# @memberjunction/ai-bedrock

MemberJunction AI provider for Amazon Bedrock. This package provides both LLM and embedding capabilities through AWS Bedrock's multi-model platform, supporting Claude, Titan, Llama, and other model families available on Bedrock.

## Architecture

```mermaid
graph TD
    A["BedrockLLM<br/>(Provider)"] -->|extends| B["BaseLLM<br/>(@memberjunction/ai)"]
    C["BedrockEmbedding<br/>(Provider)"] -->|extends| D["BaseEmbeddings<br/>(@memberjunction/ai)"]
    A -->|wraps| E["AWS Bedrock<br/>Runtime SDK"]
    C -->|wraps| E
    E -->|accesses| F["Claude / Titan /<br/>Llama / Mistral"]
    B -->|registered via| G["@RegisterClass"]
    D -->|registered via| G

    style A fill:#7c5295,stroke:#563a6b,color:#fff
    style C fill:#7c5295,stroke:#563a6b,color:#fff
    style B fill:#2d6a9f,stroke:#1a4971,color:#fff
    style D fill:#2d6a9f,stroke:#1a4971,color:#fff
    style E fill:#2d8659,stroke:#1a5c3a,color:#fff
    style F fill:#b8762f,stroke:#8a5722,color:#fff
    style G fill:#b8762f,stroke:#8a5722,color:#fff
```

## Features

- **Multi-Model Access**: Use Claude, Titan, Llama, Mistral, and other models via a single provider
- **Chat Completions**: Conversational AI through Bedrock's unified API
- **Embeddings**: Text embeddings via Amazon Titan and other embedding models
- **Streaming**: Real-time response streaming support
- **AWS Integration**: Native AWS authentication (IAM roles, access keys, SSO)
- **Private Deployments**: Support for custom and fine-tuned models on Bedrock

## Installation

```bash
npm install @memberjunction/ai-bedrock
```

## Usage

### Chat Completion

```typescript
import { BedrockLLM } from '@memberjunction/ai-bedrock';

const llm = new BedrockLLM('aws-credentials-json');

const result = await llm.ChatCompletion({
    model: 'anthropic.claude-v2',
    messages: [
        { role: 'user', content: 'Explain machine learning.' }
    ]
});
```

### Embeddings

```typescript
import { BedrockEmbedding } from '@memberjunction/ai-bedrock';

const embeddings = new BedrockEmbedding('aws-credentials-json');

const result = await embeddings.EmbedText({
    text: 'Text to embed',
    model: 'amazon.titan-embed-text-v1'
});
```

## Configuration

AWS credentials are passed as a JSON string containing:

| Field | Description |
|-------|-------------|
| `region` | AWS region (e.g., `us-east-1`) |
| `accessKeyId` | AWS access key ID |
| `secretAccessKey` | AWS secret access key |
| `sessionToken` | Optional session token for temporary credentials |

## Supported Features

| Feature | Supported | Notes |
|---------|-----------|-------|
| stopSequences | Yes | Supported for Anthropic Claude, AI21, and Amazon Titan models |
| assistantPrefill | Yes (Claude only) | Pre-seed the assistant's response for `anthropic.*` model IDs ([guide](../../Prompts/PREFILL_AND_STOP_SEQUENCES.md)) |
| streaming | Yes | Real-time streaming via `InvokeModelWithResponseStream` |

## Class Registration

- `BedrockLLM` -- Registered via `@RegisterClass(BaseLLM, 'BedrockLLM')`
- `BedrockEmbedding` -- Registered via `@RegisterClass(BaseEmbeddings, 'BedrockEmbedding')`

## Dependencies

- `@memberjunction/ai` - Core AI abstractions
- `@memberjunction/global` - Class registration
- `@aws-sdk/client-bedrock-runtime` - AWS Bedrock Runtime SDK