Priya Natarajan
Head of AI Research at Solstice Labs
We just open-sourced our evaluation harness for agentic AI systems. Three months of internal dogfooding baked in.
Open-sourcing our agent evaluation harness
solsticelabs.ai
1,842·96 comments
{"id": "7283456789012345678","url": "https://www.linkedin.com/feed/update/urn:li:activity:7283456789012345678","text": "We just open-sourced our evaluation harness for agentic AI systems. Three months of internal dogfooding baked in.","createdUtc": 1758931200,"author": {...},"engagement": {...},"attachmentType": "article","attachmentTitle": "Open-sourcing our agent evaluation harness",... 还有 8 个字段}
价格与服务商
LinkedIn 搜索帖子（完整版） API 价格：一个 API，4 个可互换的服务商
最便宜的服务商优先响应。如果它失败，下一个会在同一次调用中接手。
每次请求封顶 US$0.1321 条结果US$0.0013250 条结果US$0.066100 条结果US$0.132（封顶）按实际返回的内容计费。调低返回数量上限即可少付，而且无论返回多少，单次请求都不会超过封顶价。4Honeybee备用 3可用率 96.85%p50 12992ms占调用的 2.5%2.5%2,246 次调用US$1.94/千条结果每次请求最多 US$0.194计量方式：每条结果，按实际返回基础费 US$0.00116 + 每条结果 US$0.00193
每次请求封顶 US$0.1941 条结果US$0.0030950 条结果US$0.09766100 条结果US$0.194（封顶）按实际返回的内容计费。调低返回数量上限即可少付，而且无论返回多少，单次请求都不会超过封顶价。
试一试
一个请求获取 LinkedIn 搜索帖子（完整版）
这个示例展示了 35 个字段中的 34 个。查看全部字段
{
"found": true,
"data": {
"posts": [
{
"id": "7283456789012345678",
"url": "https://www.linkedin.com/feed/update/urn:li:activity:7283456789012345678",
"text": "We just open-sourced our evaluation harness for agentic AI systems. Three months of internal dogfooding baked in.",
"createdUtc": 1758931200,
"author": {
"id": "ACoAAAbcdef1234",
"name": "Priya Natarajan",
"headline": "Head of AI Research at Solstice Labs",
"profileUrl": "https://www.linkedin.com/in/priya-natarajan-ai",
"image": "https://media.licdn.com/dms/image/D4E03AQPriyaProfile/profile-displayphoto-shrink_800_800/0/1701000000004?e=1735689600",
"publicIdentifier": "priya-natarajan-ai",
"type": "person"
},
"engagement": {
"reactions": 1842,
"comments": 96,
"reposts": 213,
"breakdown": [
{
"type": "LIKE",
"count": 1490
},
{
"type": "PRAISE",
"count": 220
},
{
"type": "INTEREST",
"count": 132
}
]
},
"attachmentType": "article",
"attachmentTitle": "Open-sourcing our agent evaluation harness",
"attachmentSubtitle": "solsticelabs.ai",
"attachmentUrl": "https://solsticelabs.ai/blog/agent-eval-harness",
"attachmentDescription": "A look at how we test multi-step agents before they reach production.",
"attachmentImage": "https://media.licdn.com/dms/image/D4E0BAQSolsticeArticle/feedshare-shrink_800/0/1701000000005?e=1735689600"
},
{
"id": "7283109876543210987",
"url": "https://www.linkedin.com/feed/update/urn:li:activity:7283109876543210987",
"text": "",
"createdUtc": 1758758400,
"author": {
"id": "urn:li:organization:9182734",
"name": "Cortex Analytics",
"headline": "48,910 followers",
"profileUrl": "https://www.linkedin.com/company/cortex-analytics",
"image": "https://media.licdn.com/dms/image/D560BAQCortexLogo/company-logo_200_200/0/1701000000006?e=1735689600",
"publicIdentifier": "cortex-analytics",
"type": "company"
},
"engagement": {
"reactions": 340,
"comments": 12,
"reposts": 28,
"breakdown": [
{
"type": "LIKE",
"count": 310
},
{
"type": "SUPPORT",
"count": 30
}
]
},
"attachmentType": "poll",
"pollQuestion": "Which part of the ML lifecycle wastes your team the most time?",
"pollTotalVotes": 612,
"pollClosed": false,
"pollOptions": [
{
"text": "Labeling data",
"votes": 201
},
{
"text": "Feature engineering",
"votes": 158
},
{
"text": "Eval and monitoring",
"votes": 253
}
]
}
]
}
}完整的参数与响应参考：这个接口的每个字段、类型和示例。
参考
一个请求，一种响应结构。
你发送的每个字段和收到的每个字段，都附有真实示例。
最近验证于 2026-11-14 · 可用率和延迟按 30d 统计
请求体
JSON，POST 到这个接口。
- querystring
artificial intelligence
LinkedIn 帖子搜索查询，可包含 LinkedIn 搜索支持的带引号词语或布尔运算符。
- authorCompanyNamesarray
只返回与这些公司名称相关的人员发布的帖子。
- authorIndustryIdsarray
只返回与这些 LinkedIn 行业 ID 相关的作者发布的帖子。
- authorKeywordsstring
只返回作者主页头衔或职位名称中至少包含其中一个关键词的帖子。
- authorUrlsarray
只返回由这些 LinkedIn 主页或公司 URL 发布的帖子。
- contentTypeenum
只返回带有此内容类型的帖子。
- datePostedenum
last-week
只返回在此相对时间范围内发布的帖子。过去一小时以及超过一个月的时间范围会路由到支持它们的服务商。
- limitinteger
10
最多返回的帖子数（1-100，默认 10）。上限为一页 LinkedIn 搜索结果。
- mentioningMemberUrlsarray
只返回提及这些 LinkedIn 会员主页 URL 的帖子。
- requireFieldsarray
可选；省略时路由不变，由最便宜的数据源处理。列出本次请求必须能返回的输出字段，例如 `attachmentType` 或 `pollOptions`，请求就只会由能返回全部这些字段的数据源处理。未列出的字段，只要处理请求的数据源有，仍会照常返回。这可能提高你的价格：最便宜的数据源无法返回某个列出的字段时，会改由更贵的数据源处理，并按它的价格报价和扣费。如果某条帖子确实没有某个列出的字段，该字段仍可能缺失。如果列出的字段组合没有任何单一数据源能同时返回，请求会作为无效输入被拒绝，不扣费。
- sortenum
relevance
按搜索相关性或发布日期对结果排序。
响应
JSON，每个字段一个示例值。
LinkedIn posts matching the search query. Populated whenever the provider has data for the entity.
