docs(bedrock): clarify Mantle model-family API roots

[fscfede-beep]

Summary

Clarifies the route selection described in #812 without changing the existing Mantle default.

AWS currently documents different bedrock-mantle API roots for different model families:

• OpenAI model pages such as GPT-5.5 use /openai/v1.
• Other model pages such as Kimi K2.5 use /v1.

This documents that EndpointMantle intentionally preserves /openai/v1 for backwards compatibility, does not infer the route from the model name, and shows the existing Config.BaseURL escape hatch for deployments whose AWS model card requires /v1.

No SDK behavior, authentication, signing, generated code, or exported API is changed.

Validation

• git diff --check
• Fresh collision audit immediately before publication: no PR cross-reference on EndpointMantle defaults to /openai/v1, but AWS documents mantle's base URL as /v1 #812 and no matching open upstream PR.
• Cross-checked the documented roots against the current AWS GPT-5.5 and Kimi K2.5 model cards.

Local Go tests were not run because Go is not installed on this host; this is a documentation-only change and normal repository CI remains authoritative.

Closes #812.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1566b0e22b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[fscfede-beep]

Addressed the P2 in 41ff61a1f4ec9eeb1c67fb87e1dc14af85d4593a: the root Bedrock section now states that /openai/v1 is the default for OpenAI model families, notes that other Mantle model families can require /v1, and links directly to the model-specific Mantle guidance before a BaseURL override. git diff --check passes; behavior remains unchanged.

@codex review

[chatgpt-codex-connector]

Note

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[fscfede-beep]

@codex review

Please review the current head 41ff61a1f4ec9eeb1c67fb87e1dc14af85d4593a. This head addresses the prior P2 by reconciling the root Bedrock guidance with the model-family-specific Mantle routes. The previous review request on this corrected head was blocked by Codex usage limits, so there is no current-head review result to rely on. Scope remains documentation-only; no SDK behavior, auth, signing, generated code, or exported API change is claimed.

[chatgpt-codex-connector]

Note

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[fscfede-beep]

RUMBO audit note: current PR head 41ff61a1f4ec9eeb1c67fb87e1dc14af85d4593a is still open but currently reports mergeable=false. I am not pushing or requesting another automated review in this pass because the previous current-head review attempts were blocked by Codex review limits, and repeated pings would add noise without new evidence. Next useful contributor action is to reconcile the branch against current upstream main and revalidate the documentation-only diff before any further review request.