Core Concepts¶

This section explains the fundamental concepts behind AMP. Understanding these concepts will help you build more effective content automation systems.

The AMP Model¶

AMP operates on a simple premise: content marketing should be mission-driven, not task-driven.

Traditional content workflows require you to:

1. Decide what to post
2. Write the copy
3. Create visuals
4. Format for each platform
5. Schedule posts
6. Track performance
7. Adjust strategy
8. Repeat

AMP inverts this. You define what you want to achieve (the mission), and the system autonomously handles everything else.

Core Components¶

graph TB
    subgraph "Input"
        M[Mission]
        B[Brand Context]
    end

    subgraph "Processing"
        P[Pipeline]
        P --> S1[Intake]
        S1 --> S2[Strategy]
        S2 --> S3[Content]
        S3 --> S4[Publish]
        S4 --> S5[Analytics]
        S5 --> S6[Optimize]
        S6 -.-> S2
    end

    subgraph "Output"
        C[Content]
        A[Analytics]
    end

    subgraph "Infrastructure"
        PAL[Provider Abstraction]
        Q[Job Queue]
    end

    M --> P
    B --> P
    P --> C
    P --> A
    PAL --> P
    Q --> P

Mission¶

A mission is your objective. It defines:

• What you want to achieve (objectives)
• Where content should appear (platforms)
• How content should behave (constraints)
• Success criteria (KPIs)

Missions are living entities—they execute continuously until completed or paused.

Learn more about Missions →

Pipeline¶

The pipeline is AMP's execution engine. It processes missions through six sequential stages, each with specific responsibilities. The pipeline is:

• Stateful — Can resume from any stage
• Resilient — Automatic retry with rollback
• Observable — Real-time progress tracking

Learn more about the Pipeline →

Brand Context¶

Brand context is your identity distilled into a format the AI can use. It ensures every piece of generated content matches your voice, visual style, and messaging guidelines.

Learn more about Brand Context →

Content¶

Content is the output—social posts, articles, visuals. Each piece of content:

• Belongs to a mission
• Targets a specific platform
• Has a lifecycle (draft → scheduled → published)
• Tracks its own performance

Learn more about Content Lifecycle →

Design Principles¶

1. Provider Agnostic¶

AMP abstracts all external services behind the Provider Abstraction Layer (PAL). This means:

• Swap LLMs without code changes (Claude ↔ GPT ↔ Gemini)
• Switch image generators based on needs
• Add new platforms without rewriting logic

2. Multi-Tenant by Default¶

Every piece of data belongs to a tenant. This enables:

• Complete data isolation
• Tenant-specific configurations
• Per-tenant cost tracking
• Delegated administration

Learn more about Multi-Tenancy →

3. Asynchronous First¶

Content generation is inherently slow (seconds to minutes). AMP is built async-first:

• Jobs queue via NATS JetStream
• Workers process independently
• Webhook notifications for events
• Idempotent operations

4. Continuous Optimization¶

AMP doesn't just publish and forget. The optimization stage:

• Analyzes what's working
• Identifies patterns
• Adjusts strategy
• Feeds improvements back into generation

Key Terms¶

Data Flow¶

sequenceDiagram
    participant User
    participant API
    participant Queue
    participant Worker
    participant Providers
    participant Publisher

    User->>API: Create Mission
    API->>Queue: Enqueue Job
    API-->>User: Mission Created

    Queue->>Worker: Dequeue Job
    Worker->>Worker: Intake Stage
    Worker->>Providers: Strategy (LLM)
    Providers-->>Worker: Strategy Response
    Worker->>Providers: Content (LLM + Images)
    Providers-->>Worker: Content Response
    Worker->>Publisher: Schedule Posts
    Publisher-->>Worker: Scheduled
    Worker->>Queue: Complete Job

    loop Analytics Cycle
        Publisher->>Worker: Performance Data
        Worker->>Providers: Optimize (LLM)
        Providers-->>Worker: Recommendations
    end

Explore Further¶