Architecture Review Board
Multi-agent design review with structural engineering, cost estimation, and sustainability consulting — orchestrator skills and rag_augmented memory across design iterations.
What this demonstrates
Three GenericAgent instances forming a virtual design review board:
structural engineering, cost estimation, and sustainability consulting.
Cross-agent orchestrator skills run sequential pipelines with each agent
seeing prior output. Backed by rag_augmented memory with a lower
similarity threshold (0.45) to catch terminology variants across disciplines.
| Feature | How It's Used |
|---|---|
| rag_augmented memory | Design decisions and material comparisons retrieved from Qdrant |
| Orchestrator skills | full_design_review (structural → cost → sustainability), material_selection |
| Lower similarity threshold | 0.45 catches "CLT" ↔ "cross-laminated timber" ↔ "mass timber" variants |
| Structured JSON entities | Materials, jurisdictions, certifications tracked with deduplication |
| Per-agent prompts | Domain-specific summary_compression_* for engineering, cost, and sustainability |
Run it
ORCHID_CONFIG=examples/architecture-review/orchid.yml uvicorn orchid_api.main:app
orchid chat interactive --config examples/architecture-review/orchid.yml
orchid config validate examples/architecture-review/agents.yamlAgents and tools
| Agent | Tools | Expertise |
|---|---|---|
structural | analyze, code compliance, compare materials, fire strategy | Load paths, Eurocode/IBC/ASCE, fire safety |
cost | construction estimate, lifecycle costs, market rates | Budgets, LCC, regional labour/materials |
sustainability | certification, embodied carbon, strategies, carbon compare | BREEAM/LEED/DGNB, carbon accounting |
Orchestrator skills
skills:
full_design_review: # structural → cost → sustainability
material_selection: # structural → sustainabilityMemory config
supervisor:
memory:
strategy: "rag_augmented"
rag_k: 6
rag_similarity_threshold: 0.45 # lower: catches terminology variants
structured_output: true
truncation_strategy: "middle"Simulated flow
"Review a 6-story office in Berlin, 8000m², LEED Gold."
→ full_design_review skill fires
→ structural: steel + CLT recommended, Eurocode 3/5 compliant
→ cost: €22.4M, CLT 30% cheaper lifecycle vs steel-concrete
→ sustainability: LEED Gold achievable, 40% below baseline carbon
"Switch to all-timber instead."
→ material_selection skill fires
→ structural: CLT+glulam viable to 12m with fire encapsulation
→ sustainability: carbon-negative (-210 kgCO2/m³ net), BREEAM Outstanding possible
→ RAG retrieves past cost discussion from earlier design iterationFiles
examples/architecture-review/
├── README.md
├── orchid.yml
├── agents.yaml
└── tools/
├── structural.py # 6 materials, 6 building codes
├── cost.py # 3 building types, 5 regions
└── sustainability.py # 4 certifications, 7 strategies