- attachmentDescriptionstring可设为必填
A look at how we test multi-step agents before they reach production.
Short description of the attached content, as LinkedIn shows it on the preview card. Best effort on article attachments: depending on the search it may be absent, so treat it as optional. To read an article's content reliably, pass attachmentUrl to the linkedin.article endpoint.
- attachmentImagestring可设为必填
https://media.licdn.com/dms/image/D4E0BAQSolsticeArticle/feedshare-shrink_800/0/1701000000005?e=1735689600
Image URL associated with the attachment.
- attachmentSubtitlestring可设为必填
solsticelabs.ai
Subtitle of the attached structured content.
- attachmentTitlestring可设为必填
Open-sourcing our agent evaluation harness
Title of the attached article, job, or other structured content.
- attachmentTypestring可设为必填
article
Kind of structured content attached to the post, such as article, job, poll, video, or document.
- attachmentUrlstring可设为必填
https://solsticelabs.ai/blog/agent-eval-harness
Destination URL of the attachment with tracking parameters removed; a video keeps the signed query it needs to play, and a YouTube link keeps its video id. When attachmentType is article, pass this URL to the linkedin.article endpoint to read the article's full body text.
- createdUtcnumber
1758931200
UTC epoch timestamp in seconds (Unix time). Multiply by 1000 for a JS Date in milliseconds. Populated whenever the provider has data for the entity.
- idstring
7283456789012345678
Stable LinkedIn activity identifier for the post. Populated whenever the provider has data for the entity.
- pollClosedboolean可设为必填
Whether the attached poll is closed.
- pollQuestionstring可设为必填
Question asked by an attached LinkedIn poll.
- pollTotalVotesinteger可设为必填
Total votes recorded by the attached poll.
- textstring
We just open-sourced our evaluation harness for agentic AI systems. Three months of internal dogfooding baked in.
Text content of the post. Empty string when the post carries no commentary of its own, such as a bare article share or an image-only post. Populated whenever the provider has data for the entity.
- urlstring
https://www.linkedin.com/feed/update/urn:li:activity:7283456789012345678
Canonical URL of the LinkedIn post. Populated whenever the provider has data for the entity.
Public author identity attached to the post. Populated whenever the provider has data for the entity.
- headlinestring
Head of AI Research at Solstice Labs
Public headline or company follower summary for the author.
- idstring
ACoAAAbcdef1234
LinkedIn identifier for the author.
- imagestring
https://media.licdn.com/dms/image/D4E03AQPriyaProfile/profile-displayphoto-shrink_800_800/0/1701000000004?e=1735689600
Public profile or company image URL for the author.
- namestring
Priya Natarajan
Display name of the author.
- profileUrlstring
https://www.linkedin.com/in/priya-natarajan-ai
Canonical LinkedIn profile or company URL for the author.
- publicIdentifierstring
priya-natarajan-ai
LinkedIn public identifier (vanity slug) for the author, for example "williamhgates". Stable across lanes, unlike the raw id.
- type"person" | "company" | "showcase"
person
Kind of LinkedIn entity that authored the post, derived from the author URL.
Reaction counts grouped by LinkedIn reaction type.
- countinteger
1490
Number of reactions of this type.
- typestring
LIKE
LinkedIn reaction type.
Public engagement totals attached to the post. Populated whenever the provider has data for the entity.
- commentsinteger
96
Total comment count.
- reactionsinteger
1842
Total reaction count.
- repostsinteger
213
Total repost or share count.
Options and public vote counts for an attached poll.
- textstring
Poll option text.
- votesinteger
Votes recorded for this option.
- foundboolean
true
Whether matching LinkedIn posts were found.
- reason"not_found"
Present only when `found` is false, and says why there is no result. `not_found`: the source states the target does not exist, or returned nothing for it. A `found: false` answer is a successful call, not an error, and `costUsd` is what it actually cost.
你在出售这个 API？查看你的接口必须返回的完整数据结构。
常见问题
关于 LinkedIn 搜索帖子（完整版） API
AnyAPI 的 LinkedIn 搜索帖子（完整版） API 只需一次 POST 请求 /v1/run/linkedin.search_posts_full，就以统一格式的 JSON 返回 LinkedIn 数据。搜索公开的 LinkedIn 帖子，附带丰富的作者、互动、附件和投票详情。AnyAPI 会在 4 个数据源之间路由每个请求，某个数据源失败时自动切换。费用：每千次请求 US$4 起，以美元计价，无需订阅，也没有每月最低消费。过去 30 天，通过 AnyAPI 发起的 LinkedIn 搜索帖子（完整版） API 调用中有 99.9% 成功，响应时间中位数为 3.6 秒，共测量 91,115 次调用。