{"id":12103,"date":"2026-04-07T09:46:06","date_gmt":"2026-04-07T16:46:06","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/cosmosdb\/?p=12103"},"modified":"2026-04-13T09:30:48","modified_gmt":"2026-04-13T16:30:48","slug":"scalable-ai-with-azure-cosmos-db-tredence-intelligent-document-processing-idp-march-2026","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/cosmosdb\/scalable-ai-with-azure-cosmos-db-tredence-intelligent-document-processing-idp-march-2026\/","title":{"rendered":"Scalable AI with Azure Cosmos DB: Tredence Intelligent Document Processing (IDP)"},"content":{"rendered":"<p data-start=\"264\" data-end=\"655\"><strong data-start=\"264\" data-end=\"427\">Azure Cosmos DB enables scalable AI-driven document processing, addressing one of the biggest barriers to operational scale in today\u2019s enterprise AI landscape.<\/strong> Organizations continue to manage massive volumes of structured and unstructured content\u2014contracts, regulatory filings, operational records, images, and field documentation\u2014yet many workflows remain fragmented, manual, and slow.<\/p>\n<p data-start=\"657\" data-end=\"1112\">In this month&#8217;s edition of our <strong>Scalable AI with Azure Cosmos DB series<\/strong>, we featured <a href=\"https:\/\/www.tredence.com\/\" target=\"_blank\" rel=\"noopener\">Tredence<\/a> and their production-grade Intelligent Document Processing (IDP) solution\u2014an Azure-native architecture designed to transform document-heavy business processes into scalable, AI-driven systems.<\/p>\n<p data-start=\"657\" data-end=\"1112\">Presented by <strong>Unmesh Kulkarni<\/strong> \u2013 SVP of AI &amp; <strong>Ashwini Sharma<\/strong> \u2013 Director, GenAI from <a href=\"https:\/\/www.tredence.com\/\">Tredence<\/a>, this edition showcases how Azure Cosmos DB, Azure Document Intelligence, and agentic AI patterns power enterprise-scale document understanding and decision-making.<\/p>\n<p><center><iframe src=\"\/\/www.youtube.com\/embed\/0gSS9PTZwlw?si=36nHkhm3bkYRERY7\" width=\"560\" height=\"314\" allowfullscreen=\"allowfullscreen\"><\/iframe><\/center>This post recaps the architecture presented in the session, including the document pipeline, agentic orchestration layer, and a real-world telecom scenario where the solution helped reduce decision cycles by <strong>80%<\/strong> and unlock <strong>multi-million-dollar savings<\/strong>.<\/p>\n<h2>Enterprise document processing challenges: moving from legacy systems to AI at scale<\/h2>\n<p>Enterprises in industries such as telecom, retail, and media often operate across disconnected systems while handling high volumes of business-critical documents. These may include contracts, legal agreements, compliance records, field inspection reports, OSS\/BSS data, scanned files, and image-based documentation. Traditional approaches to processing this information are often manual, inconsistent, and difficult to scale.<\/p>\n<p>Common pain points include:<\/p>\n<ul>\n<li>Contract analysis that takes weeks or even months<\/li>\n<li>Slow and error-prone revenue mapping across operational systems<\/li>\n<li>Outdated site documentation during infrastructure transformation programs such as 5G to 6G transitions<\/li>\n<li>Siloed data estates that prevent unified analysis<\/li>\n<li>Compliance risk caused by incomplete or inconsistent extraction<\/li>\n<\/ul>\n<p>Tredence addresses this challenge through its <strong>IDP platform<\/strong>, built natively on Azure and designed for enterprise-grade scale, traceability, and automation.<\/p>\n<p><figure id=\"attachment_12118\" aria-labelledby=\"figcaption_attachment_12118\" class=\"wp-caption aligncenter\" ><a href=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/1.png\"><img decoding=\"async\" class=\"wp-image-12118 size-large\" src=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/1-1024x579.png\" alt=\"ATOM.AI agentic intelligent document processing architecture on Microsoft platform showing Azure Cosmos DB unified data layer, multimodal AI extraction, OCR, vector embeddings, and enterprise document workflows.