The complete, consolidated architecture for an AI-powered Architecture Intelligence Platform. 8 layers, 51 components, 5 intelligence engines, enterprise-grade B2B.
ArchPilot is NOT a chatbot. NOT a documentation tool. It's a decision intelligence layer that ingests meetings, diagrams, code, and ADRs — and provides expert-level review, best practices, risk analysis, and visual editing across all of them.
Real-time audio capture → STT → contextual AI analysis → live suggestions during technical discussions. Understands decisions as they happen. Works with ANY audio app.
Upload any diagram (PNG/PDF/Draw.io/Excalidraw/Mermaid/IaC) → AI parses into graph → detects 40+ anti-patterns → suggests replacements with pros/cons/risk/cost → visual editing → version comparison.
Connect GitHub/GitLab → Tree-sitter AST parsing → detect patterns in actual code → compare decisions vs implementation → flag drift → PR-level architecture review.
Auto-generate ADRs from meetings → review uploaded ADRs (12-criteria scoring) → cross-reference for conflicts → detect superseded decisions → maintain living repository.
Living knowledge base across 11 domains. Surfaces contextually in meetings, diagrams, code reviews, ADRs. Every recommendation includes: pros, cons, risks, confidence, cost impact, effort, reversibility.
Live or recorded
7 formats + IaC
GitHub/GitLab
MD/PDF/Confluence
AWS, Azure, GCP, multi-cloud, hybrid
Monolith, microservices, serverless, CQRS, DDD
Design patterns, SOLID, clean architecture
SQL/NoSQL, lakes, streaming, data mesh
REST, GraphQL, gRPC, WebSocket
CI/CD, K8s, Terraform, GitOps
Zero trust, OAuth, encryption, compliance
SPA, SSR, micro-frontends, state mgmt
Native, cross-platform, offline-first
Model serving, MLOps, RAG, vector DBs
Kafka, event sourcing, saga patterns
Every component in the system. Original 26 + 25 new from gap analysis. ■ NEW = added from architecture review to resolve gaps.
How data moves through the system. Click ▶ to animate each flow.
~5ms
Check/Queue
~200ms
Process
~50ms
~150ms
~50ms
pgvector
1-3s
Zod
~100ms
~50ms
7 formats
Auto
Vision AI
Nodes+Edges
40+ rules
Pros/Cons/Risk
React Flow
8 dimensions
OAuth
AST Parse
Services/APIs
60+ checks
Decision vs Code
PR Comments
Discussion
Decisions
Draft
12 criteria
Conflicts
Approve
Repository
Periodic
From meeting
Standards
Blast radius
Notification
Workflow
Track PRs
Verify
How each intelligence pillar works under the hood.
| Format | Parsing Strategy | Output |
|---|---|---|
| PNG/JPG/Screenshot | GPT-5.2 Vision → component extraction + OCR | Structured JSON graph |
| PDF-to-image → Vision pipeline | Structured JSON graph | |
| Draw.io (.xml) | XML parsing → mxCell extraction | Native graph with metadata |
| Excalidraw | JSON parsing → elements + arrows | Positional graph |
| Mermaid (.md) | Mermaid parser → AST | Semantic graph |
| SVG | SVG DOM + AI label extraction | Structured graph |
| Terraform/CDK/Pulumi | IaC parser → resource mapping | Infrastructure graph |
3-5 alternatives
Pros/Cons/Risk/Cost
Connections
Impact on all
| Dimension | Version A (Original) | Version B (AI Optimized) | Version C (Cost Optimized) |
|---|---|---|---|
| Reliability | 62/100 | 89/100 | 74/100 |
| Scalability | 45/100 | 92/100 | 71/100 |
| Security | 58/100 | 85/100 | 60/100 |
| Monthly Cost | $2,400 | $4,100 | $1,800 |
| Complexity | Low | Medium | Low |
| Migration Effort | — | 3-4 weeks | 1-2 weeks |
| Anti-Patterns | 7 | 1 | 3 |
| Compliance | No | SOC2+HIPAA | SOC2 only |
Module boundaries, dependency graphs, "is this actually a monolith disguised as microservices?"
