feat: add Google Vertex AI as a first-class Claude provider

[elbertwang]

Add support for running Claude models on Google Vertex AI, using the Anthropic Python SDK's AnthropicVertex client which shares the same .messages.create() / .messages.stream() interface as the regular Anthropic client — making it a near drop-in replacement.

Configuration is fully compatible with Claude Code's Vertex AI setup:

export CLAUDE_CODE_USE_VERTEX=1
export ANTHROPIC_VERTEX_PROJECT_ID=<project-id>
export CLOUD_ML_REGION=us-east5   # or "global"

Key changes:

• agent/anthropic_adapter.py: new build_vertex_client() function
• hermes_cli/auth.py: register vertex provider with gcloud_adc auth type, aliases, and auto-detection via CLAUDE_CODE_USE_VERTEX
• hermes_cli/providers.py: HermesOverlay, aliases, display label
• hermes_cli/runtime_provider.py: vertex resolution (skips credential pooling since Vertex uses Google ADC)
• hermes_cli/main.py: add Vertex AI to the top-level provider menu in hermes model / hermes setup with interactive project/region configuration
• run_agent.py: vertex-aware client construction in AIAgent init, switch_model, error recovery, and fallback paths via new _build_anthropic_client_for_provider() helper
• cli.py: skip api_key/base_url validation for vertex provider; pass project_id/region through to AIAgent
• gateway/run.py: forward project_id/region in runtime kwargs
• agent/auxiliary_client.py: _try_vertex() for sub-tasks
• agent/model_metadata.py: add vertex to provider prefixes
• agent/models_dev.py: map vertex → anthropic in models.dev registry
• pyproject.toml: anthropic → anthropic[vertex] (adds google-auth)

Tests: 16 new tests (6 client + 10 provider), 192 total passing.

What does this PR do?

Related Issue

Fixes #

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

How to Test

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:

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

For New Skills

• This skill is broadly useful to most users (if bundled) — see Contributing Guide
• SKILL.md follows the standard format (frontmatter, trigger conditions, steps, pitfalls)
• No external dependencies that aren't already available (prefer stdlib, curl, existing Hermes tools)
• I've tested the skill end-to-end: hermes --toolsets skills -q "Use the X skill to do Y"

Screenshots / Logs

[alt-glitch]

Likely duplicate of #3569 — both add Google Vertex AI as a Claude provider. Also related to #16010 (Vertex AI for Gemini) and #13484 (feature request).

[alt-glitch]

Likely duplicate of #3569

[teknium1]

Thanks for the Claude-on-Vertex implementation. The requested Claude behavior remains absent on current main, but this April branch predates the established Vertex/Gemini provider and needs a targeted salvage.

Problems

• Current main already maps vertex and vertex-ai to Gemini's OpenAI-compatible chat_completions route at hermes_cli/runtime_provider.py:1573-1604. This PR instead unconditionally routes vertex to anthropic_messages at hermes_cli/runtime_provider.py:754-772, so it would replace Gemini rather than add a Claude path.
• The Vertex fallback block at run_agent.py:4957 cannot run: the preceding mode selection only marks Anthropic or /anthropic endpoints as anthropic_messages (run_agent.py:4938-4955).
• hermes_cli/main.py:2655-2660 writes project, region, and an enable flag to .env; current main places non-secret Vertex routing settings in vertex: config (hermes_cli/config.py:3287-3303).

Suggested changes

• Add model-based dual routing within the existing vertex provider, preserving the Gemini route and choosing AnthropicVertex for Claude IDs.
• Reuse the current config/profile-safe credential path and add Gemini-regression plus Vertex-fallback coverage.

This is an automated hermes-sweeper review.

[teknium1]

Current main already uses vertex/vertex-ai for Gemini through the OAuth2 OpenAI-compatible path. This unconditional route must become model-based so Claude selects AnthropicVertex without replacing the existing Gemini behavior.

[teknium1]

This branch is unreachable for a Vertex fallback: above, fb_api_mode becomes anthropic_messages only for anthropic or a base URL ending in /anthropic; Vertex falls through as chat_completions and never enters this block.

[teknium1]

Project ID and region are non-secret routing settings. Persist them under the existing vertex: config section rather than .env, matching the repository configuration policy and current Vertex provider.