\" width=\"1024\" height=\"579\" srcset=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/1-1024x579.png 1024w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/1-300x170.png 300w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/1-768x434.png 768w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/1.png 1338w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/a><figcaption id=\"figcaption_attachment_12118\" class=\"wp-caption-text\"><center><strong>Atom.ai (IDP) Data Flow<\/strong><\/center><\/figcaption><\/figure><\/p>\n<h2>A three-layer architecture for scalable enterprise AI<\/h2>\n<p>The solution presented follows a clean three-layer architecture:<\/p>\n<h3>1. Experience layer<\/h3>\n<p>This includes web and mobile user experiences, along with conversational interfaces that allow business users to interact with extracted knowledge and system recommendations in natural language.<\/p>\n<h3>2. Agentic intelligence layer<\/h3>\n<p>This layer orchestrates multi-agent workflows using modern AI frameworks such as <strong>LangChain<\/strong>, <strong>LangGraph Semantic Kernel<\/strong>, and <strong>Microsoft Agent Framework<\/strong>. Specialized agents coordinate tasks such as extraction, validation, reasoning, search, and decision support. These agents communicate through agent-to-agent and agent-to-tool patterns, including data access through the <strong>Azure Cosmos DB MCP Toolkit<\/strong>.<\/p>\n<h3>3. Unified data plane<\/h3>\n<p>At the foundation is <strong>Azure Cosmos DB<\/strong>, which acts as the unified data and memory layer for the solution. It stores:<\/p>\n<ul>\n<li>Operational application data<\/li>\n<li>Structured extraction outputs<\/li>\n<li>Metadata and document state<\/li>\n<li>Vector embeddings for semantic retrieval<\/li>\n<li>Graph relationships for contextual reasoning<\/li>\n<li>Session memory and interaction context for agents<\/li>\n<\/ul>\n<p>This unified model allows the system to support both transactional and AI-native workloads without introducing additional data silos.<\/p>\n<p><a href=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/2.png\"><img decoding=\"async\" class=\"aligncenter wp-image-12119 size-large\" src=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/2-1024x578.png\" alt=\"Azure Cosmos DB data plane for agentic AI illustrating unified storage of documents, metadata, vector embeddings, graph relationships, and agent memory for scalable enterprise AI workloads.\" width=\"1024\" height=\"578\" srcset=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/2-1024x578.png 1024w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/2-300x169.png 300w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/2-768x434.png 768w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/2.png 1342w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/a><\/p>\n<p><center><span style=\"font-size: 10pt;\"><strong>Data plane for Agents &amp; AI Apps<\/strong><\/span><\/center><\/p>\n<h2>Why Azure Cosmos DB is the backbone<\/h2>\n<p>A key theme from the session was that the value of the solution comes not only from AI models, but from the strength of the underlying data architecture.<\/p>\n<p><strong>Azure Cosmos DB<\/strong> enables this architecture through several capabilities:<\/p>\n<ul>\n<li><strong>Multi-model flexibility<\/strong> \u2013 supporting document-style access, graph relationships, and vector-based retrieval patterns in a single platform<\/li>\n<li><strong>Low-latency global scale<\/strong> \u2013 enabling consistent application responsiveness across regions<\/li>\n<li><strong>Automatic horizontal scaling<\/strong> \u2013 allowing the system to absorb variable ingestion and query workloads<\/li>\n<li><strong>Support for hybrid workloads<\/strong> \u2013 combining transactional operations, metadata filtering, semantic retrieval, and relationship traversal<\/li>\n<\/ul>\n<p>This is especially important for agentic systems, where AI agents need fast access to both current state and historical context in order to reason effectively and act reliably.<\/p>\n<h2>AI-driven document pipeline: precision, auditability, and scale<\/h2>\n<p>At the core of the solution is a document intelligence pipeline built using <strong>Azure AI services<\/strong>, including <strong>Azure Document Intelligence<\/strong> and <strong>Content Understanding<\/strong>.