REST endpoints, GraphQL schemas, gRPC protos. Validates versioning, error handling, auth.
ORM usage, N+1 queries, missing indexes, connection pooling, transaction boundaries.
Try/catch coverage, retry logic, circuit breakers, timeouts, graceful degradation.
Auth middleware, input validation, SQL injection, XSS, hardcoded secrets, CORS.
Repository, factory, strategy, CQRS, event sourcing, DDD aggregates. Flags misuse.
| Decision (Meeting/ADR) | Code Reality | Drift Type | Severity |
|---|---|---|---|
| "Use event-driven for orders" | Synchronous HTTP calls found | Architecture Drift | Critical |
| "All services need circuit breakers" | 3 of 8 services missing | Implementation Gap | High |
| "Use PostgreSQL for user data" | PostgreSQL confirmed ✓ | Aligned | None |
| "API versioning via URL path" | Mix of URL + header versioning | Inconsistency | Medium |
| Criterion | Checks | Common Failure |
|---|---|---|
| Context Completeness | WHY was this needed? | Jumps to solution |
| Options Considered | 3+ alternatives? | Only chosen option listed |
| Trade-off Analysis | Pros/cons per option? | Only pros of chosen |
| Decision Reasoning | WHY this won? | No explanation |
| Consequences | Impact documented? | Missing in 70% of ADRs |
| Risk Assessment | What could go wrong? | Overly optimistic |
| Reversibility | How hard to reverse? | Missing entirely |
| Scale Assumptions | At what scale? | Made for 100, applied at 1M |
| Cost Implications | Cost impact? | No cost analysis |
| Compliance Impact | Regulatory? | Not considered |
| Expiry/Review Date | When to revisit? | Decisions become stale |
| Cross-References | Related ADRs linked? | ADRs exist in isolation |
Every recommendation includes this structure — no vague advice:
What makes ArchPilot enterprise-ready: compliance, standards enforcement, admin controls, and change management.
Monitors: access controls, audit logging, encryption, change management, availability. Auto-generates evidence packages.
PHI data flows, access controls, audit trails, BAA requirements, data retention.
Traces PII through every component. Where it enters, stores, processes, shares. Right-to-deletion feasibility.
Maps cardholder data environment, identifies out-of-scope, flags segmentation gaps, validates tokenization.
Compliance audits cost $500K-2M/year. ArchPilot generating audit evidence automatically pays for itself 5x over.
Simple syntax: DENY service.exposure == "public" AND service.type == "database"
Advisory: Warning only. Soft Block: Requires justification. Hard Block: Prevents save/export.
AWS Well-Architected, OWASP Top 10, 12-Factor App, Zero Trust, HIPAA Baseline.
Enforced across ALL pillars: Meetings (flag policy violations in speech), Diagrams (block non-compliant designs), Code (reject PRs violating standards), ADRs (flag contradictions with policies).
| Role | Permissions |
|---|---|
| Org Admin | Full control: users, teams, billing, settings, policies, integrations |
| Architecture Reviewer | Approve/reject change requests, view all team architectures |
| Team Lead | Manage team settings, view team analytics, manage team projects |
| Engineer | Use all intelligence features, submit change requests, provide feedback |
| Compliance Officer | Audit logs, compliance dashboard, policy management, export reports |
| External Auditor | Read-only: compliance reports, audit logs, evidence packages |
Also includes: SAML 2.0 SSO, SCIM provisioning (Okta/Azure AD), forced SSO, MFA enforcement, IP allowlisting, session controls, data residency, DLP controls.
| Stage | Action | ArchPilot's Role |
|---|---|---|
| 1. Propose | Engineer proposes change | Auto-generates request from meeting/diagram edit |
| 2. Impact | What does this affect? | AI traces blast radius across services, teams, compliance |
| 3. Review | Board evaluates | AI report: pros/cons/risks/best-practice alignment |
| 4. Approve | Decision recorded | ADR created, knowledge graph updated, teams notified |
| 5. Implement | Engineers build it | PR reviews verify implementation matches approval |
| 6. Verify | Post-implementation | Drift detection confirms code matches approved design |
What justifies $200K/year contracts: org-wide visibility into architecture health, risk, and ROI.
