<?xml version="1.0" encoding="utf-8"?><?xml-stylesheet type="text/xsl" href="atom.xsl"?>
<feed xmlns="http://www.w3.org/2005/Atom">
    <id>https://your-docusaurus-site.example.com/blog</id>
    <title>Docs Blog</title>
    <updated>2026-05-01T00:00:00.000Z</updated>
    <generator>https://github.com/jpmonette/feed</generator>
    <link rel="alternate" href="https://your-docusaurus-site.example.com/blog"/>
    <subtitle>Docs Blog</subtitle>
    <icon>https://your-docusaurus-site.example.com/img/favicon.ico</icon>
    <entry>
        <title type="html"><![CDATA[PDF OCR pipeline simplification — faster, cheaper, same API]]></title>
        <id>https://your-docusaurus-site.example.com/blog/2026/05/01/pdf-ocr-pipeline-simplification</id>
        <link href="https://your-docusaurus-site.example.com/blog/2026/05/01/pdf-ocr-pipeline-simplification"/>
        <updated>2026-05-01T00:00:00.000Z</updated>
        <summary type="html"><![CDATA[We rebuilt the PDF OCR backend on a leaner pipeline that processes pages directly from PDF, without an intermediate PNG conversion step. Same API, lower latency, and a cleaner cost story.]]></summary>
        <content type="html"><![CDATA[<p>We just shipped a backend overhaul of <strong>PDF OCR</strong>. The user-facing API is unchanged — your existing integrations keep
working — but everything behind it is leaner.</p>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="what-changed-and-what-didnt">What changed (and what didn't)<a href="https://your-docusaurus-site.example.com/blog/2026/05/01/pdf-ocr-pipeline-simplification#what-changed-and-what-didnt" class="hash-link" aria-label="Direct link to What changed (and what didn't)" title="Direct link to What changed (and what didn't)">​</a></h2>
<p><strong>Same</strong></p>
<ul>
<li>Endpoints, authentication, request and response shapes.</li>
<li>Webhook events and Svix signature verification.</li>
<li>Pricing and subscription tiers.</li>
</ul>
<p><strong>Faster behind the scenes</strong></p>
<ul>
<li>Pages now go straight from PDF to OCR. We dropped the intermediate PDF→PNG conversion that used to run in AWS Lambda
for every page.</li>
<li>Worker count for the OCR pipeline went from five to two, which removes hops, queues, and retry surface area.</li>
<li>Per-page latency in our internal benchmarks improved noticeably, and tail latency on bursty workloads is much
tighter now that fewer queues need to drain.</li>
</ul>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="why-we-did-it">Why we did it<a href="https://your-docusaurus-site.example.com/blog/2026/05/01/pdf-ocr-pipeline-simplification#why-we-did-it" class="hash-link" aria-label="Direct link to Why we did it" title="Direct link to Why we did it">​</a></h2>
<p>The old pipeline was three-staged: split the PDF into per-page PDFs, convert each page to PNG (Lambda), then OCR the
PNG. The PNG conversion cost real money and added a queue retry hop where, on a transient failure, we'd redo a step
that didn't need redoing.</p>
<p>The OCR model already accepts PDF inputs natively. Once we wired the page-OCR worker to send page PDFs directly, the
PNG conversion step became dead weight. Removing it eliminated:</p>
<ul>
<li>One Lambda function and one API Gateway.</li>
<li>Two Cloudflare Queues (and their dead-letter queues).</li>
<li>Three of the five OCR-pipeline workers.</li>
</ul>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="do-i-need-to-change-anything">Do I need to change anything?<a href="https://your-docusaurus-site.example.com/blog/2026/05/01/pdf-ocr-pipeline-simplification#do-i-need-to-change-anything" class="hash-link" aria-label="Direct link to Do I need to change anything?" title="Direct link to Do I need to change anything?">​</a></h2>
<p><strong>No.</strong> Same endpoints, same auth, same payloads.</p>
<p>If you want a slightly simpler upload path, we now also document the single-request multipart upload at
<code>https://api.commapdf.cobbling.ai/v1/parsing/upload</code> alongside the existing presigned-URL flow at
<code>https://api.pdf-ocr.cobbling.ai/v1/upload/direct</code>. Both still work; pick whichever fits your client.</p>