<\/p>\n<p>The pipeline supports:<\/p>\n<ul>\n<li><strong>Multimodal ingestion<\/strong> from PDFs, scanned images, operational systems, and visual field inputs such as drone imagery<\/li>\n<li><strong>OCR and layout understanding<\/strong> for extracting structure from semi-structured and unstructured files<\/li>\n<li><strong>Table, chart, and content extraction<\/strong> for downstream analytics and reasoning<\/li>\n<li><strong>Classification and confidence scoring<\/strong> to improve extraction reliability<\/li>\n<li><strong>Verbatim extraction<\/strong> for legal and compliance-sensitive scenarios where paraphrasing is not acceptable<\/li>\n<li><strong>Smart backfill capabilities<\/strong> to add or correct fields later without requiring the full document corpus to be reprocessed<\/li>\n<li><strong>Post-processing and validation<\/strong> across documents and systems<\/li>\n<li><strong>Human-in-the-loop workflows<\/strong> to continuously improve extraction quality and governance<\/li>\n<\/ul>\n<p>The output persisted directly into <strong>Azure Cosmos DB<\/strong> as structured fields, metadata, embeddings, and contextual relationships\u2014creating an immediately usable knowledge layer for downstream agents and applications.<\/p>\n<h2>From extraction to action: agentic AI for document processing and decision-making<\/h2>\n<p>A major differentiator in the Tredence approach is that the system does not stop at extraction. Instead, it uses <strong>agentic AI<\/strong> to turn document understanding into business decisions.<\/p>\n<p>In this architecture:<\/p>\n<ul>\n<li>Orchestrator agents coordinate specialized sub-agents<\/li>\n<li>Sub-agents handle domain tasks such as contract interpretation, revenue analysis, or document validation<\/li>\n<li>Persistent memory in <strong>Azure Cosmos DB<\/strong> allows the system to maintain context across sessions<\/li>\n<li>AI models reason over extracted knowledge to identify patterns, anomalies, and opportunities<\/li>\n<li>Tools integrated through the <strong>Cosmos DB MCP Toolkit<\/strong> allow agents to securely perform data access operations, semantic retrieval, and contextual graph exploration<\/li>\n<\/ul>\n<p>This means agents can do more than extract fields. They can correlate data, validate findings, surface risks, and recommend next-best actions\u2014while maintaining traceability for governance and review.<\/p>\n<p><figure id=\"attachment_12120\" aria-labelledby=\"figcaption_attachment_12120\" class=\"wp-caption aligncenter\" ><a href=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/3.png\"><img decoding=\"async\" class=\"wp-image-12120 size-large\" src=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/3-1024x578.png\" alt=\"Azure Cosmos DB-powered Project Zenith architecture illustrating AI-driven telecom network optimization, integrating contracts, operational data, vector embeddings, and agentic AI for real-time enterprise decisions.\" width=\"1024\" height=\"578\" srcset=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/3-1024x578.png 1024w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/3-300x169.png 300w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/3-768x434.png 768w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/3.png 1342w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/a><figcaption id=\"figcaption_attachment_12120\" class=\"wp-caption-text\"><center><strong>Project Zenith Architecture<\/strong><\/center><\/figcaption><\/figure><\/p>\n<h2>Real-world telecom scenario: optimizing network rollout and reducing operational cost<\/h2>\n<p>The session highlighted a telecom scenario in which the solution was applied to support rollout planning and infrastructure optimization.<\/p>\n<p>The objective was to accelerate network modernization by identifying opportunities to optimize existing infrastructure, reduce redundant sites, and improve decision-making across contract, operations, and revenue data.<\/p>\n<p>The solution included several core modules:<\/p>\n<h3>Strategic cell site contract engine<\/h3>\n<p>Contracts were processed to extract key commercial and operational terms such as rent, vendor details, equipment references, renewal timelines, and obligations. This information was then linked with operational metrics stored in <strong>Azure Cosmos DB<\/strong>.<\/p>\n<h3>Revenue propagation and prioritization<\/h3>\n<p>Coverage, utilization, and business impact signals were combined to estimate revenue potential and help prioritize the highest-value sites.