feat: add Vertex AI provider for enterprise grade GCP integration

[realtusharnargunde]

Enhance OpenWiki to allow Enterprise level readiness via Google Vertex AI integration. As a supported provider, Vertex AI will help enable Claude models via GCP's Model Garden. This allows organizations on GCP to use OpenWiki with their existing cloud infrastructure.

Requirement : How Vertex AI Helps

1. Unified billing : AI costs on existing GCP invoices
2. IAM integration : Fine-grained access control, audit logs
3. No new vendor : Use already-approved cloud provider

Usage details. :

For local development :

One-time setup

gcloud auth application-default login

Run OpenWiki

OPENWIKI_PROVIDER=vertexai
ANTHROPIC_VERTEX_PROJECT_ID=my-project
CLOUD_ML_REGION=us-east5
openwiki --init

Interactive Setup

openwiki # Select "Vertex AI" provider, press Enter to use ADC

For CI/CD Setup - work with org policies to facilitate appropriate usage

Vertex AI authentication in CI/CD depends on your organization's policies:

• Workload Identity Federation (recommended) — keyless auth from GitHub/GitLab
• Service Account Key ( best for PoC ) — if allowed by org policy

Testing :

[SadriG91]

Heads-up: #179 (opened Jul 7) implements the same thing — Claude models via Vertex AI with ADC auth — so maintainers may want to look at the two together. The main design differences, for easier triage:

• Env vars: feat: add google vertex ai (claude) provider #179 uses the standard GCP names (GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION) passed explicitly to the AnthropicVertex client, rather than the Anthropic-SDK-specific ANTHROPIC_VERTEX_PROJECT_ID / CLOUD_ML_REGION — the standard names line up with gcloud, workload-identity CI setups, and other GCP tooling out of the box.
• Default endpoint: feat: add google vertex ai (claude) provider #179 defaults the location to global (overridable) for availability, rather than a regional example.
• feat: add google vertex ai (claude) provider #179 also carries the README + openwiki/ doc updates, onboarding steps for project/location, tests across constants/env/commands, and a documented live E2E against a GCP project, and it's currently rebased conflict-free on 0.0.4.

Happy to consolidate in either direction — if there are ideas here that #179 is missing, I'll gladly fold them in.

[Brad Huffman (ppsplus-bradh)]

Coordination pointer: this overlaps a few other in-flight Vertex PRs — #154 adds Vertex as a gemini-enterprise provider that routes Gemini/Gemma + Claude + Llama/Mistral/DeepSeek/Qwen (MaaS) through one provider via resolveVertexSurface(), #179 adds a dedicated Vertex Claude provider, and #238 adds Vertex + Bedrock. All touch src/constants.ts, src/agent/index.ts, and src/credentials.tsx, so first-to-land will conflict with the others.

Summary of how they relate: #154 (comment) — raising it so the maintainers can settle on one Vertex direction. Happy to help consolidate.

[realtusharnargunde]

Closing this PR based on direction and progress on others. Cheers