Streamlit web app over the pipeline (issue #5)

.gitignore

@@ -0,0 +1,27 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+.pytest_cache/
+
+# Virtual environments
+.venv/
+venv/
+env/
+
+# Streamlit secrets — never commit API keys
+.streamlit/secrets.toml
+
+# Local Chroma / embeddings persistence
+.chroma/
+chroma/
+*.sqlite3
+
+# Jupyter
+.ipynb_checkpoints/
+
+# OS / editor cruft
+.DS_Store
+Thumbs.db
+.vscode/
+.idea/

.streamlit/config.toml

@@ -0,0 +1,10 @@
+# Neo Kinpaku product theme (adapted from the impeccable design system):
+# warm-dark lacquer surfaces, kinpaku gold as the single restrained accent,
+# neutral near-white text. Never pure black/white; flat with hairline borders.
+[theme]
+base = "dark"
+primaryColor = "#F4B93C"              # kinpaku gold — primary action / active only
+backgroundColor = "#0F0E0B"           # lacquer ground (warm, not pure black)
+secondaryBackgroundColor = "#181714"  # raised lacquer — sidebar / panels / cards
+textColor = "#DBDBDB"                 # neutral near-white body
+font = "sans serif"

README.md

@@ -0,0 +1,45 @@
+# AI Résumé Builder
+
+Tailors a résumé to a specific job using **retrieval-grounded generation**: it matches your real
+experience to each job requirement, so it reframes what you actually have instead of inventing new
+experience. Built for the AI-201 workshop, now with a web UI.
+
+Works with **Anthropic, Gemini, DeepSeek, or Groq** — pick one; embeddings run locally, so retrieval
+is identical whichever you choose.
+
+## Web app (Streamlit)
+
+```bash
+pip install -r requirements.txt
+streamlit run app.py
+```
+
+Then set an API key one of three ways (in order of preference):
+
+1. **Streamlit secrets** — create `.streamlit/secrets.toml`:
+   ```toml
+   ANTHROPIC_API_KEY = "sk-ant-…"   # or GOOGLE_API_KEY / DEEPSEEK_API_KEY / GROQ_API_KEY
+   ```
+2. **Environment variable** — `export ANTHROPIC_API_KEY=…`
+3. **Sidebar field** — paste it in the running app (session-only; never committed).
+
+Deploys as-is to **Streamlit Community Cloud** or **Hugging Face Spaces** (add the key as a secret there).
+
+### What you get
+- A grounded, **source-tagged** tailored résumé (`[SOURCE: RESUME | GITHUB | REFRAMED]`).
+- A **coverage report** — which requirements matched real experience, and which are honest gaps.
+- An independent **fabrication check** and a deterministic **diff** against your original.
+
+### Architecture
+- **`pipeline.py`** — provider-agnostic pipeline (LLM client, scraping, chunking, retrieval, generation,
+  eval, fabrication check). No Streamlit or Colab dependency; importable anywhere.
+- **`app.py`** — the Streamlit UI over that pipeline.
+
+## Notebooks (workshop)
+- `AI_Resume_Builder_v1.ipynb` — the basics: a single tailoring prompt.
+- `AI_Resume_Builder_v2_Phase2.ipynb` — production patterns: RAG, a multi-tool agent, and safety guardrails.
+
+Open them in Colab and add your provider's API key to the Secrets tab (🔑).
+
+## License
+GNU GPL v3 — see [LICENSE](LICENSE).