| Dimension | Weight | Data Sources | Example |
|---|---|---|---|
| Reliability | 20% | Diagrams, code, anti-patterns, SPOF | 3 SPOFs, 2 missing circuit breakers → 65 |
| Security | 20% | Code scans, compliance, auth patterns | All APIs auth'd ✓, 1 hardcoded secret → 72 |
| Scalability | 15% | Topology, DB patterns, async ratio | 80% async, proper cache, shared DB → 78 |
| Decision Quality | 15% | ADR scores, drift alignment, feedback | ADRs avg 72%, 3 drift violations → 68 |
| Tech Debt | 10% | Anti-pattern age, known issues | 12 debt items, 3 over 6mo old → 55 |
| Compliance | 10% | Policy violations, audit readiness | SOC2: 94%, HIPAA: 2 gaps → 82 |
| Ops Excellence | 10% | CI/CD, monitoring, incidents | CI/CD on all, 3 missing monitoring → 70 |
Show score as line chart over 12 months. CTO to board: "Architecture health improved from 62 to 84 since ArchPilot." That's what renews $200K contracts.
| Level | Name | Indicators | ArchPilot Features Needed |
|---|---|---|---|
| 1 | Ad Hoc | No ADRs, no diagrams, tribal knowledge | Meetings + basic ADRs |
| 2 | Emerging | Some docs, inconsistent, hero-dependent | + Diagram review + search |
| 3 | Defined | Formal process, standards sometimes followed | + Code integration + policies |
| 4 | Managed | Decisions tracked, measured, improved | + Health scores + drift detection |
| 5 | Optimized | Architecture = competitive advantage | Full platform + predictive |
| Incident | Correlated Decision | ArchPilot's Analysis |
|---|---|---|
| DB connection pool exhaustion | ADR-023: "Use shared PostgreSQL" | "Flagged as risk in original discussion. Recommended separate DBs. This is the consequence." |
| Payment cascade failure ($50K) | Diagram: "Missing circuit breaker" | "Detected 4 months ago. Severity: Critical. Not actioned. Cost of inaction: $50K." |
| Auth overload on Black Friday | Meeting: "We'll scale auth later" | "Tech debt recorded. Acknowledged as future risk. Trigger: >10K concurrent." |
PagerDuty tracks incidents. Datadog monitors metrics. Nobody connects those to architecture decisions. That's ArchPilot's unique position.
What, why, who, when
Services, DBs, queues
In use, where, quality
Who knows/owns what
Debt, time bombs
Where, severity
Per-service, trends
Evolution history
Billion-dollar B2B companies build ecosystems, not just products.
| API | Use Case | Model |
|---|---|---|
| Architecture Analysis | Send diagram → get analysis + anti-patterns + scores | Per-call |
| Code Review | Send PR diff → get architectural impact | Per-call |
| Compliance Check | Send architecture → get compliance report | Per-check |
| Knowledge Query | Query org's architecture knowledge graph | Included |
| Decision Webhook | Get notified on decisions affecting your service | Included |
"Healthcare Rules" (HIPAA), "Fintech Security" (PCI-DSS), "Startup Scale-Up" (anti-premature optimization).
Slack, Teams, Jira, Linear, Confluence, Notion, PagerDuty, Datadog, GitHub, Terraform Cloud.
"E-commerce Microservices", "SaaS Multi-Tenant", "Event-Driven", "Real-Time Analytics".
"ML/AI Architecture", "IoT System", "Mobile-First", "Blockchain Infra".
500+ rules, 50+ integrations, 100+ templates → switching costs become enormous. Salesforce, Datadog, HubSpot became billion-dollar companies through marketplace lock-in.
Consulting firms brand ArchPilot as their own. Custom logo/domain. Revenue share: 30%.