<\/p>\n<h3>Zero-touch drone surveys<\/h3>\n<p>Visual field inputs were used to automate site validation and update documentation more efficiently.<\/p>\n<p>Because contracts, metadata, embeddings, and site relationships were all stored in a common data platform, the system could support rich operational questions such as:<\/p>\n<blockquote><p>Which sites associated with a specific vendor are nearing expiration and showing low utilization?<\/p><\/blockquote>\n<p>This type of multi-dimensional reasoning is difficult to achieve in fragmented architectures, but becomes practical when structured data, unstructured data, semantic context, and graph relationships are all available in one place.<\/p>\n<p><figure id=\"attachment_12121\" aria-labelledby=\"figcaption_attachment_12121\" class=\"wp-caption alignnone\" ><a href=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/4.png\"><img decoding=\"async\" class=\"wp-image-12121 size-large\" src=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/4-1024x580.png\" alt=\"Azure Cosmos DB-powered telecom network map visualizing AI-driven site optimization, integrating contracts, operational data, and agentic AI for real-time enterprise decision intelligence.\" width=\"1024\" height=\"580\" srcset=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/4-1024x580.png 1024w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/4-300x170.png 300w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/4-768x435.png 768w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/4.png 1344w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/a><figcaption id=\"figcaption_attachment_12121\" class=\"wp-caption-text\"><center><strong>Project Zenith network visual<\/strong><\/center><\/figcaption><\/figure><\/p>\n<h2>Quantified impact<\/h2>\n<p>According to the session, the solution delivered strong business outcomes, including:<\/p>\n<ul>\n<li><strong>Decision cycles reduced from 2 months to 1\u20132 weeks<\/strong><\/li>\n<li><strong>80% reduction in processing time<\/strong><\/li>\n<li><strong>$2.49M in annual savings in one Dallas market<\/strong><\/li>\n<li><strong>Projected national savings of $50M\u2013$100M<\/strong><\/li>\n<li>Support for <strong>millions of documents<\/strong> with near real-time analytics<\/li>\n<\/ul>\n<p>These outcomes underscore an important point: scalable AI is not just about model quality. It is about data architecture, operational integration, and system design.<\/p>\n<h2>Why AI-driven document processing with Azure Cosmos DB matters<\/h2>\n<p>Traditional IDP systems often plateau when they encounter enterprise-scale complexity\u2014multiple systems, multiple content types, compliance constraints, and the need for continuous reasoning.<\/p>\n<p>Tredence\u2019s architecture works because it combines:<\/p>\n<ul>\n<li>A unified data plane in <strong>Azure Cosmos DB<\/strong><\/li>\n<li>Enterprise-grade AI extraction services<\/li>\n<li>Agentic design patterns for reasoning and orchestration<\/li>\n<li>Azure-native governance, security, and extensibility<\/li>\n<\/ul>\n<p>This is the shift many enterprises are now making\u2014from isolated AI experiments to integrated production systems that can reason, scale, and drive measurable business value.<\/p>\n<p><figure id=\"attachment_12110\" aria-labelledby=\"figcaption_attachment_12110\" class=\"wp-caption aligncenter\" ><a href=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/Tredence-Cosmos-data.png\"><img decoding=\"async\" class=\"wp-image-12110 size-full\" src=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/Tredence-Cosmos-data.png\" alt=\"Azure Cosmos DB query interface visualizing JSON documents, data explorer results, and real-time querying for agentic AI, vector search, and enterprise-scale applications.\" width=\"680\" height=\"402\" srcset=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/Tredence-Cosmos-data.png 680w, https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-content\/uploads\/sites\/52\/2026\/04\/Tredence-Cosmos-data-300x177.png 300w\" sizes=\"(max-width: 680px) 100vw, 680px\" \/><\/a><figcaption id=\"figcaption_attachment_12110\" class=\"wp-caption-text\"><center><strong>Azure Cosmos DB Query panel (showing query generated by agent)<\/strong><\/center><\/figcaption><\/figure><\/p>\n<h2>Final thoughts<\/h2>\n<p>The Tredence session is a strong example of what production-ready <strong>agentic AI<\/strong> looks like when paired with the right data foundation. By combining <strong>Azure Cosmos DB<\/strong>, <strong>Azure Document Intelligence<\/strong>, and modern AI orchestration frameworks, enterprises can move beyond manual document processing and toward intelligent systems that extract, understand, validate, and act.