Changelog

Try the all-new PDF Pals v2 (beta): ask your documents questions and get answers that cite the page, now across your whole library, with private local models and more file types. Your license activates v2 during the beta. This update also fixes Anthropic API keys and brings many fixes to Classic.

Introducing PDF Pals v2 (beta)

• A brand-new app, rebuilt from the ground up. Open it from PDF Pals › Try the New PDF Pals (macOS 14 or later).
• Answers cite the page they came from. Click a citation to jump to the passage.
• A Library for all your documents and chats in one place.
• More than PDFs: books (EPUB), Word, Excel, PowerPoint and Markdown.
• Better answers: v2 finds the right passage more often.
• Local models run on your Mac, free and private. With local models, nothing leaves your Mac.
• Bring over your documents, prompts and AI providers from Classic in a few clicks. Your Classic chats stay where they are.
• Switch between Classic and v2 anytime in Settings › Experience. Switching never moves or deletes a library.

Classic

• Anthropic: fixed API keys failing because PDF Pals used retired Claude 3 models. Your key is now checked against Anthropic's model list, and chats on retired models move to current versions automatically.
• Long chats and documents now fit the model you use, with fewer "prompt too long" errors, especially with local models such as Ollama.
• Models with a large context window read more passages from your documents, for fuller answers.
• OpenAI: fixed reasoning models (o-series, GPT-5 and later) failing in chats.
• OpenRouter: fixed valid keys failing when the selected model was retired.
• Perplexity, Mistral, Groq and Google AI: up-to-date model lists and default models.
• Chats find documents you moved or renamed in Finder. Documents moved to the Trash show as missing.
• OCR detects the document's language, so scanned non-English PDFs are read correctly.
• Fixed PDF pages that were unreadable in Dark Mode, and Settings › Experience appearing blank in Dark Mode.
• Fixed short answers that sometimes didn't appear in the chat, and a blank area above the messages in some chats.
• Other bug fixes and improvements

• PDF Pals v2: new chats keep the last 10 questions and answers as context (was 2).
• PDF Pals v2: when the free trial ends, a window shows how to buy or enter your license key.
• PDF Pals v2: Help › PDF Pals Help opens the help center.
• Contact Support opens an email to [email protected].
• Other bug fixes and improvements

• New: try the new PDF Pals (v2 beta) right from the app. Answers cite the exact page and open it, it reads Word, PowerPoint, Excel and EPUB files, and it can run local models so nothing leaves your Mac. You can switch back to Classic at any time, and your Classic library stays as it is.
• New: PDF Pals › Try the New PDF Pals… opens the v2 offer any time.
• New: Settings › License shows what your license covers, including v2.
• Books and Office documents can be opened with PDF Pals from Finder.
• License activation is more reliable and no longer unlinks on a network hiccup.
• Other bug fixes and improvements

• New: Added support for Google Gemini Models
• Other bug fixes & improvements

Minor fixes. No detailed notes were published for this build.

• New: Fetch claude models automatically from Anthropic
• Other bug fixes & improvements

This is a major release with many useful features. Upgrade is highly recommended.

• New: Fetch models automatically from OpenAI, OpenRouter, Groq, Mistral, Perplexity and more
• New: Added support for Anthropic Claude 3, including vision capability.
• New: All-new chat message list implementation. Faster and better!
• New: Improved LaTeX rendering
• New: Customize font size for the chat input
• Improvement: Allow insecure network connection. This is neccessary to connect to local inference servers
• Fix: Fixed a crash issue related to SwiftUI internal rendering engine
• Fix: Fixed the issue where cost estimation was inaccurate in some cases
• Improvement: BoltAI stops streaming on user's interuption

• New: A new Prompt Library
• Natively support local AI models via Ollama
• Allow setting number of parallel indexing tasks (default: 5)
• Support custom user instruction
• Added support for Groq
• Added "Enter to send message" option
• Allow toggling the Document Viewer
• Improved rendering performance

• Fixed an issue when using GPT-4 Vision
• Fixed a race condition issue for some users

This release integrates with GPT 4 Vision and supports more AI service providers, including Local LLM Inference Server.

Read full release note →

• New: Supports more AI service providers: Mistral AI, Perplexity and Together AI
• New: Supports custom OpenAI-compatible servers: OpenAI proxy, LocalAI and LM Studio Inference Server...
• New: Incorporates ShotSolve into PDF Pals: take a screenshot of your PDF and ask AI about it
• New: Switch to the new embedding model `text-embedding-3`: better & cheaper
• New: Added support for the new GPT 4 Turbo model (`gpt-4-0125-preview`)
• New: Allow searching within the current chat
• Fix: Fixed the bug where the Upgrade window keeps showing
• Fix: Fixed the bug where users cannot add new AI Service Provider

This release integrates with GPT 4 Vision and supports more AI service providers, including Local LLM Inference Server.

• New: Supports more AI service providers: Mistral AI, Perplexity and Together AI
• New: Supports custom OpenAI-compatible servers: OpenAI proxy, LocalAI and LM Studio Inference Server...
• New: Incorporates ShotSolve into PDF Pals: take a screenshot of your PDF and ask AI about it
• New: Switch to the new embedding model `text-embedding-3`: better & cheaper
• New: Added support for the new GPT 4 Turbo model (`gpt-4-0125-preview`)
• New: Allow searching within the current chat
• Fix: Fixed the bug where the Upgrade window keeps showing

Minor bug fixes and improvements

Read full v1.4 release note →

Biggest release since launch: 9 new major features and significant performance improvement.

