Bedrock Mantle Responses API is incompatible with official OpenAI .NET SDK

I'm using Bedrock Mantle's OpenAI-compatible Responses API endpoint with the official OpenAI .NET SDK (openai-dotnet). I've found two compatibility issues that prevent the SDK from working with Bedrock Mantle out of the box. The same code works correctly against Azure Foundry without any modifications. Also the chat completions API works without any issues

Endpoint: https://bedrock-mantle.eu-north-1.api.aws/v1
Model: openai.gpt-oss-120b
SDK: openai-dotnet (Responses API, GetResponsesClient())
Runtime: .NET 10

Issue 1: created_at returned in float / scientific notation

When a request succeeds, the response contains created_at as a float in scientific notation:

{
  "created_at": 1.772195752E9
}

The OpenAI Responses API returns this as an integer (e.g., 1772195752). The OpenAI .NET SDK uses JsonElement.GetInt64() to deserialize this field (source), and System.Text.Json throws a FormatException on any number containing a decimal point or scientific notation, even if the underlying value is a whole number.

Stack trace:

System.FormatException: One of the identified items was in an invalid format.
   at System.Text.Json.ThrowHelper.ThrowFormatException()
   at System.Text.Json.JsonElement.GetInt64()
   at OpenAI.Responses.ResponseResult.DeserializeResponseResult(...)

This means even a successful Bedrock Mantle response cannot be read by the official SDK.

Issue 2: content as array of content parts is rejected

The OpenAI .NET SDK serializes input messages with content as an array of content parts. This is the format produced by ResponseItem.CreateUserMessageItem() and ResponseItem.CreateSystemMessageItem():

{
  "type": "message",
  "role": "user",
  "content": [
    {
      "type": "input_text",
      "text": "Hello, can you help me with a task?"
    }
  ]
}

Bedrock Mantle rejects this with:

{
  "status": "failed",
  "error": {
    "code": "invalid_prompt",
    "message": "ErrorEvent { error: APIError { type: \"BadRequestError\", code: Some(400), message: \"invalid conversation JSON: invalid type: sequence, expected a string at line 1 column 4880\", param: None } }"
  }
}

Minimal reproduction

This request will fail:

{
  "model": "openai.gpt-oss-120b",
  "max_output_tokens": 1024,
  "tools": [],
  "input": [
    {
      "type": "message",
      "role": "system",
      "content": [
        {
          "type": "input_text",
          "text": "You are a helpful assistant."
        }
      ]
    },
    {
      "type": "message",
      "role": "user",
      "content": [
        {
          "type": "input_text",
          "text": "What is a singleton pattern?"
        }
      ]
    }
  ],
  "parallel_tool_calls": true
}

Environment

• AWS Region: eu-north-1
• Endpoint: https://bedrock-mantle.eu-north-1.api.aws/v1
• Model: openai.gpt-oss-120b
• SDK: openai-dotnet (2.8.0), using the Responses API client
• Runtime: .NET 10, C# 14
• Comparison: The identical code and request structure works against Azure Foundry with no modifications or workarounds