[Feature Request]: Implement Bedrock V2 Client (ModelClientV2)

[priyansh4320]

Is your feature request related to a problem? Please describe.

Implement Bedrock V2 Client (ModelClientV2)

Why Implement Bedrock V2 Client?

The current Bedrock client (api_type: "bedrock") uses the legacy ModelClient interface, which returns flattened ChatCompletion responses. This limits access to rich content types and creates inconsistencies across providers.

Current Limitations (V1 Client)

1. Lost Rich Content: Tool calls, images, and other structured content are flattened or lost during response processing
2. Provider-Specific Parsing: Requires manual parsing of nested structures to access different content types
3. No Type Safety: Responses are untyped dictionaries, making it easy to introduce bugs
4. Limited Forward Compatibility: New Bedrock features require code changes to support
5. Inconsistent API: Different response formats across OpenAI, Anthropic, Gemini, and Bedrock clients

Benefits of V2 Client

1. Rich Content Preservation

   • All content types (text, images, tool calls, citations) preserved as typed ContentBlock objects
   • No data loss during response transformation
   • Direct access to structured content

2. Better Developer Experience

   # V1 - Manual parsing
   response = client.create(params)
   messages = client.message_retrieval(response)
   content = messages[0] if messages else ""
 
   # V2 - Direct property access
   response = client.create(params)
   text = response.text                    # All text content
   tool_calls = response.get_tool_calls()   # Tool calls as objects
   images = response.get_content_by_type("image")  # Image content blocks
   

Type Safety

• Pydantic models with automatic validation
• Typed content blocks (TextContent, ImageContent, ToolCallContent, etc.)
• Enum-based content types
• IDE autocomplete and type checking support

3. Forward Compatibility

   • Unknown content types automatically handled via GenericContent
   • No code changes needed when Bedrock adds new features
   • Future-proof architecture

4. Provider-Agnostic Format

   • Unified response format across OpenAI, Anthropic, Gemini, and Bedrock
   • Easier to switch between providers
   • Consistent developer experience across all LLM providers

5. Backward Compatible

   • Can use create_v1_compatible() to get V1 format when needed
   • Works seamlessly with existing V1 clients in group chats
   • No breaking changes to existing code

6. Alignment with AG2 Architecture

   • Follows the ModelClientV2 protocol already implemented for OpenAI
   • Supports the unified response model across all providers

Use Cases Enabled

• Structured Outputs: Better support for Pydantic models and JSON schemas
• Multimodal Applications: Proper handling of images and other media types
• Tool Calling: Rich access to tool call information
• Cost Tracking: Built-in cost calculation per response
• Future Features: Ready for new Bedrock capabilities without code changes

Migration Path

Migration is simple - just change api_type:

# V1
{"api_type": "bedrock", "model": "...", ...}

V2 (recommended)

{"api_type": "bedrock_v2", "model": "...", ...} # All other configuration parameters remain identical.

Implementation Status

✅ Completed: Bedrock V2 client implementation with full ModelClientV2 protocol support
✅ Completed: Comprehensive unit tests
✅ Completed: Integration tests
✅ Completed: Documentation updates
✅ Completed: Backward compatibility layer

Related

• OpenAI V2 Client: Reference implementation

Describe the solution you'd like

No response

Additional context

No response