fix(bedrock): preserve dots in Bedrock model IDs during normalization

[chhanz]

What does this PR do?

normalize_model_name() unconditionally replaces all dots with hyphens, which breaks Bedrock model IDs where dots are structural separators (e.g.
global.anthropic.claude-opus-4-6-v1 becomes the invalid global-anthropic-claude-opus-4-6-v1). This causes all Claude models on Bedrock to fail with ValidationException: The provided model identifier is invalid, while Nova models (which use the bedrock_converse path) work fine.

This PR adds a _is_bedrock_model_id() helper to detect Bedrock-native IDs by their regional/vendor prefixes and skips dot-to-hyphen conversion for them.

Root Cause

normalize_model_name() was designed for OpenRouter-style names (claude-opus-4.6 → claude-opus-4-6) but did not account for Bedrock-native IDs where dots delimit
region/vendor/model hierarchy.

Related Issue

• [Feature]: Auto-detect Bedrock inference profile prefix for Claude models #10774 — Auto-detect Bedrock inference profile prefix for Claude models
• normalize_model_name() breaks model names for custom base URLs #7164 — normalize_model_name() breaks model names for custom base URLs

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• agent/anthropic_adapter.py: Added _is_bedrock_model_id() helper that detects Bedrock model ID patterns by regional prefixes (global., us., eu., ap., apac., jp.) and
  foundation model prefix (anthropic.claude)
• agent/anthropic_adapter.py: Modified normalize_model_name() to return early (preserving dots) when _is_bedrock_model_id() returns True
• tests/agent/test_anthropic_adapter.py: Added TestIsBedrockModelId class and test_preserves_dots_in_bedrock_model_ids test method (6 new test cases)

How to Test

1. Configure Hermes with Bedrock provider (e.g. region ap-northeast-2)
2. Set model to any Claude inference profile (e.g. global.anthropic.claude-opus-4-6-v1)
3. Send any message

Before fix: ValidationException: The provided model identifier is invalid.
After fix: Normal response from Claude via AnthropicBedrock SDK

Affected Model ID Patterns

Pattern	Example	Dots are...
Global inference profile	global.anthropic.claude-opus-4-6-v1	structural (preserved ✓)
Regional inference profile	apac.anthropic.claude-sonnet-4-20250514-v1:0	structural (preserved ✓)
Foundation model	anthropic.claude-opus-4-6-v1	structural (preserved ✓)
OpenRouter format	claude-opus-4.6	version separator (converted → claude-opus-4-6 ✓)
Aggregator slug	anthropic/claude-sonnet-4.6	version separator (converted → claude-sonnet-4-6 ✓)

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: Amazon Linux 2023 (EC2, ap-northeast-2, Python 3.11.14)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility
  guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Screenshots / Logs

Error log (before fix)

ERROR root: API call failed after 3 retries. An error occurred (ValidationException)
when calling the ConverseStream operation: The provided model identifier is invalid.
| provider=bedrock model=global.anthropic.claude-opus-4-7 msgs=2 tokens=~3,106

Test results (after fix)

$ venv/bin/python -m pytest tests/agent/test_anthropic_adapter.py -q
126 passed in 5.12s

Verification

>>> from agent.anthropic_adapter import normalize_model_name
>>> normalize_model_name("global.anthropic.claude-opus-4-6-v1")
'global.anthropic.claude-opus-4-6-v1'  # dots preserved ✓
>>> normalize_model_name("claude-opus-4.6")
'claude-opus-4-6'  # still converts OpenRouter format ✓
>>> normalize_model_name("anthropic/claude-sonnet-4.6")
'claude-sonnet-4-6'  # still strips prefix and converts ✓

[alt-glitch]

Multiple competing PRs for Bedrock dot preservation: #13821, #13112, #11700. This one uses prefix detection; #13821 uses URL heuristic.

[teknium1]

Thanks for the contribution @chhanz! This fix has been implemented on main independently.

This is an automated hermes-sweeper review.

The same bug was fixed by a parallel PR (#12295) which merged on 2026-04-23 as commit f2fba4f9a:

• agent/anthropic_adapter.py now contains _is_bedrock_model_id() at line 989 covering the same prefixes (anthropic., us., eu., ap., jp., global.)
• normalize_model_name() already guards against dot-to-hyphen conversion for Bedrock IDs at line 1027

Member @alt-glitch also flagged the competing PRs (#13821, #13112, #11700) in their review comment. Your analysis and test cases were solid — they just arrived slightly after another fix landed. Closing as implemented on main.