Small local demos for OCR, text-to-speech, semantic PDF search, and Streamlit data browsing.
Journalists often deal with scanned texts, but the best tools are locked behind code. How can we help non-technical users try out our favorite libaries without forcing them through installing Python and running notebooks?
Newsrooms' C-suites have loved auto-generated podcasts recently, but the rest of the publication is often split. How can we let everyone have a hand in demoing the product to show its strengths and weaknesses?
A larger TTS demo with Kokoro ONNX and MMS options.
Semantic search is a useful tool for investigative work, but you don't always want to upload all of your docs into a Google product. Can a home-grown version work just as well?
A notebook walkthrough that reads local PDFs, embeds each page, and ranks semantic search results. This will only work on codespaces!
Semantic search over local PDFs using sentence-transformer embeddings.
Tired of doing data analysis for all of your coworkers? Give them the tools to browse the data directly themselves!
A notebook that opens up a CSV and does a little light analysis.
A simple Streamlit browser for local property transfer data.
A fuller Streamlit transfer-data app with charts, filters, and summaries.
Use Braintrust for evaluation workflows, then use the CSV files here as small local datasets to test with.
Download the CSVs from the section's Download materials link.
Use Pydantic AI with OpenRouter to ask questions and request structured outputs.