await response.json();
console.log(data, meta.charged_usd);What comes back
Records from a real request on 2026-11-21. Names, avatars and profile links of ordinary users show as <masked>; long text is cut here, not in the API.
{
"data": [
{
"type": "comment",
"commentId": "10655838129",
"answerId": null,
"parentCommentId": "10655110604",
"replyToCommentId": null,
"text": "有gc的语言没法直接利用析构函数,像c++…",
"likeCount": 43,
"replyCount": 0,
"isPinned": false,
"isByContentAuthor": "<masked>",
"createdAt": "2023-08-20T14:46:21Z",
"author": "<masked>",
"authorId": "<masked>",
"authorUrlToken": "<mas…",
"authorAvatar": "<masked>",
"authorHeadline": "<masked>",
"authorGender": null,
"authorFollowerCount": null,
"authorVerification": null,
"authorIpLocation": "<masked>"
},
{
"type": "comment",
"commentId": "10657441511",
"answerId": null,
"parentCommentId": "10655110604",
"replyToCommentId": "<masked>",
"text": "我猜下一步就是该跟你杠,敲到哪格式化…",
"likeCount": 23,
"replyCount": 0,
"isPinned": false,
"isByContentAuthor": "<masked>",
"createdAt": "2023-08-22T17:03:54Z",
"author": "<masked>",
"authorId": "<masked>",
"authorUrlToken": "<mas…",
"authorAvatar": "<masked>",
"authorHeadline": "<masked>",
"authorGender": "<masked>",
"authorFollowerCount": null,
"authorVerification": null,
"authorIpLocation": "<masked>"
}
],
"cursor": "c1.Hq8vR2mTn5wK.yJ0dXA…",
"meta": {
"request_id": "req_01JD8Q…",
"charged_usd": "0.007180",
"balance_usd": "9.875000",
"items": 2
}
}Fields 20
| Field | Type | Example value |
|---|---|---|
| type | string | comment |
| commentId | string | 10655838129 |
| answerId | null | null |
| parentCommentId | string | 10655110604 |
| replyToCommentId | null | null |
| text | string | 有gc的语言没法直接利用析构函数,像c++那样利用RAII,不用with用什么?Java,c#不也都有类似的东西吗? |
| likeCount | number | 43 |
| replyCount | number | 0 |
| isPinned | boolean | false |
| isByContentAuthor | string | <masked> |
| createdAt | string | 2023-08-20T14:46:21Z |
| author | string | <masked> |
| authorId | string | <masked> |
| authorUrlToken | string | <masked> |
| authorAvatar | string | <masked> |
| authorHeadline | string | <masked> |
| authorGender | null | null |
| authorFollowerCount | null | null |
| authorVerification | null | null |
| authorIpLocation | string | <masked> |
Parameters 2
| Name | Type | Required | Description | Example |
|---|---|---|---|---|
| limit | integer | no | How many records to return, 1 to 100. A request never returns more than this. | 5 |
| commentIds | array | yes | Comment IDs (the commentId of a comment) to fetch all replies of. | ["10655110604"] |
Paging
Send the cursor of the last answer back with the same parameters to get the next page. Each page is a request of its own, charged on its own; the last page has cursor null.
const params = {"commentIds":["10655110604"]};
const records = [];
let cursor = null;
do {
const response = await fetch("https://api.silklinedata.com/v1/zhihu/replies", {
method: "POST",
headers: { Authorization: `Bearer ${process.env.SILKLINE_KEY}`, "Content-Type": "application/json" },
body: JSON.stringify({ ...params, limit: 100, ...(cursor ? { cursor } : {}) }),
});
const answer = await response.json();
records.push(...answer.data);
cursor = answer.cursor;
} while (cursor);
console.log(records.length);What a request costs
You pay per record; when a request returns few records or none, it is charged at most the endpoint minimum for each page it reads, possibly less, and errors or timeouts on our side are not charged.
Here a page holds up to 3 records and its minimum is $0.00125; a request for 100 records reads 34 pages.
- 1 record: up to $0.00359
- 20 records: up to $0.0718
- 100 records: up to $0.359
The response's meta says what was charged (charged_usd) and what is left (balance_usd). Errors on our side are not charged.
Questions
What parameters does Replies to comments take?
Required: commentIds. Every request also takes limit (1 to 100).
How many records does one request return?
Up to 100 (limit 1 to 100). A page holds up to 3 records, so 100 records take 34 pages in one request.
Can I page through more results?
Yes. While the answer carries a cursor, send it back with the same parameters for the next batch.
How much does it cost?
Per record: record $0.00359. The minimum is $0.00125 per page: with few records or none, a request pays at most that for each page it reads, possibly less. 100 records cost at most $0.359.
What does a request cost when it fails or comes back empty?
You pay per record; when a request returns few records or none, it is charged at most the endpoint minimum for each page it reads, possibly less, and errors or timeouts on our side are not charged. Each answer's meta says what the request was charged and the balance left.