feat(providers): add Google Vertex AI provider for Claude

[eilonc-pillar]

Summary

Adds vertex-ai as a first-class provider, enabling Claude models via Google Cloud Vertex AI with Application Default Credentials (ADC).

• Uses AnthropicVertex from anthropic[vertex] SDK
• Reuses the existing Anthropic Messages API adapter — zero new adapters needed
• Supports main agent, auxiliary tasks, compression, delegation, and credential pool routing

Configuration

provider: vertex-ai
model: claude-sonnet-4-6

# Environment
export VERTEX_PROJECT=my-gcp-project   # required
export VERTEX_REGION=us-east5           # optional, default: us-east5

# Auth (any of these):
# 1. Attached service account on GCP VMs (ADC) — zero config
# 2. GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json
# 3. gcloud auth application-default login (local dev)

Changes

File	Lines	What
agent/anthropic_adapter.py	+52	build_vertex_client() + resolve_vertex_credentials()
agent/auxiliary_client.py	+34	_try_vertex(), auto-resolver guard, resolve_provider_client routing, aux model entry
hermes_cli/auth.py	+7	Provider registry entry
hermes_cli/runtime_provider.py	+19	Vertex-ai resolution + credential pool api_mode fix
run_agent.py	+18	Provider detection + client init
pyproject.toml	+2	vertex optional extra

What changed since v1

• Added hermes_cli/runtime_provider.py — vertex-ai resolution was missing, causing the credential pool to fall back to chat_completions and create an OpenAI client instead of AnthropicVertex
• Added credential pool api_mode fix in _resolve_runtime_from_pool_entry()
• Rebased onto current main (resolved auxiliary_client.py conflict from upstream's _resolve_forced_provider removal)

Tested

Locally and on GCP with a Vertex AI service account using ADC.

Install

pip install 'hermes-agent[vertex]'

[eilonc-pillar]

Hi all - @leon @intertwine @timknip @rewbs

Very good job on the Hermes agent! It delivers a lot of value and innovation.
I've opened this PR to enable powering hermes agent with Google's vertex AI juice (both ACP and SA json). Would appreciate if you consider to add this.

Best,
Eilon, Pillar Security

[DeadMeme5441]

Would like this feature to be merged, would work great.

[teknium1]

Thanks for the Claude-on-Vertex contribution. The requested behavior is not present on current main, but this April implementation predates the current Vertex provider and needs a targeted salvage rather than a direct application.

Problems

• Current main already owns vertex and maps vertex-ai to it; its runtime path is Gemini/OpenAI-compatible chat_completions at hermes_cli/runtime_provider.py:1573-1604. The PR's separate vertex-ai Anthropic path must coexist with that established Gemini path.
• The PR's VERTEX_PROJECT / VERTEX_REGION configuration bypasses current vertex: config and profile-aware secret handling (agent/vertex_adapter.py:50-103, hermes_cli/config.py:3287-3303).
• The diff adds no tests while changing runtime selection, initialization, auxiliary routing, and credential handling. Existing Vertex coverage is in tests/agent/test_vertex_adapter.py and tests/hermes_cli/test_vertex_provider.py.

Suggested changes

• Implement Claude model detection as a second transport path of the existing vertex provider, preserving Gemini support.
• Reuse the current config and profile-safe credential-resolution mechanisms.
• Add coverage for Claude routing/client construction, auxiliary use, and the existing Gemini regression path.

Automated hermes-sweeper review.