启动猎人新闻:Captain (YC W26) – 文件自动化检索增强生成
Launch HN: Captain (YC W26) – Automated RAG for Files

原始链接: https://www.runcaptain.com/

## Captain Odyssey:快速部署RAG Captain Odyssey是一个新平台,旨在**快速构建和部署检索增强生成 (RAG) 管道**——从手动构建的约78%准确率提升到**几分钟内达到95%准确率**。它通过使用您的数据(由您托管或利用Captain的托管基础设施)来简化AI代理开发。 主要功能包括**通用索引**(自动OCR、文件转换、嵌入)、**托管向量存储**(无需外部数据库)和**代理/混合搜索**,以提高相关性。Captain与Azure、GCP、Amazon S3、SharePoint等流行的云服务集成。 Captain专为企业需求而设计,提供**细粒度、安全的基于角色的访问控制**,并已通过**SOC 2认证**。该平台采用API优先策略,旨在消除传统RAG实施中所需的时间和维护。Captain目前可用,未来将推出确定性AI等新功能。

## Captain:文件自动化RAG - 摘要 Captain (runcaptain.com) 是一项新服务,旨在简化为非结构化数据构建和维护检索增强生成 (RAG) 管道。它可自动化从云存储 (S3, GCS) 和 SaaS 来源 (Google Drive) 进行索引,处理文本提取、分块、嵌入和搜索等任务。演示“Ask PG’s Essays” (pg.runcaptain.com) 展示了它快速索引和查询 Paul Graham 著作的能力。 创始人强调了生产级 RAG 管道的复杂性,并指出维护,尤其是在大型文件集合的情况下,需要大量精力。Captain 旨在通过为索引和查询提供标准化的 API,抽象底层基础设施来解决这个问题。 主要功能包括支持各种数据类型(转换为 Markdown),利用 Gemini 3 Pro 和 Voyage 等模型进行嵌入和重新排序,以及提供确定性的页面引用。虽然承认 DIY RAG 解决方案和 Vertex File Search 等竞争对手的兴起,但 Captain 将自己定位为可扩展、托管的解决方案,适用于需要可靠、持续索引和检索的组织。定价从每月 295 美元开始,可索引高达 1000 页的新内容。
相关文章

原文
Just shipped: Captain Odyssey – Our Private Market Dataset Read More →
Avg. Accuracy 78% -> 95%Backed byY CombinatorCombinator

Ship enterpriseagentic searchin minutes

Power AI agents with your data or ours

Data Sources

Connect Your Existing Systems

Integrate quickly with your cloud services.

Azure BlobAzure Blob
GCP StorageGCP Storage
Amazon S3Amazon S3
SharePointSharePoint
Google DriveGoogle Drive
DropboxDropbox
1,000 Custom Options
ConfluenceConfluence
SlackSlack
GmailGmail
NotionNotion
SharePoint
Upload from SharePoint
Bring the power of Microsoft 365 to Captain
S3/GCS/Azure
Index your Cloud
Connect S3 / GCS / Azure

Stop burning time on spotty RAG

Effortlessly ship standardized, fully-managed context pipelines instead.

captain
Building RAG manually
Universal IndexingAuto OCR + VLM, file conversions, best-in-class embeddings
Pre-Processing & OCR
Chunking Strategy
Embedding Model
Captain CollectionsManaged vector storage (no external database needed)
Vector Database
Agentic + Hybrid SearchWeighted search for keywords and semantic relevance
Query Embedding
Similarity Lookup
Re-Ranking
Prompt Engineering
+95% Accuracy

Deploy in minutes · Zero maintenance
~78% Accuracy

3-6 months · Scale and maintain

Built by engineers from

Boar's Head
Sony
IEEE
Reality Interactive
Purdue
Rocketbook

Granular and Secure.

Map Role-Based Access and pass SOC 2 requirements.

Role-Based Governance

Attach custom metadata to files at index time, then filter queries with granular operators to enforce role-based access across any collection.

Read the Docs
SOC 2 Type II Certified

SOC 2 Certified

Enterprise-grade infrastructure security;

Independently audited and pentested. Read our security report and compliance posture below:

Learn more

On the Radar

Find the latest news, updates, and stories on Captain.

Coming Soon

Unlocking Determinism from AI Randomness

Our system will take care of all the accuracy, all the indexing, all the overhead, and you can just throw in the files and ask it questions.

Lewis Polansky, CEO @ Captain

If data drives decisions, waiting isn't an option.

Command your data on captain

Join the AI movement. Ship production RAG in minutes.

联系我们 contact @ memedata.com