Skip to content

Running Labs in Google Colab

Use Google Colab when you do not want to install Docker or a local Python stack. Each lab notebook can be opened with one click via the Open in Colab badge on the lab pages.

Repository: fischer3-net/accurate_secure_rag_systems

Open a notebook in Colab

Lab Notebook Launch
1.1 Chunking & Metadata lab-1.1-chunking.ipynb Open In Colab
1.2 Hybrid Retrieval lab-1.2-hybrid-rerank.ipynb Open In Colab
2.1 Storage Benchmark lab-2.1-benchmark.ipynb Open In Colab
2.2 Graph-Augmented RAG lab-2.2-graph-rag.ipynb Open In Colab
3.1 Modular Skills lab-3.1-modular-skills.ipynb Open In Colab
3.2 Skill Router lab-3.2-skill-router.ipynb Open In Colab
4.1 Golden Dataset lab-4.1-golden-dataset.ipynb Open In Colab
4.2 CI Quality Gate lab-4.2-ci-gate.ipynb Open In Colab

The badge opens the notebook from the main branch on GitHub. If your fork uses another branch or org name, change the URL accordingly.

First-time setup in Colab

Notebooks include a Colab bootstrap cell near the top. Run it first. It will:

  1. Clone this repository into /content/accurate_secure_rag_systems (if needed)
  2. Change into the correct lab directory
  3. Install lightweight offline dependencies (pydantic, langchain-text-splitters, pytest, …)
  4. Put the lab src/ package on sys.path

If a notebook is older and missing that cell, paste this at the top and set LAB_DIR for the lab you are running:

# --- Colab bootstrap (paste if missing) ---
import os, sys
from pathlib import Path

REPO_URL = "https://github.com/fischer3-net/accurate_secure_rag_systems.git"
REPO_ROOT = Path("/content/accurate_secure_rag_systems")
LAB_DIR = "labs/01-chunking"  # change per lab: 02-storage, 03-skills, 04-evaluation

if not REPO_ROOT.exists():
    !git clone --depth 1 {REPO_URL} {REPO_ROOT}

LAB = REPO_ROOT / LAB_DIR
os.chdir(LAB)
sys.path.insert(0, str(LAB))
# Week 2+ may also need Week 1 on the path:
sys.path.insert(0, str(REPO_ROOT / "labs" / "01-chunking"))

%pip install -q pydantic python-dotenv langchain-text-splitters langchain-core pytest pyyaml pandas

print("LAB =", LAB)
print("cwd =", Path.cwd())

Then run the remaining cells as usual.

What works offline in Colab

Feature Colab without GCP Notes
Document-aware chunking, metadata Yes Lab 1.1
Hybrid BM25 + dense + RRF Yes Lab 1.2 (hashing embedder fallback)
Storage benchmark / graph RAG Yes In-memory stores
Skills + router Yes Deterministic skills
Golden dataset + eval gate Yes Lab 4
Live Vertex AI / BigQuery writes Optional Requires GCP project + auth

Optional GCP in Colab:

from google.colab import auth
auth.authenticate_user()
!gcloud config set project YOUR_PROJECT_ID

Saving your work

  • Copy to Drive: Colab menu → File → Save a copy in Drive
  • Download: File → Download → .ipynb
  • Changes in /content/... are not pushed back to GitHub unless you configure git credentials and push from a cell

Troubleshooting

Symptom Fix
ModuleNotFoundError: src Re-run the bootstrap cell; confirm LAB points at the lab folder that contains src/
FileNotFoundError for data/... Bootstrap must os.chdir(LAB) so relative data/ paths resolve
Badge 404 Repo must be public (or you must be logged into GitHub with access); path must match main
Private fork Open Colab → File → Upload notebook, or use a GitHub token to clone your fork URL in the bootstrap cell
Kernel restart Re-run the bootstrap cell before other cells