• New: Support text-to-speech and speech-to-text using OpenAI API. Your fingers will thank you.
• New: PDF search & better navigation
• New: Workspace (Conversation) Switcher. Quickly switch between conversations
• New: Prompt with highlighted text & PDF page content
• New: Export chat to HTML, Markdown or JSON (compatible with OpenAI’s finetuning format)
• New: Better chat management. You can reset the chat context, delete all messages…
• New: Custom font & theme for code blocks
• New: Support advantaged parameters like context limit, temperature, presence penalty, frequent penalty...
• New: The markdown renderer has been rebuilt from scratch. It significantly improved the scroll performance now!
• Other quality-of-life improvements:
  • Filter conversation by title
  • Support multiline copy
  • Code highlighting during streaming
  • Responsive copy button
  • Automatically scroll to bottom on AI response
  • Improved message edit dialog
  • Fixed issue with OpenRouter model list

• Fixed the bug where the input field does not show for OpenRouter users

• Supports new models: GPT-4 Turbo (Preview) and GPT-3.5 Turbo 1106 (New)

• Fixed a minor issue for Azure OpenAI Service

IMPORTANT: this release temporarily removes the ability to customize System Instruction (read more below)

• Improved the accuracy of page references
• Improved AI responses: you can ask PDF Pals about a specific page. For example: "Summarize the content of page 12"
• Improvement: You can continue to ask questions while new documents are being indexed
• Improvement: You can remove a document from a multi-doc chat now

This is a minor release to fix the indexing issue with multi-doc chat

• Fixed the issue where PDF Pals might not be able to find relevant content from newly added documents
• Fixed the issue where Page References might be incorrect in multi-doc chats
• Added the option to Save message to file
• Added the option to close Welcome Window automatically
• Improvement: re-open the Welcome Window from Dock icon or app menubar

IMPORTANT: This release added the Indexing Strategy option to improve the accuracy of AI answers. Please read the full release note to configure yours.

• Indexing Strategy: choose how to process your documents: smart, fast or accurate
  • Fast: use local AI model, fast and free but less accurate
  • Accurate: use OpenAI model, high accuracy but slower and not free ($1 per 3000 pages)
  • Smart (recommended): use fast indexing for large documents (configurable, default >300 pages)
• App Appearance: customize your app's color scheme, font family, text size & paragraph spacing
• Better keyboard support: ⌘E to Edit Last Message, ⌘R to Regenerate & ⌘⇧C to Copy Last Response
• Support more AI models available in OpenRouter
• Set a custom Finder tag (default: Blue)
• Improve the chat TextField
• Improved the context menu in a multi-doc chat (right-click, start new chat)
• Bug fix: incorrect icon when using an OpenRouter model
• Bug fix: fix the issue where sometimes clicking on page references does not navigate to a new page

Fixed the issue where some users cannot open PDFs from Finder

IMPORTANT: this releases updates the default System Instruction. If you're using a custom prompt, please reset to the new default value. Read guide

• Improve default System Instruction
• Better chat management: you can now Edit / Regenerate / Delete messages
• Allow reprocessing a PDF, including the option to force using OCR
• Improve scanned PDF detection
• General bug fixes & improvements

Build 12 · Aug 25, 2023

• Fixes the streaming issue with Azure OpenAI for some users
• Improves rendering performance for long response

Build 10 · Aug 17, 2023

This release fixes the issue with OpenRouter for some users

Build 9 · Aug 15, 2023

• Web Browsers Integration (Experimental): Added the ability to open a PDF url in browser with PDF Pals using a shortcut key. Read set up guide: https://pdfpals.com/help/how-to-open-pdfs-from-web-browsers
• Minor bug fixes & improvements

Build 8 · Aug 10, 2023

• Fix the issue where the app crashes for some users on Intel-based Macs

Build 7 · Aug 6, 2023

• Fix the issue where the app crashes for some users on macOS 14

Build 6 · Aug 6, 2023

Biggest release yet. This release includes many new features that you will love 🥳

• Chat with multiple PDFs 🎉
• Highlight relevant paragraph in the PDF Viewer when clicking on the referenced page
• You can customize the system instruction template now
• Added support for Azure OpenAI & OpenRouter
• Added the ability to customize OpenAI API host & Organization ID
• Added previous GPT models (gpt-3.5-turbo-0301 etc.)
• Better drag & drop support
  • Drag & drop into the Welcome Window to start a new chat.
  • Drag & drop documents into a Chat Window to add to the conversation
• You can delete a conversation from the Welcome Window now

Build 5 · Jul 28, 2023

This release fixes the issue where some users couldn't activate the license key

Build 4 · Jul 21, 2023

• By default, opening a PDF starts a new chat session instead of opening the last one. You can change this behavior in Preferences
• Added the ability to continue previous chats
• Added the ability to rebuild the similarity index. Click on the Chat Configuration popover > Similarity Index
• Save & restore chat window size
• Fixed the issue where the PDF Viewer disappears for some users

Build 3 · Jul 17, 2023

Hot fix

• Fix the issue where the main window closes quickly after launch for some users

Build 2 · Jul 17, 2023

First Public Release of PDF Pals 🎉

• Chat with a single PDF (chatting with multiple PDFs coming soon)
• Each answer includes referenced pages from the source PDF
• Accepts PDFs in any language. You can change the output language anytime during the conversation.
• Supports scanned PDFs and complex forms using a powerful OCR engine
• Automatically redacts sensitive data from your PDF such as credit card numbers or monetary amounts etc.
• Maximum flexibility: you can switch GPT models and adjust the temperature anytime during the conversation
• Supports Markdown and Code Syntax
• Supports LaTeX and Math Expressions
• Secure: your OpenAI API key is stored in the standard Apple Keychain. Your PDFs stay in your local machine. No uploading required.

Build 1 · Jul 15, 2023

• The MVP Release