Select llama.cpp GGUF on CPU Basic Spaces

Default extraction to the fine-tuned MiniCPM-V Hub checkpoint.

Expand lab knowledge graph to 107 markers with curated YouTube videos.

Polish upload preview and deployment dependencies

Route ZeroGPU auto backend to Transformers

Fall back from Transformers to llama.cpp

Select extraction backend by CUDA availability

Switch extraction to PDF text only

Use prebuilt llama.cpp wheel on Space

Switch llama.cpp Space back to CPU wheel

Try cu121 llama-cpp wheel lane

Switch llama-cpp wheel lane to cu125

Use cu132 llama-cpp wheel index

Reduce Space build time

Fix llama.cpp JSON extraction path

Make llama.cpp ZeroGPU backend default

Wire llama.cpp ZeroGPU badge path

Switch Space deployment to ZeroGPU

fix(train): use MiniCPM-V 4.6 model id and official mmproj

chore: .gitignore + RUNBOOK