Release notes from paperless-ai

v3.0.9 | fix tag caching Issue from latest security fix breaking PP-AI

2025-11-04T07:28:45Z

Full Changelog: v3.0.8...v3.0.9

Fixed #763

Release v3.0.8 | Security Fix | Updates | Merges

2025-11-03T14:04:27Z

What's Changed

fix: Add an event listener to the submit button by @larskraemer in #609
Fix: Tag serialization issue causing "[object Object]" in AI prompts (#600) by @Helmi in #628
fix: multiple event handlers for tags in settings.js by @larskraemer in #608
fix: truncate long titles to 128 characters. by @beanieboi in #603
Implement sorting history by tags by @larskraemer in #638
feat: add pre-existing document types to AI context by @Helmi in #630
Fixed setup for azure openai model by @romanlum in #649

New Contributors

@larskraemer made their first contribution in #609
@Helmi made their first contribution in #628
@beanieboi made their first contribution in #603
@romanlum made their first contribution in #649

Full Changelog: v3.0.7...v3.0.8

v3.0.7

2025-06-22T17:49:29Z

News:

This release refactors the prompt handling to support user-customized placeholders in the system message. It introduces a dedicated RestrictionPromptService for placeholder replacement, removes legacy prompt-building logic from multiple services, and updates the associated test scripts.

Now you can define with %RESTRICTED_CORRESPONDENTS% and %RESTRICTED_TAGS% placeholders where in the prompt the existing Data is injected.

What's Changed

refactor prompt handing to support placeholders by @nielsmaerten in #584

New Contributors

@nielsmaerten made their first contribution in #584

Full Changelog: v3.0.6...v3.0.7

v3.0.6 | Critical fix for Custom Providers

2025-06-18T14:18:13Z

What's Changed

Fix 400 Error with Gemini API in document analysis by @clusterzx in #587

Full Changelog: v3.0.5...v3.0.6

v3.0.5 | Introduce fine-grained restrictions for tags and correspondents

2025-06-16T15:53:07Z

What's Changed

Fix punkt_tab download by @clusterzx in #542
RAG Chat for Azure Open AI by @habitoti in #537
Bump tar-fs from 2.1.2 to 2.1.3 in the npm_and_yarn group across 1 directory by @dependabot in #559
Enhance OpenAI service with external API integration and restrictions by @clusterzx in #582

Introduce fine-grained restrictions for tag and correspondent creation, along with the ability to fetch external data via an API. Update settings management and ensure proper handling of new features.

Full Changelog: v3.0.4...v3.0.5

v3.0.4 - RAG Performance and stability update

2025-05-21T19:03:53Z

Whats changed:

Hey dear community,

this new release will increase the performance of the RAG to startup by 95% (measured against 889 documents). The unnecessary step to re-index all documents after a restart of the container is now gone and it will start with all currently indexed documents. Also this made it possible to get real-time accurate information and health status of the RAG backend.

Don't miss this version!

Increase stability RAG process, use already indexed data instead of doing a re-index at every startup. Reducing the time by 95% to start. 43ff808

Full Changelog: v3.0.3...v3.0.4

v3.0.3

2025-05-20T19:37:45Z

Added / implemented a real working AI Status badge for the RAG Chat UI - to see if it is available and which model is loaded.
Removed the filter tab on RAG Chat as this feature is highly unstable and has to be fixed in the future.

Full Changelog: v3.0.1...v3.0.3

v3.0.1 - Fix date format as paperless-ngx updated their code

2025-05-20T15:18:22Z

What's Changed

[FIX] Send correct formatted created date on document update by @JonathanSaudhof in #533

New Contributors

@JonathanSaudhof made their first contribution in #533

Full Changelog: v3.0.0...v3.0.1

v3.0.0 - Introduce the RAG Chat

2025-05-18T15:32:52Z

🚀 New Feature Announcement
Paperless-AI now includes a powerful, integrated RAG-powered Chat interface!
Introducing a whole new way to interact with your Paperless-NGX archive: instead of browsing, filtering, or guessing which tags to search for — just ask.
Thanks to Retrieval-Augmented Generation (RAG), you can now search semantically across the full content of your documents and get human-like answers instantly.

🔍 No more guessing. Just ask.
Want to know “When did I receive my electricity contract?”, “How much did I pay for the last car repair?” or “Which documents mention my health insurance?” — Paperless-AI will find it for you, even if you don’t remember the exact title, sender, or date.

💡 What does RAG bring to Paperless-NGX?

True full-text understanding of your documents

Context-aware responses — beyond keyword search

Useful when dealing with large or chaotic document archives

Saves time, avoids frustration, and unlocks insights you may have forgotten you had stored

Blazingly fast answers backed by your own trusted data

⚠️ Important Note: If you're installing Paperless-AI for the first time, please restart the container after completing the setup routine (where you enter your API keys and preferences). This ensures that all services initialize correctly and your RAG index is built properly.
➕ This step is not required when updating an existing installation.

v2.7.8

2025-05-16T19:50:06Z

What's Changed

Add a configurable token limitation and response size by @habitoti in #496
Exactly compute the new content size according to given token limit by @habitoti in #498
Fix wrong model reference for Azure Playground by @habitoti in #511
Bump the npm_and_yarn group across 1 directory with 3 updates by @dependabot in #514
fix issue #516 by @KoffeinKaio in #521

New Contributors

@habitoti made their first contribution in #496
@KoffeinKaio made their first contribution in #521

Full Changelog: v2.7.6...v2.7.8