LlamaIndex:产品、工程与评测·· 2026-05-26AI 评分45
LlamaIndex 周报:新增 HEIC 解析、延迟指标及 Google Agents API 集成
LlamaIndex Newsletter 5-26-26
AI 导读
LlamaIndex 发布本周通讯,主要更新包括 LlamaParse 原生支持 HEIC 格式文件解析、上线包含队列与处理时间的完整延迟指标监控,以及提供模板以集成 Google 新的 Agents API。此外还展示了基于 LiteParse 的金融尽职调查 Agent 案例,并提及在 AI Engineer Singapore 和 Google I/O 的活动参与情况。
正文 · 原文
Hi there, Llama Enthusiasts! 🦙
Welcome to this week's edition of the LlamaIndex newsletter! We're excited to share major updates including native HEIC file parsing, comprehensive latency metrics, and our exciting integration with Google's new Agents API. Plus, we've been busy at conferences and building powerful financial analysis tools. Check out these developments along with our upcoming webinar on ParseBench!
🎯 Last Call: ParseBench Webinar
- Don't miss our upcoming webinar on ParseBench - the first document OCR benchmark designed specifically for AI agents. We'll reveal all the magic behind how to know if your document parser is truly production-ready. Register now before it's too late!
☁️ LlamaParse Feature Highlights
- Native HEIC File Support: LlamaParse now parses Apple's default HEIC image format directly - no more converting to JPEG first! Perfect for enterprise file systems with iPhone photos of whiteboards, scanned docs, and receipts.
- Latency Metrics Now Live: Get complete latency breakdowns for every Parse, Extract, and Classify job with queue time, processing time, and total latency metrics. New Metrics tab includes scatter plots and histograms to analyze usage patterns. Check it out in Parse History
🤩 Get Building
- Google Agents API Integration: We built a template that gives Google's new sandboxed agent environment access to LlamaParse and LiteParse for autonomous document processing. Configure, clone, install, and watch agents work with messy real-world documents automatically. GitHub Repository, Get Started with LlamaParse
- Financial Due Diligence Agent: Built a demo agent that ingests SEC filings and answers questions with exact citations highlighted on original PDF pages. About 600 lines of Next.js, no vector DB, just LiteParse. Blog Post
- AI Engineer Singapore Wrap-Up: Thanks to all the developers who joined our workshop, keynote, and executive dinner in Singapore! See you at the world fair in SF in a few weeks. Here's a copy of the slides from our presentation.
- Google I/O Presence: We were featured in the Google I/O developer keynote! Exciting new features coming to the Gemini API, and we're thrilled to provide document infrastructure for Google ecosystem builders.
来源:LlamaIndex:产品、工程与评测 · llamaindex.ai