Files
Michele Dolfi 19af03f539 feat: Support for DeepSeek-OCR in VLM pipeline (#2798)
* add parsing of annotated markdown and definition of new ResponseFormat for the VLM pipeline

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix broken html in test

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* update result with initial text

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* move parsing to vlm pipeline

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* restore md from main

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* process table structure

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* simplify and refactor

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* factor out deepseekocr utils

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* renaming

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* refactor common logic in vlm parsing logic

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add deepseek-ocr with ollama

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* update tests for new annotation format

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix parsing of title

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* more test data

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add picture item

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix bbox parsing

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* remove old tests

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add test parsing deepseek md

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* rename test

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* add test with ollama conversion

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

* fix test and mark methods as private

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>

---------

Signed-off-by: Michele Dolfi <dol@zurich.ibm.com>
2026-01-09 18:42:40 +01:00
..