Amazon Bedrock Mantle

Overview

Amazon Bedrock Mantle is a separate DeepIntShield provider identity from the Bedrock Runtime provider. Use bedrock-mantle/<model> for the regional OpenAI-compatible endpoint; use bedrock/<model> for Bedrock Runtime’s Converse and Invoke APIs.

Supported operations

Operation	Non-streaming	Streaming	DeepIntShield route
List models	✅	—	GET /v1/models
Chat Completions	✅	✅ SSE	POST /v1/chat/completions
Responses	✅	✅ SSE	POST /v1/responses
Stored Responses retrieve	✅	✅ SSE	GET /v1/responses/{response_id}
Stored Responses delete/cancel	✅	—	/v1/responses/{response_id}
Responses compaction	✅	—	POST /v1/responses/compact
Embeddings	✅	—	POST /v1/embeddings
Count tokens	✅	—	POST /v1/responses/input_tokens
Files: upload/list/get/delete/content	✅	—	/v1/files

Input-item listing is not enabled for this provider. Realtime sessions and Responses WebSocket mode are also unsupported.

Configure the regional endpoint

1. Create a short-lived Amazon Bedrock API key with access to the models and operations you intend to use.
2. Open Workspace → AI Providers, select Amazon Bedrock Mantle, and add the API key.
3. Set Network config → Base URL to the endpoint for your AWS Region.

The default is https://bedrock-mantle.us-east-1.api.aws. For another region, set the regional hostname without a trailing /v1; DeepIntShield normalizes a configured /v1 suffix if one is present.

Caution

Long-lived Bedrock API keys are intended for exploration. For production, follow AWS guidance for short-term credentials, expiration, least privilege, rotation, and CloudTrail review.

Make a Responses request

curl https://app.deepintshield.com/v1/responses \
  -H "Authorization: Bearer $DEEPINTSHIELD_VIRTUAL_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bedrock-mantle/openai.gpt-oss-120b-1:0",
    "input": "Summarize the deployment risks.",
    "store": true
  }'

The gateway sends the configured Bedrock API key as the upstream bearer token; the caller sees only the DeepIntShield Virtual Key. For deterministic follow-up operations when several keys can serve the model, pin the same configured key as described in Stored Responses and compaction.

Embeddings, files, and token counting

Use a provider-qualified model on model-bearing requests:

curl https://app.deepintshield.com/v1/embeddings \
  -H "Authorization: Bearer $DEEPINTSHIELD_VIRTUAL_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bedrock-mantle/amazon.titan-embed-text-v2:0",
    "input": "Reference text"
  }'

Token counting uses the provider’s Anthropic-compatible count-tokens operation and is therefore model-dependent. File operations use the Mantle Files API; retain the same provider and key affinity for later resource operations.

Model discovery and pricing

GET /v1/models?provider=bedrock-mantle discovers the account’s regional catalog. Availability varies by model and region. AWS states that per-token pricing for a model is the same on Bedrock Runtime and Mantle; verify current Amazon Bedrock pricing and configure contracted rates through pricing overrides.

See AWS’s Mantle Responses guide, endpoint comparison, and API key guidance.