AI-Ready Analytics Platform — built for the CMRIT FDP "Agentic AI: Developing Intelligent Agents with Modern AI Frameworks" (Day 5: Applications, Deployment & Future Trends).
Upload any Excel workbook → get automatic profiling, KPIs, interactive charts, forecasts, anomaly detection — and, with a Groq API key, an agentic AI analyst that plans tool calls, inspects your data, and answers business questions.
Works with or without an AI key. No key → 🟡 Analytics Mode (everything except AI narration). Key present → 🟢 AI Mode (Groq Llama 3.3 70B).
┌──────────────────────────────────────────────┐
│ Streamlit UI │
│ Overview │ KPIs │ Charts │ AI Summary │ Ask │
└──────┬───────────────────────────┬───────────┘
│ │
┌──────────▼──────────┐ ┌──────────▼──────────┐
│ Analytics Engine │ │ Agentic AI Layer │
│ pandas / numpy │◄────┤ ReAct tool loop │
│ profiling, KPIs, │tools│ LangChain + Groq │
│ forecast, anomalies │ │ llama-3.3-70b │
└──────────┬──────────┘ └──────────┬──────────┘
│ │ (optional)
┌──────▼───────┐ ┌──────▼───────┐
│ Excel Upload │ │ Groq API │
│ / Demo Data │ └──────────────┘
└──────────────┘
Git push ──► GitHub repo ──► ArgoCD (auto-sync) ──► Kubernetes
▲ │
└── self-heal ◄──────┘
docker build ──► Docker Hub (appars/excel-insightforge-agent)
The agentic part: the "Ask Data" tab runs a minimal ReAct loop. The LLM receives a tool catalog (get_kpis, group_aggregate, trend_and_forecast, anomalies, declining_products…), decides which to call, observes JSON results, iterates up to 5 steps, then answers — with the full trace visible in the UI.
git clone https://github.com/appars/excel-insightforge-agent
cd excel-insightforge-agent
python3.11 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.py
# → http://localhost:8501Click Generate Demo Dataset for the built-in Global E-Commerce Analytics workbook (50k orders, 10k customers, 500 products, marketing + supply-chain sheets with planted insights).
pip install pytest
pytest -qGet a free key at https://console.groq.com/keys. Priority order:
- Sidebar API key field (highest)
.envfile →cp .env.example .envand setGROQ_API_KEY=gsk_...- Environment variable →
export GROQ_API_KEY=gsk_... - Nothing → 🟡 Analytics Mode (never crashes, gracefully degrades)
| Mode | Indicator | Features |
|---|---|---|
| Analytics | 🟡 | Upload, profiling, KPIs, charts, forecast, anomalies, rule-based summary |
| AI | 🟢 | Everything above + LLM executive summary + agentic Q&A with tool trace |
# Build
docker build -t appars/excel-insightforge-agent:latest .
# Run (Analytics Mode)
docker run -p 8501:8501 appars/excel-insightforge-agent:latest
# Run (AI Mode)
docker run -p 8501:8501 -e GROQ_API_KEY=gsk_xxx appars/excel-insightforge-agent:latest
# Push to Docker Hub
docker login
docker push appars/excel-insightforge-agent:latestImage highlights: python:3.11-slim, non-root user, layer-cached deps, built-in HEALTHCHECK on Streamlit's /_stcore/health.
- Install Rancher Desktop → enable Kubernetes (containerd or dockerd).
- Verify:
kubectl get nodesshowslima-rancher-desktop(or similar) Ready. - NodePort services are reachable at
localhost:<nodePort>.
# Optional: real Groq secret (skip → Analytics Mode)
kubectl create secret generic groq-secret --from-literal=GROQ_API_KEY=gsk_xxx
kubectl apply -f k8s/
kubectl get pods -l app=excel-insightforge-agent # 2 replicas Running
# Access
open http://localhost:30851 # NodePort on Rancher Desktop
# or: kubectl port-forward svc/excel-insightforge-agent 8501:80Manifests include: 2 replicas, readiness + liveness probes on /_stcore/health, resource requests/limits, optional secret injection.
⚠️ k8s/secret.yamlis a template with an empty key — never commit real secrets. Create them withkubectl create secretor a sealed-secrets/SOPS workflow.
# Install ArgoCD
kubectl create namespace argocd
kubectl apply -n argocd -f https://raw.githubusercontent.com/argoproj/argo-cd/stable/manifests/install.yaml
# UI access
kubectl port-forward svc/argocd-server -n argocd 8080:443
# password:
kubectl -n argocd get secret argocd-initial-admin-secret -o jsonpath="{.data.password}" | base64 -d
# Register the app
kubectl apply -f argocd/application.yamlArgoCD now watches github.com/appars/excel-insightforge-agent (path k8s/, branch main) with auto-sync + self-heal + prune.
Live demo moment 🎬: edit k8s/deployment.yaml in GitHub (e.g. replicas: 2 → 3), commit, and watch ArgoCD detect, sync, and roll out — no kubectl needed. Then try kubectl scale deploy excel-insightforge-agent --replicas=1 and watch self-heal revert it. That's GitOps.
curl -fsS http://localhost:8501/_stcore/health # → "ok"Used by Docker HEALTHCHECK and both K8s probes.
| Symptom | Fix |
|---|---|
ImagePullBackOff |
Image not pushed / wrong name → docker push appars/excel-insightforge-agent:latest |
Pod OOMKilled |
Raise memory limit (demo dataset needs ~1Gi) |
| Probes failing | Check kubectl logs; Streamlit needs ~10–20 s to boot |
ArgoCD Unknown/ComparisonError |
Repo URL/branch/path wrong, or repo is private (add repo credentials in ArgoCD) |
| 🟡 mode despite secret | Secret name/key must be groq-secret / GROQ_API_KEY; restart pods after creating it |
| AI errors / rate limits | App auto-falls back to rule-based summary — demo never breaks |
| NodePort unreachable | kubectl port-forward svc/excel-insightforge-agent 8501:80 |
 |
 |
 |
 |
(placeholders — add after first run)
excel-insightforge-agent/
├── app.py # Streamlit UI (tabs, modes, sidebar)
├── config.py # Hybrid API-key resolution, logging
├── services/
│ ├── analytics.py # Profiling, KPIs, forecast, anomalies, rule-based summary
│ ├── ai_service.py # Groq LLM + agentic ReAct tool loop
│ ├── dataset_generator.py # 50k-row demo dataset with planted insights
│ └── visualization.py # Plotly figure builders
├── tests/test_app.py # Unit + smoke tests
├── k8s/ # deployment / service / secret manifests
├── argocd/application.yaml # GitOps app (auto-sync, self-heal)
├── Dockerfile # python:3.11-slim, non-root, healthcheck
├── requirements.txt
└── .env.example
MIT — built for teaching. Reuse freely in your FDP sessions.