v7.5.5 - THINKING models

This section tracks minor updates following the 7.5.0 release.

Version	Changes
7.5.1	Fixed setup_windows.py to install the correct version of tesseract based on whether a user is using Python 3.11 or Python 3.12. Previously, the 3.11 version was always installed leading to installation errors for Python 3.12 users.
7.5.2	Added monkey patch to module_ocr.py unless/until ocrmypdf library implements PR 1493.
7.5.3	Reverted to CUDA 12.4.1 to be more compatible with flash attention 2 and revised setup_windows.py to clear the cache of compiled code that triton creates, which was necessary b/c triton changed their sourcecode.
7.5.4	Fixed set_cuda_paths yet again b/c triton...and added QwQ - 32b chat model.
7.5.5	Added the following chat models: Qwen - 3b, Exaone Deep - 2.4b, Olympic Coder - 7b, Exaone Deep - 7.8b, OLMo 2 - 13b, Reka Flash - 21b, Exaone Deep - 32b, Olympic Coder - 32b Added more python libraries to scrape. Updated Jeeves knowledge. Updated dependencies. Refactored chat module.

The full release notes for v7.5.0 are below.

Vision Models

• Basically every vision model was revamped to use less VRAM and produce higher quality results. This was a monumental effort. Also, the following were added or removed:

Change	Details	Notes
Added	Ovis2 - 1b	Supersedes Ovis1.6-Llama3.2 - 3b
Added	Ovis2 - 2b	Supersedes Ovis1.6-Llama3.2 - 3b
Added	Qwen VL - 3b	Excellent new model
Added	Qwen VL - 7b	Excellent new model
Added	Granite Vision - 2b	Excellent new small vision model
Removed	Moondream2 - 1.9b	Outdated source code in the parent repository
Removed	Mississippi - 2b	Superseded by more recent, higher-quality models
Removed	Ovis1.6-Llama3.2 - 3b	Superseded by Ovis2 models

Chat Models

Change	Details	Notes
Added	granite-3.2-8b-instruct	Supersedes granite-3.1-8b-instruct
Removed	Mistral Small - 22b	Superseded by Mistral Small - 24b, which was previously added

Optical Character Recognition

• MASSIVE effort to improve OCR, which was ultimately successful. OCR text is now embedded into a new .pdf, which is saved to same directory as the original .pdf instead of creating a .txt file.
• Temporarily removed the GOT_OCR backend due to excessive VRAM usage – may be reinstated later upon further testing.

General Improvements

• Updated scraping libraries across the board.
• Updated all dependencies.
• Added new AMD GPUs to the GPU comparison chart.
• Improved the metrics bar to run smoother.
• Updated Jeeves knowledge-base

Installation

• Just follow the installation instructions on the github readme.

Upgrading from a Prior Version

INSTRUCTIONS FOR UPGRADING FROM A PRIOR VERSION

To upgrade from a prior version while still keeping your databases and models that were downloaded, perform the following steps:

• In your directory that contains all of the files currently...delete the Include, Lib, Scripts folders. (pertain to your virtual environment)
• Delete the pyvenv.cfg file. (pertains to your virtual environment)
• At this point, any and all files pertaining to your virtual environment are gone.
• Delete the Assets and CSS folders. (new ones will be provided in their entirety further below)
• Extract all the files from the .zip file for this release v7.5.0 and go into the src folder.
• Copy all files and folders EXCEPT THE CONFIG.YAML FILE into the aforementioned folder where you virtual environment USED TO BE. If there are any prompts indicating that files will be overwritten, CLICK YES since we're only updating necessary files now.
• EXCEPT THE CONFIG.YAML FILE!
• EXCEPT THE CONFIG.YAML FILE!
• last time...EXCEPT THE CONFIG.YAML FILE!

Afterwards, open a command prompt and run the installation instructions on this repository's page, namely:

python -m venv .

.\Scripts\activate

python setup_windows.py