74 lines
2.3 KiB
Markdown
74 lines
2.3 KiB
Markdown
# Brand Monitor — Capstone Reference Implementation
|
|
|
|
## What This Is
|
|
|
|
A reference implementation of Capstone Option 1 (Brand Monitor) from the Agentic Engineering Course.
|
|
|
|
This is NOT a production system. It's a teaching tool that demonstrates the concepts from Modules 1-7:
|
|
- Multi-agent architecture with domain locking
|
|
- Security (damage-control rules, L3+)
|
|
- Cost estimation and cascade routing
|
|
- Mental models and agent expertise
|
|
- Observability and evaluation
|
|
|
|
## Structure
|
|
|
|
```
|
|
brand-monitor/
|
|
├── ARCHITECTURE.md # System design document
|
|
├── README.md # This file
|
|
├── config/
|
|
│ ├── brand.yaml # Brand configuration
|
|
│ └── damage-control-rules.yaml # Security rules
|
|
├── agents/
|
|
│ ├── scanner.md # Scanner agent persona
|
|
│ ├── analyzer.md # Analyzer agent persona
|
|
│ └── reporter.md # Reporter agent persona
|
|
├── data/
|
|
│ ├── sample_scan.jsonl # Example scan results
|
|
│ └── sample_analysis.jsonl # Example analysis results
|
|
└── tests/
|
|
└── test_pipeline.py # Test suite (run with pytest)
|
|
```
|
|
|
|
## How to Use
|
|
|
|
```bash
|
|
# 1. Study the architecture
|
|
cat ARCHITECTURE.md
|
|
|
|
# 2. Review agent definitions
|
|
cat agents/scanner.md
|
|
cat agents/analyzer.md
|
|
cat agents/reporter.md
|
|
|
|
# 3. Run the test suite
|
|
python -m pytest tests/test_pipeline.py -v
|
|
|
|
# 4. Compare against your own capstone implementation
|
|
```
|
|
|
|
## Key Lessons Demonstrated
|
|
|
|
| Concept | Where | Lesson |
|
|
|---------|-------|--------|
|
|
| Multi-agent architecture | ARCHITECTURE.md | M4 Orchestration |
|
|
| Domain permissions | Each agent .md file | M4 Domain Locking |
|
|
| Damage control | config/damage-control-rules.yaml | M3 Security |
|
|
| Cost estimation | tests/test_pipeline.py (TestCost) | M6 Economics |
|
|
| Cascade routing | Scanner=Flash, Analyzer=Sonnet, Reporter=Opus | M6 Cascade Routing |
|
|
| Mental models | Agent system prompts | M4 Mental Models |
|
|
| Evaluation | tests/ directory | M6 Evals |
|
|
| Security audit | Config files | M3 Defense-in-Depth |
|
|
|
|
## Cost Summary
|
|
|
|
| Agent | Model | Cost/Run |
|
|
|-------|-------|----------|
|
|
| Scanner | Gemini Flash | ~$0.0009 |
|
|
| Analyzer | Claude Sonnet | ~$0.021 |
|
|
| Reporter | Claude Opus | ~$0.225 |
|
|
| **Total** | | **~$0.247** |
|
|
| Daily (4 runs) | | ~$0.99 |
|
|
| Monthly (30d) | | ~$29.64 |
|