<\/p>\n<p>For organizations working with large volumes of unstructured content, this pattern offers a clear path to reducing compliance risk, accelerating decision-making, and unlocking meaningful operational savings.<\/p>\n<p>You can watch the full session here: <a href=\"https:\/\/aka.ms\/scalableai-live-mar26\">https:\/\/aka.ms\/scalableai-live-mar26<\/a>\nYou can also explore the broader series here: <a href=\"https:\/\/devblogs.microsoft.com\/cosmosdb\/scalable-ai-with-azure-cosmos-db-video-series\/\">Scalable AI with Azure Cosmos DB \u2013 Video Series<\/a><\/p>\n<p>If you\u2019re building AI solutions that need to operate over massive volumes of unstructured data\u2014with memory, semantic retrieval, full text, highly performant vector search and intelligent orchestration\u2014Azure Cosmos DB is the answer to it.<\/p>\n<h2 id=\"about-azure-cosmos-db\"><strong>About Azure Cosmos DB<\/strong><\/h2>\n<p>Azure Cosmos DB is a fully managed and serverless NoSQL and vector database for modern app development, including AI applications. With its SLA-backed speed and availability as well as instant dynamic scalability, it is ideal for real-time NoSQL and MongoDB applications that require high performance and distributed computing over massive volumes of NoSQL and vector data.<\/p>\n<p>To stay in the loop on Azure Cosmos DB updates, follow us on\u00a0<a href=\"https:\/\/twitter.com\/AzureCosmosDB\" target=\"_blank\" rel=\"noopener\">X<\/a>,\u00a0<a href=\"https:\/\/aka.ms\/AzureCosmosDBYouTube\" target=\"_blank\" rel=\"noopener\">YouTube<\/a>, and\u00a0<a href=\"https:\/\/www.linkedin.com\/company\/azure-cosmos-db\/\" target=\"_blank\" rel=\"noopener\">LinkedIn<\/a>.\u00a0 Join the discussion with other developers on the\u00a0<a href=\"https:\/\/discord.gg\/pczdC2SU\" target=\"_blank\" rel=\"noopener\">#nosql channel on the Microsoft Open Source Discord<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Azure Cosmos DB enables scalable AI-driven document processing, addressing one of the biggest barriers to operational scale in today\u2019s enterprise AI landscape. Organizations continue to manage massive volumes of structured and unstructured content\u2014contracts, regulatory filings, operational records, images, and field documentation\u2014yet many workflows remain fragmented, manual, and slow. In this month&#8217;s edition of our Scalable [&hellip;]<\/p>\n","protected":false},"author":13641,"featured_media":12113,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[1980,14],"tags":[],"class_list":["post-12103","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-azure-cosmos-db","category-core-sql-api"],"acf":[],"blog_post_summary":"<p>Azure Cosmos DB enables scalable AI-driven document processing, addressing one of the biggest barriers to operational scale in today\u2019s enterprise AI landscape. Organizations continue to manage massive volumes of structured and unstructured content\u2014contracts, regulatory filings, operational records, images, and field documentation\u2014yet many workflows remain fragmented, manual, and slow. In this month&#8217;s edition of our Scalable [&hellip;]<\/p>\n","_links":{"self":[{"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/posts\/12103","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/users\/13641"}],"replies":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/comments?post=12103"}],"version-history":[{"count":0,"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/posts\/12103\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/media\/12113"}],"wp:attachment":[{"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/media?parent=12103"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/categories?post=12103"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devblogs.microsoft.com\/cosmosdb\/wp-json\/wp\/v2\/tags?post=12103"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}