<p>See the refreshed <a href="https://your-docusaurus-site.example.com/docs/pdf-ocr">PDF OCR documentation</a> for the full reference.</p>
<p>As always, ping us if anything looks off — we read every report.</p>]]></content>
        <author>
            <name>Chao Cheng</name>
            <uri>https://github.com/chengchao</uri>
        </author>
        <category label="Product Updates" term="Product Updates"/>
    </entry>
    <entry>
        <title type="html"><![CDATA[PDF OCR replaces commaPDF]]></title>
        <id>https://your-docusaurus-site.example.com/blog/2025/04/10/pdf-ocr-upgrade</id>
        <link href="https://your-docusaurus-site.example.com/blog/2025/04/10/pdf-ocr-upgrade"/>
        <updated>2025-04-10T00:00:00.000Z</updated>
        <summary type="html"><![CDATA[Introducing the upgraded PDF OCR service with better performance, higher accuracy, and AdvancedOCR pricing parity.]]></summary>
        <content type="html"><![CDATA[<p>We are sunsetting <strong>commaPDF</strong> and rolling out the new <strong>PDF OCR</strong> experience for every Cobbling AI customer. The upgraded
service delivers better recognition accuracy, faster results, and richer structured output while keeping pricing aligned
with the existing AdvancedOCR plan.</p>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="why-the-change">Why the change?<a href="https://your-docusaurus-site.example.com/blog/2025/04/10/pdf-ocr-upgrade#why-the-change" class="hash-link" aria-label="Direct link to Why the change?" title="Direct link to Why the change?">​</a></h2>
<p>The document processing landscape has evolved quickly. PDF OCR is built on our latest recognition models and orchestration
pipeline, enabling:</p>
<ul>
<li>Higher fidelity extraction for complex layouts, tables, and mathematical content</li>
<li>Reduced processing latency across large document batches</li>
<li>More consistent Markdown and LaTeX output for downstream automation</li>
</ul>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="what-stays-the-same">What stays the same?<a href="https://your-docusaurus-site.example.com/blog/2025/04/10/pdf-ocr-upgrade#what-stays-the-same" class="hash-link" aria-label="Direct link to What stays the same?" title="Direct link to What stays the same?">​</a></h2>
<ul>
<li>Your AdvancedOCR subscription already includes PDF OCR—there is no price increase</li>
<li>Existing API keys continue to work without any changes</li>
<li>Webhook events (<code>job.created</code>, <code>job.started</code>, <code>job.completed</code>, <code>job.failed</code>) remain available so your integrations stay
in sync</li>
</ul>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="what-do-i-need-to-do">What do I need to do?<a href="https://your-docusaurus-site.example.com/blog/2025/04/10/pdf-ocr-upgrade#what-do-i-need-to-do" class="hash-link" aria-label="Direct link to What do I need to do?" title="Direct link to What do I need to do?">​</a></h2>
<ul>
<li>Update any hard-coded commaPDF endpoints to the new <code>api.pdf-ocr.cobbling.ai</code> domain</li>
<li>Review the refreshed <a href="https://your-docusaurus-site.example.com/docs/pdf-ocr">PDF OCR documentation</a> for the latest workflow and webhook details</li>
<li>Remove references to commaPDF in your internal documentation—PDF OCR is now the standard service</li>
</ul>
<p>We're excited to see what you build with the faster, smarter PDF OCR pipeline. If you have questions or need assistance
migrating, please reach out to the Cobbling AI support team.</p>]]></content>
        <author>
            <name>Chao Cheng</name>
            <uri>https://github.com/chengchao</uri>
        </author>
        <category label="Product Updates" term="Product Updates"/>
    </entry>
    <entry>
        <title type="html"><![CDATA[Making Document Intelligence Accessible and Affordable]]></title>
        <id>https://your-docusaurus-site.example.com/blog/our-first-blog</id>
        <link href="https://your-docusaurus-site.example.com/blog/our-first-blog"/>
        <updated>2025-03-19T00:00:00.000Z</updated>
