|
| 1 | +--- |
| 2 | +sidebar_label: Overview |
| 3 | +sidebar_position: 1 |
| 4 | +--- |
| 5 | + |
| 6 | +# Architecture Overview |
| 7 | + |
| 8 | +AOF's architecture is designed around composability, safety, and multi-model intelligence. This section covers the core architectural concepts that power the framework. |
| 9 | + |
| 10 | +## Core Principles |
| 11 | + |
| 12 | +1. **Composability**: Build complex workflows from simple, reusable components |
| 13 | +2. **Safety**: Context injection ensures environment boundaries are respected |
| 14 | +3. **Intelligence**: Multi-model consensus improves accuracy and reduces errors |
| 15 | +4. **Scalability**: Multi-tenant routing supports enterprise deployments |
| 16 | + |
| 17 | +## Architecture Documents |
| 18 | + |
| 19 | +### [Composable Design](./composable-design.md) |
| 20 | + |
| 21 | +The foundational architecture of AOF, organized into four layers: |
| 22 | + |
| 23 | +``` |
| 24 | +Layer 4: TRIGGERS ─ Platform routing (Slack, Discord, Teams, etc.) |
| 25 | +Layer 3: FLOWS ─ Multi-step workflows with orchestration |
| 26 | +Layer 2: FLEETS ─ Agent composition for collaboration |
| 27 | +Layer 1: AGENTS ─ Single-purpose AI specialists |
| 28 | +``` |
| 29 | + |
| 30 | +Each layer builds on the previous, enabling you to start simple and add complexity as needed. |
| 31 | + |
| 32 | +### [Context Injection](./context-injection.md) |
| 33 | + |
| 34 | +How AOF ensures the same agent can operate safely across different environments: |
| 35 | + |
| 36 | +- **Environment boundaries**: Isolate production from staging |
| 37 | +- **Approval workflows**: Require human approval for destructive operations |
| 38 | +- **Rate limiting**: Prevent runaway agents |
| 39 | +- **Audit trails**: Track all agent actions for compliance |
| 40 | + |
| 41 | +### [Multi-Model Consensus](./multi-model-consensus.md) |
| 42 | + |
| 43 | +Leverage multiple AI models to improve accuracy: |
| 44 | + |
| 45 | +- **Cross-validation**: Multiple models verify each other's conclusions |
| 46 | +- **Weighted voting**: Assign different weights to model opinions |
| 47 | +- **Confidence scoring**: Know when to trust the output |
| 48 | +- **Fault tolerance**: One wrong model gets outvoted |
| 49 | + |
| 50 | +### [Multi-Tenant Flows](./multi-tenant-agentflows.md) |
| 51 | + |
| 52 | +Scale AOF across organizations, teams, and projects: |
| 53 | + |
| 54 | +- **Platform routing**: Different platforms → different agents |
| 55 | +- **Channel isolation**: Team-specific agent configurations |
| 56 | +- **User/role matching**: Admin vs developer permissions |
| 57 | +- **Organization boundaries**: Enterprise multi-org support |
| 58 | + |
| 59 | +## Visual Overview |
| 60 | + |
| 61 | +``` |
| 62 | +┌─────────────────────────────────────────────────────────────────────────────┐ |
| 63 | +│ AOF Architecture │ |
| 64 | +│ │ |
| 65 | +│ ┌───────────────────────────────────────────────────────────────────────┐ │ |
| 66 | +│ │ TRIGGER LAYER │ │ |
| 67 | +│ │ Slack │ Discord │ Teams │ Telegram │ WhatsApp │ GitHub │ │ |
| 68 | +│ └─────────────────────────────────────────────────────────────────────────┘│ |
| 69 | +│ │ │ |
| 70 | +│ ▼ │ |
| 71 | +│ ┌───────────────────────────────────────────────────────────────────────┐ │ |
| 72 | +│ │ FLOW LAYER │ │ |
| 73 | +│ │ Multi-step workflows with nodes, conditions, and approval gates │ │ |
| 74 | +│ └─────────────────────────────────────────────────────────────────────────┘│ |
| 75 | +│ │ │ |
| 76 | +│ ▼ │ |
| 77 | +│ ┌───────────────────────────────────────────────────────────────────────┐ │ |
| 78 | +│ │ FLEET LAYER │ │ |
| 79 | +│ │ Agent composition: collectors → analyzers → synthesizers │ │ |
| 80 | +│ └─────────────────────────────────────────────────────────────────────────┘│ |
| 81 | +│ │ │ |
| 82 | +│ ▼ │ |
| 83 | +│ ┌───────────────────────────────────────────────────────────────────────┐ │ |
| 84 | +│ │ AGENT LAYER │ │ |
| 85 | +│ │ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │ │ |
| 86 | +│ │ │ k8s-ops │ │ aws-ops │ │ docker │ │terraform│ │ git │ ... │ │ |
| 87 | +│ │ └─────────┘ └─────────┘ └─────────┘ └─────────┘ └─────────┘ │ │ |
| 88 | +│ └─────────────────────────────────────────────────────────────────────────┘│ |
| 89 | +│ │ |
| 90 | +│ ┌───────────────────────────────────────────────────────────────────────┐ │ |
| 91 | +│ │ CROSS-CUTTING CONCERNS │ │ |
| 92 | +│ │ Context Injection │ Multi-Model Consensus │ Multi-Tenant Routing │ │ |
| 93 | +│ └─────────────────────────────────────────────────────────────────────────┘│ |
| 94 | +└─────────────────────────────────────────────────────────────────────────────┘ |
| 95 | +``` |
| 96 | + |
| 97 | +## Getting Started |
| 98 | + |
| 99 | +- **New to AOF?** Start with [Composable Design](./composable-design.md) to understand the core concepts |
| 100 | +- **Deploying to production?** Read [Context Injection](./context-injection.md) for safety best practices |
| 101 | +- **Building RCA workflows?** See [Multi-Model Consensus](./multi-model-consensus.md) |
| 102 | +- **Enterprise deployment?** Check [Multi-Tenant Flows](./multi-tenant-agentflows.md) |
0 commit comments