v2.7.1

# Paperless AI v2.7.1 – New Features and Improvements

## Overview of Changes

Paperless AI v2.7.1 is a substantial update bringing both **new features** and **enhancements** for users and developers. This release merges a series of development updates into the main branch (via PR [#394](https://github.com/clusterzx/paperless-ai/pull/394) and [#395](https://github.com/clusterzx/paperless-ai/pull/395)) to deliver improvements including:

- **Azure OpenAI Service Integration** – a new provider option to use Microsoft’s Azure-hosted OpenAI models.
- **Structured Output Implementation** – the AI’s responses can now be returned in a consistent, structured format (e.g. JSON) for easier parsing.
- **DeepSeek-R1 Model Support** – compatibility with the advanced DeepSeek-R1 large language model, expanding the range of AI models you can use.
- Various **UI refinements, bug fixes, and performance tweaks** carried over from the development branch merge.

These changes aim to improve Paperless AI’s **flexibility** (supporting more AI backends), **reliability** of AI outputs, and **overall user experience**.

---

## 🔹 Azure OpenAI Integration

**What’s new:** Paperless AI now supports **Azure OpenAI**, allowing users to connect to Microsoft's hosted AI models, such as GPT-4 deployed on Azure.

**Technical details:**  
- A dedicated **“Azure” provider** option has been added in the settings.  
- Uses Azure’s API structure (via the `AzureOpenAI` class).  
- Requires Azure-specific settings like **Resource Name, Deployment ID**, and an **API Key**.  

**Impact:**  
- Enterprise users can now leverage Azure’s AI models with **better availability** and **higher context limits** (depending on Azure’s offering).  
- Developers gain **first-class support** for Azure OpenAI, simplifying integrations.  
- Contributed by **@mb-ai-dev** in [PR #378](https://github.com/clusterzx/paperless-ai/pull/378).

---

## 🔹 Structured Output Implementation

**What’s new:** AI-generated responses can now be returned in a structured format (like **JSON**), making it easier for **automated processing** and **data extraction**.

**Technical details:**  
- The AI can now **return structured JSON** responses instead of plain text.  
- Improves **document metadata extraction** by ensuring a predictable format.  
- Reduces manual parsing and **makes AI output more machine-readable**.  

**Impact:**  
- Users will **receive clean, structured data** from AI responses, reducing the need for manual corrections.  
- Automations and integrations with **Paperless-ngx** become more reliable.
- Improves **data accuracy** for document categorization, tagging, and metadata handling.

---

## 🔹 DeepSeek-R1 Model Support

**What’s new:** Paperless AI now **supports DeepSeek-R1**, a powerful **open-source LLM** known for **strong reasoning and coding capabilities**.

**Technical details:**  
- Works via both **local deployments** (e.g., **Ollama**) and **API endpoints**.  
- Uses the OpenAI API format, making it easy to integrate.  

**Impact:**  
- Users can now choose DeepSeek-R1 as an **alternative to OpenAI models**.  
- Ideal for those wanting **a self-hosted AI solution** with advanced reasoning.  
- Expands model flexibility, keeping Paperless AI open to **multiple AI backends**.

---

## 🔹 Usability, Performance, and AI Model Compatibility

- **More AI model choices** – OpenAI, Azure, DeepSeek, and local models all work seamlessly.  
- **Faster, structured AI responses** – reduces unnecessary AI hallucinations and improves automation workflows.  
- **More reliable error handling** – better logging and debugging when using different AI providers.  
- **Future-proofed AI pipeline** – with structured outputs and wider model support, Paperless AI is becoming a **more robust document AI solution**.

---

## 🌟 Community Contributions

This release was made possible thanks to contributions from the community:

- **@mb-ai-dev** made their first contribution by implementing **Azure OpenAI support** ([PR #378](https://github.com/clusterzx/paperless-ai/pull/378)).  
- Feature requests from the community directly influenced the addition of **structured outputs** and **DeepSeek-R1 support**.  
- Thanks to all contributors who tested and provided feedback!

## What's Changed
* Merge latest version to dev by @clusterzx in https://github.com/clusterzx/paperless-ai/pull/394
* Merge pull request #394 from clusterzx/main by @clusterzx in https://github.com/clusterzx/paperless-ai/pull/395
* Add support for Azure OpenAI by @mb-ai-dev in https://github.com/clusterzx/paperless-ai/pull/378
* Implemented Structured Output
* Added Deepseek-r1 support

## New Contributors
* @mb-ai-dev made their first contribution in https://github.com/clusterzx/paperless-ai/pull/378

**Full Changelog**: https://github.com/clusterzx/paperless-ai/compare/v2.6.7...v2.7.1