DocWire 2026 update

blog/2026-03-11-DocWire-up-to-date-2026.md

@@ -0,0 +1,109 @@
+---
+title: 'DocWire: A 2025 Retrospective'
+authors: krzysztof
+tags: [news, update]
+---
+
+For a while, this blog has been quiet — but DocWire hasn’t.
+
+Behind the scenes, 2025 became a defining year for the project. What started as a document-processing toolkit gradually transformed into something far more ambitious: a modular, message-driven framework capable of powering modern data pipelines and AI-driven workflows.
+
+This post is a short retrospective of that transformation — and a look at the engineering decisions that made it possible.
+
+
+<!--truncate-->
+
+## A New Core: Message-Driven Architecture
+
+The most fundamental shift happened deep inside the SDK.
+
+Earlier versions relied on tightly coupled data representations, which made extending the system increasingly difficult. Adding new data types often required changes across multiple components.
+
+In 2025, DocWire introduced a message-driven core — a decoupled model where processing stages communicate through generic messages flowing across pipelines.
+
+This shift unlocked several key benefits:
+- clearer data flow across processing stages  
+- independent, composable components  
+- easier extensibility without breaking existing pipelines  
+- the ability to expose pipelines as standalone microservices  
+
+Rather than a collection of utilities, DocWire started to behave like a coherent processing platform.
+
+
+
+## AI Integration: Local and Cloud by Design
+
+The new architecture enabled deeper integration with language models and semantic processing.
+
+DocWire pipelines can now leverage both:
+- local models for offline and privacy-sensitive scenarios  
+- cloud-based models for advanced conversational or generative workflows  
+
+This hybrid approach gives developers flexibility to choose the right execution context for each task — from fully offline semantic search to cloud-augmented analysis pipelines.
+
+Even lightweight local models can provide multilingual embeddings and semantic understanding directly within applications, without network dependencies.
+
+
+
+## High-Fidelity Document Understanding
+
+Alongside architectural changes, the core parsing capabilities evolved significantly.
+
+Throughout 2025, DocWire introduced:
+- a redesigned HTML parsing pipeline  
+- a new PDF engine with improved stability  
+- image extraction from complex document formats  
+- position-aware layout reconstruction  
+
+This progression culminated in what we describe as high-fidelity document understanding: the ability to capture not only textual content but also the structural and spatial context of documents.
+
+Such layout awareness is essential for accurate indexing, semantic retrieval, and downstream AI processing.
+
+
+
+## Developer Experience and Production Readiness
+
+A major focus of the year was making DocWire not only powerful but also pleasant to build with.
+
+Key improvements included:
+- a zero-overhead diagnostic system for development-time insights  
+- structured JSON logging for easier observability  
+- richer error reporting with contextual metadata  
+- safer assertions and defensive programming utilities  
+- modernization of the build system and CI workflows  
+
+These changes may be less visible than new features, but they are critical for building reliable production systems.
+
+
+
+## Watch: 2025 Engineering Recap
+
+To illustrate this transformation visually, we prepared a short overview of the architectural and functional changes introduced throughout the year.
+
+[![Video title](https://img.youtube.com/vi/vBgrIh04R-I/hqdefault.jpg)](https://youtu.be/vBgrIh04R-I)
+
+
+
+## What This Means
+
+Looking back, 2025 was less about incremental improvement and more about redefinition.
+
+DocWire is no longer just a library integrated into applications. The foundations now exist for treating it as an application backbone — a processing layer capable of running independently as part of larger data infrastructures.
+
+This evolution opens the door to new usage patterns:
+- pipeline-based document ingestion  
+- service-oriented deployment models  
+- local AI processing workflows  
+- backend systems centered around structured data extraction  
+
+In short, the toolkit matured into a framework.
+
+
+
+## What Comes Next
+
+This retrospective sets the stage for what follows.
+
+In the next post, we’ll introduce a concept that helps describe this evolution at a higher level: DocWire as the invisible engine behind secure data pipelines.
+
+For now, if you’ve been following the project — or discovering it for the first time — 2025 represents a turning point worth exploring.

blog/authors.yml

@@ -2,3 +2,8 @@ ferid:
   name: Ferid Obeidat
   title: Author
   image_url: /img/Ferid-headshot.png
+krzysztof:
+  name: Krzysztof Nowicki
+  title: Author
+  image_url: /img/krzysztof.png
+