Career Tree RAG (Microsoft AI Innovators Summer Internship)

Are you feeling lost in the world of AI? This is a local career-exploration RAG project, which can help you find your career path. We meticulously designed career tree ourselves which you can check from https://www.ibz04.pro/blog/career-tree.

---

Requirements

• Python 3.11+
• Foundry Local (for chat and CV analysis; not needed for indexing)
• Network once to download the embedding model and Foundry model

Quick start

git clone <repo-url>
cd Lost   # or your local folder name

chmod +x setup.sh index.sh ask.sh retrieve.sh regenerate.sh recommend.sh ui.sh _env.sh
./setup.sh
./index.sh
./ui.sh             # http://127.0.0.1:8501
# or: ./ask.sh --interactive

Scripts activate the venv themselves via _env.sh.

What each command does

Command	Purpose
./setup.sh	Create venv and install dependencies
./index.sh	Build the search index (required before asking)
./ui.sh	Open the Streamlit UI
./ask.sh "…"	Ask a question in the terminal
./recommend.sh --cv path.pdf	Analyze a CV
./retrieve.sh "…"	Search only (no LLM)
./regenerate.sh	Re-download the career tree from the website (overwrites local edits to data/career-tree.json)

UI

1. Career Chat — ask career questions
2. CV Analysis — upload a CV → structured JSON
3. Indexing — rebuild the search index

First chat answer can take 1–3 minutes while the local model loads.

Models

Model	Role
all-MiniLM-L6-v2	Embeddings (Hugging Face)
qwen2.5-0.5b	Default local LLM (Foundry). Small/fast; use --model <alias> for a stronger one

Add content

• Articles: put .md files in content/extra/, then ./index.sh
• Career roles: edit data/career-tree.json, then ./index.sh (do not run ./regenerate.sh after that, or your edits will be wiped)

Notes

• Venv and Chroma live under ~/.career_tree_rag/ (avoids iCloud Desktop freezes)
• CV parsing stays local; cache is in data/cv_cache/ (gitignored)
• License: MIT — see [LICENSE](LICENSE)