Add portfolio demo: README, Streamlit app, synthetic data, CI

[Copilot]

Transforms patient_analyzer into a reproducible portfolio project with an interactive demo, synthetic data, and CI smoke tests. Uses lightweight TF-IDF embeddings to avoid large model downloads in CI.

Changes

README.md

• TL;DR, problem statement, quickstart (venv + pip), demo instructions
• Privacy statement emphasizing 100% synthetic data
• Screenshot of running demo

demo/

• sample_patients.csv: 10 synthetic patients with age, gender, diagnosis_code, text_note
• app.py: Streamlit app with TF-IDF embeddings, nearest neighbor search, 2D PCA visualization
  • Comments indicate where to plug sentence-transformers/Bio_ClinicalBERT for production

tests/smoke_test.py

• 6 tests covering data loading, embedding computation, neighbor search, PCA reduction, full pipeline

.github/workflows/ci.yml

• Installs minimal deps, runs pytest smoke tests
• Explicit permissions block for security

requirements.txt

• Minimal deps: numpy, pandas, scikit-learn, streamlit, matplotlib
• Heavy deps (torch, transformers) commented as optional

Screenshot

Usage

pip install -r requirements.txt
streamlit run demo/app.py

Do not merge without owner approval.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• checkip.amazonaws.com
  • Triggering command: /home/REDACTED/.local/bin/streamlit streamlit run demo/app.py --server.headless true --server.port 8501 (dns block)
  • Triggering command: /home/REDACTED/.local/bin/streamlit streamlit run demo/app.py --server.headless true --server.port 8502 (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

Original prompt

Goal: Improve patient_analyzer to be a showcase portfolio project by adding a clear README, a small interactive demo (Streamlit app) that visualizes patient embeddings and nearest neighbors using synthetic data, a sample dataset, and a CI smoke test that runs the demo script headlessly. Create branch portfolio/docs-patient-analyzer and open PR to main. Do not merge.

Required changes:

1. README.md - TL;DR, problem statement (comparing patients in embedding space), quickstart (venv + pip install -r requirements.txt), how to run the Streamlit demo locally, privacy & data-use statement, expected outputs screenshot placeholder, license.
2. requirements.txt - minimal deps: "numpy", "pandas", "scikit-learn", "streamlit", "sentence-transformers" (or omit heavy transformer and instead include a synthetic embedding generator). To keep CI light, the smoke test will not download large models.
3. demo/sample_patients.csv - small synthetic CSV with 10 patients and a few features (age, gender, diagnosis_code, text_note). Add a script to generate synthetic embeddings for them.
4. demo/app.py - Streamlit app that reads sample_patients.csv, computes simple TF-IDF or random embeddings (to avoid large model download in CI), shows a selected patient and top-3 nearest patients with a simple visualization (table and 2D scatter via PCA). Include comments on where to plug real embeddings (sentence-transformers) in production.
5. tests/smoke_test.py - script that runs python -m demo.app (or uses streamlit to run in headless mode) or simply imports app functions and runs a small function to compute neighbors; this should be invoked by CI to ensure demos run.
6. .github/workflows/ci.yml - runs tests/smoke_test.py; installs requirements and executes the test.

PR description: Add README, Streamlit demo with synthetic data, and CI smoke test to make patient_analyzer reproducible and demonstrable for portfolio. Emphasize synthetic data & privacy. Do not merge without owner approval.

Notes: Keep CI lightweight and avoid downloading large transformer models. Where heavy models are helpful, add instructions in README for optional steps ('extras' extras_require)."

This pull request was created as a result of the following prompt from Copilot chat.

Goal: Improve patient_analyzer to be a showcase portfolio project by adding a clear README, a small interactive demo (Streamlit app) that visualizes patient embeddings and nearest neighbors using synthetic data, a sample dataset, and a CI smoke test that runs the demo script headlessly. Create branch portfolio/docs-patient-analyzer and open PR to main. Do not merge.

Required changes:

1. README.md - TL;DR, problem statement (comparing patients in embedding space), quickstart (venv + pip install -r requirements.txt), how to run the Streamlit demo locally, privacy & data-use statement, expected outputs screenshot placeholder, license.
2. requirements.txt - minimal deps: "numpy", "pandas", "scikit-learn", "streamlit", "sentence-transformers" (or omit heavy transformer and instead include a synthetic embedding generator). To keep CI light, the smoke test will not download large models.
3. demo/sample_patients.csv - small synthetic CSV with 10 patients and a few features (age, gender, diagnosis_code, text_note). Add a script to generate synthetic embeddings for them.
4. demo/app.py - Streamlit app that reads sample_patients.csv, computes simple TF-IDF or random embeddings (to avoid large model download in CI), shows a selected patient and top-3 nearest patients with a simple visualization (table and 2D scatter via PCA). Include comments on where to plug real embeddings (sentence-transformers) in production.
5. tests/smoke_test.py - script that runs python -m demo.app (or uses streamlit to run in headless mode) or simply imports app functions and runs a small function to compute neighbors; this should be invoked by CI to ensure demos run.
6. .github/workflows/ci.yml - runs tests/smoke_test.py; installs requirements and executes the test.

PR description: Add README, Streamlit demo with synthetic data, and CI smoke test to make patient_analyzer reproducible and demonstrable for portfolio. Emphasize synthetic data & privacy. Do not merge without owner approval.

Notes: Keep CI lightweight and avoid downloading large transformer models. Where heavy models are helpful, add instructions in README for optional steps ('extras' extras_require)."

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[edgarbc]

lgtm