feat: 9 new waste detectors + extended CSV data + AI dashboard (dashboard_AI.py)

[smahima27]

Summary

Adds 9 new AWS waste detection rules, extended sample data, and an AI-powered dashboard — built during the Perforce Global Jam 2026 hackathon.

---

New detection rules (detection_engine.py)

#	Function	What it detects	Est. saving
1	detect_nat_idle	NAT Gateways with <1 GB/week traffic ($32/mo base charge)	Full monthly cost
2	detect_idle_load_balancers	ALBs/NLBs with 0 target groups or 0 requests	Full monthly cost
3	detect_old_gen_instances	t2/m4/c4/r4 families — recommend upgrade to t3/m5/c5/r5	4–11% per instance
4	detect_orphan_snapshots	EBS snapshots >90 days with no source volume	Full monthly cost
5	detect_gp2_volumes	gp2 volumes eligible for gp3 migration (zero downtime)	20% per volume
6	detect_ondemand_no_coverage	EC2 on-demand >30 days without Reserved Instance or Savings Plan	~40%
7	detect_infinite_log_retention	CloudWatch log groups with no expiry policy set	~70%
8	detect_stopped_ec2_with_ebs	Stopped EC2 still incurring EBS storage costs	Full EBS cost
9	detect_underutilized_cache_redshift	ElastiCache/Redshift clusters at <10% CPU	Full monthly cost

Bug fix: detect_idle_ec2 now excludes stopped instances to prevent double-counting with rule #8.

---

Sample data (aws_cost_data.csv)

Extended from 60 to 98 rows. Added 38 new rows covering:

• 3 NAT Gateway rows (low traffic: 0.05–0.3 GB/week)
• 3 ALB/NLB rows (0 requests or 0 target groups)
• 4 old-gen EC2 rows (t2.large, m4.xlarge, c4.4xlarge, r4.2xlarge)
• 5 orphaned EBS snapshots (90–200 days old)
• 5 gp2 EBS volumes (100–1000 GB)
• 4 on-demand EC2 (running 180–365 days without RI/SP)
• 5 CloudWatch log groups (infinite retention)
• 4 stopped EC2 (paying for attached EBS only)
• 4 ElastiCache/Redshift clusters (1.8–4.2% CPU)

---

New file: dashboard_AI.py

Enhanced Streamlit dashboard featuring:

• FinOps AI chatbot (Claude claude-sonnet-4-20250514) embedded in a side panel — answers questions about the actual findings data
• Slack webhook integration — fire a top-finding alert to any channel
• Executive summary + category donut chart
• Filterable findings cards with remediation CLI commands toggle

Run with:

export ANTHROPIC_API_KEY='sk-ant-...' && streamlit run dashboard_AI.py

---

Detection results after changes

Total findings    : 77
Monthly waste     : $7,686.55
Annual waste      : $92,238.60

Waste by category:
  Idle Resource             $5,769.20/mo
  Zombie Resource           $1,012.60/mo
  RI/SP Optimisation        $507.00/mo
  Storage Optimisation      $210.00/mo
  Old Generation            $116.35/mo
  Log Retention             $71.40/mo