I’ve made a searchable version of the handwritten Chandra Shum Shere catalogue at Oxford, with linked folio images and subject filters (~7000 MSS in all):

https://suhasmahesh.com/oxford-css-manuscripts/

Would you mind sharing this on Indology list?

The GPT-6 Astra AI model performed the bulk of the image analysis. Although caution about AI in general is warranted, current AI models are extremely reliable at recognizing handwritten Roman text. I spent significant compute resources verifying the results by pitting various models against each other. The original images are linked on the website for each entry. I've run spot checks on a hundred entries, and found the results accurate enough to trust.

Best,
Suhas