AWS/GCP/Azure partners refer customers. SI partners implement. Commission: 15-20% Y1.
Silver (1+ deploy), Gold (5+ deploys), Platinum (dedicated support + co-marketing).
| Segment | Companies | ACV | TAM |
|---|---|---|---|
| Enterprise (1000+ eng) | ~5,000 | $100-500K/yr | $2.5B |
| Mid-Market (100-1000) | ~25,000 | $20-100K/yr | $1.5B |
| Startups (20-100) | ~100,000 | $5-20K/yr | $1.0B |
| Consulting (white-label) | ~2,000 | $50-200K/yr | $0.3B |
| API Platform | All | Usage-based | $0.5B |
| Total TAM | ~$5.8B | ||
| Revenue Stream | % at $100M ARR | Margin |
|---|---|---|
| Enterprise Subscriptions | 55% | 85% |
| API Platform | 15% | 90% |
| Marketplace | 10% | 95% |
| Professional Services | 10% | 40% |
| White-Label/Partner | 10% | 80% |
40% of enterprise TAM needs something other than cloud SaaS.
Supabase Cloud + Vercel. Fully managed, lowest ops. Best for startups and cloud-native enterprises.
Runs in customer's AWS/GCP/Azure. Customer controls infra, ArchPilot manages software. Unlocks $100K+ deals.
Helm chart for K8s. Customer provides PostgreSQL, storage, compute. Unlocks $200K+ deals (banks, defense).
Zero internet. Local Whisper for STT, Ollama for LLM. For defense contractors, classified environments.
Local STT + cloud AI. Transcripts never leave laptop. Only anonymized queries sent to cloud.
Bring Your Own Model: Azure OpenAI, AWS Bedrock, GCP Vertex, self-hosted. Meeting data never leaves their infra.
Every technology used across all 8 layers.
| Service | Purpose | Latency |
|---|---|---|
| Deepgram Nova-3 | Real-time streaming STT + speaker diarization | ~200ms |
| Claude Opus 4.6 | Deep arch reasoning, trade-offs, ADRs | 3-5s |
| Claude Sonnet 4.5 | Cost estimates, diagram analysis | 1-2s |
| GPT-5.2 Thinking | Code analysis, agentic workflows | 3-8s |
| GPT-5.2 Instant | Quick Q&A, summaries | 0.5-1s |
| Groq (Llama 4 Scout) | Ultra-fast triage, fallback | ~200ms |
| Service | Replaces |
|---|---|
| PostgreSQL + pgvector | RDS + DynamoDB + Pinecone |
| Auth (SAML/SCIM) | Auth0 / Clerk |
| Realtime | Redis + Socket.io |
| Storage | AWS S3 |
| Edge Functions | AWS Lambda |
| Vault | AWS KMS |
| Category | Tools |
|---|---|
| Frontend | Next.js, React Flow, D3.js, Tailwind CSS, Framer Motion |
| Desktop | Electron, desktopCapturer, AudioWorklet, electron-builder |
| Code Analysis | Tree-sitter (AST), GitHub/GitLab API |
| Enterprise | SAML 2.0, SCIM, Presidio (PII), Zod (validation), cockatiel (circuit breakers) |
| Cost Data | Infracost API |
| Deployment | Vercel, Supabase Cloud, Helm, Docker, electron-builder |
| Monitoring | Sentry, PostHog, Logflare, OpenTelemetry |
Exact build order with sprint breakdowns, dependencies, and milestones. 5 phases, 34 sprints, 18 months to enterprise-ready.
First customer
$3K MRR
10 customers
$30K MRR
30 customers
$1.8M ARR
First enterprise
$5M ARR
SOC2 certified
$10M ARR
API+Marketplace
$30M ARR
Category leader
$100M ARR
IPO-ready
$200M+ ARR
Nobody occupies this category. Otter.ai does meetings. Copilot does code. LucidChart does diagrams. Datadog does monitoring. ArchPilot is the ONLY tool that connects: what was discussed → what was decided → what was built → what broke → what should change. That cross-cutting intelligence is the billion-dollar insight.