        <summary type="html"><![CDATA[Docusaurus Plushie]]></summary>
        <content type="html"><![CDATA[<p><img decoding="async" loading="lazy" alt="Docusaurus Plushie" src="https://your-docusaurus-site.example.com/assets/images/banner-e76f8e11964295be0c87c1bc32f1cef9.png" width="2494" height="1024" class="img_nBFF"></p>
<p>Cobbling AI delivers high-quality PDF, image, and website parsing to markdown at affordable prices, eliminating the need to choose between expensive premium services or budget options with poor results.</p>
<p>Let's face it – document processing has come a long way. OCR technology is impressive, and many LLMs can handle documents pretty well these days. The real problem? Getting that quality without emptying your wallet.</p>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="the-price-quality-dilemma">The Price-Quality Dilemma<a href="https://your-docusaurus-site.example.com/blog/our-first-blog#the-price-quality-dilemma" class="hash-link" aria-label="Direct link to The Price-Quality Dilemma" title="Direct link to The Price-Quality Dilemma">​</a></h2>
<p>We've spent countless hours testing the market options, and here's what we discovered:</p>
<ol>
<li><strong>Want quality parsing? Be prepared to pay premium prices</strong> (like "maybe we should cut the office snacks" kind of prices)</li>
<li><strong>Want affordable parsing? Lower your expectations</strong> (hope you enjoy playing "guess what this paragraph was supposed to say!")</li>
</ol>
<p>It's that frustrating middle ground that was missing – high-quality document processing that normal humans and bootstrapped startups can actually afford.</p>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="cobbling-ai-the-sweet-spot-solution">Cobbling AI: The Sweet Spot Solution<a href="https://your-docusaurus-site.example.com/blog/our-first-blog#cobbling-ai-the-sweet-spot-solution" class="hash-link" aria-label="Direct link to Cobbling AI: The Sweet Spot Solution" title="Direct link to Cobbling AI: The Sweet Spot Solution">​</a></h2>
<p>We didn't build Cobbling AI because we invented some magical new OCR technology (though our engineers are pretty clever). We built it because we're obsessed with finding that elusive balance:</p>
<ul>
<li>Quality parsing that preserves document structure and formatting</li>
<li>Pricing that won't make your finance team break into a cold sweat</li>
<li>An API simple enough that you'll actually enjoy using it</li>
</ul>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="markdown-the-universal-translator">Markdown: The Universal Translator<a href="https://your-docusaurus-site.example.com/blog/our-first-blog#markdown-the-universal-translator" class="hash-link" aria-label="Direct link to Markdown: The Universal Translator" title="Direct link to Markdown: The Universal Translator">​</a></h2>
<p>We chose markdown as our output format because it's the perfect middle ground – structured enough to preserve document hierarchy but clean enough for LLMs to work with efficiently. It's like the diplomatic translator between your messy human documents and your picky AI models.</p>
<h2 class="anchor anchorWithStickyNavbar_QgMV" id="join-the-great-document-democratization">Join the "Great Document Democratization"<a href="https://your-docusaurus-site.example.com/blog/our-first-blog#join-the-great-document-democratization" class="hash-link" aria-label="Direct link to Join the &quot;Great Document Democratization&quot;" title="Direct link to Join the &quot;Great Document Democratization&quot;">​</a></h2>
<p>We believe quality document processing shouldn't be a luxury feature. Whether you're processing ten documents a month or ten thousand, Cobbling AI scales with you without the terrifying bill shock.</p>
<p>Give us a try and see for yourself – we're not claiming to reinvent the wheel, just making sure you can afford the car.</p>]]></content>
        <author>
            <name>Chao Cheng</name>
            <uri>https://github.com/chengchao</uri>
        </author>
        <category label="Hello" term="Hello"/>
        <category label="Docusaurus" term="Docusaurus"/>
    </entry>
</feed>