Setup with Google Cloud (Vertex AI)

Time needed: ~15-20 minutes Difficulty: Advanced — requires a Google Cloud project When to use: You want EU data residency (when available), GCP billing integration, or enterprise controls.

Most users should use the Gemini API Key setup instead. It takes 2 minutes and uses the same AI model. This guide is for advanced users who need GCP-specific features.

---

Prerequisites

• A Google account
• A credit card for Google Cloud billing (you won’t be charged until you exceed free tier)

---

Step 1: Create a Google Cloud Project

1. Go to Google Cloud Console
2. Sign in with your Google account
3. Click the project dropdown at the top → “New Project”
4. Name it (e.g., “golden-retriever”) → click “Create”
5. Make sure your new project is selected in the dropdown

[Screenshot placeholder: GCP Console project creation dialog]

---

Step 2: Enable billing

1. Go to Billing
2. Link a billing account to your project
3. If you don’t have a billing account, create one (requires credit card)

[Screenshot placeholder: Billing account linking page]

---

Step 3: Enable required APIs

You need two APIs enabled:

Vertex AI API

1. Go to API Library
2. Search for “Vertex AI API”
3. Click it → click “Enable”

[Screenshot placeholder: Vertex AI API page with Enable button]

Cloud Storage JSON API

1. Back in the API Library, search for “Cloud Storage JSON API”
2. Click it → click “Enable”

[Screenshot placeholder: Cloud Storage JSON API page with Enable button]

---

Step 4: Create a Cloud Storage bucket

1. Go to Cloud Storage
2. Click “Create”
3. Name your bucket (e.g., “golden-retriever-yourname”)
4. Choose a region:
   • us-central1 — Recommended (lowest latency for embeddings)
   • europe-west1 — For EU data at rest (note: embedding model currently only runs in US)
5. Leave other settings as default → click “Create”
6. Note your bucket name — you’ll need it in Step 6

[Screenshot placeholder: Bucket creation dialog with name and region fields]

---

Step 5: Create OAuth credentials

Configure the consent screen (first time only)

1. Go to OAuth Consent Screen
2. Choose “External” → click “Create”
3. Fill in:
   • App name: Golden Retriever
   • User support email: your email
   • Developer contact email: your email
4. Click “Save and Continue” through Scopes (no changes needed)
5. Under Test Users, click “Add Users” → add your email
6. Click “Save and Continue” → “Back to Dashboard”

[Screenshot placeholder: OAuth consent screen configuration]

Create the client credentials

1. Go to Credentials
2. Click ”+ Create Credentials” → “OAuth client ID”
3. Application type: “Desktop app”
4. Name: “Golden Retriever”
5. Click “Create”
6. A dialog appears with your Client ID and Client Secret — copy both

[Screenshot placeholder: OAuth client ID creation, highlighting Desktop app selection]

[Screenshot placeholder: The credentials dialog showing Client ID and Client Secret with copy buttons]

---

Step 6: Configure Golden Retriever

1. Launch Golden Retriever
2. On the setup wizard, click “Google Cloud (Vertex AI)”
3. The GCP Setup Wizard opens with 4 phases:

Phase 1: Credentials

• Paste your Client ID
• Paste your Client Secret
• Click “Sign In & Continue”
• Your browser opens for Google sign-in — authorize the app

[Screenshot placeholder: GCP wizard Phase 1 with credential fields]

Phase 2: Project Configuration

• Enter your GCP Project ID (from Step 1, e.g., “golden-retriever”)
• Enter your Bucket name (from Step 4)
• Select your Region from the dropdown
• Click “Validate Project”

[Screenshot placeholder: GCP wizard Phase 2 with project config fields]

Phase 3: Validation

• The wizard automatically checks:
  • OAuth token validity
  • Vertex AI API enabled
  • Cloud Storage API enabled
  • Bucket accessibility
  • Embedding endpoint reachability
• All checks should show green checkmarks
• Click “Continue”

[Screenshot placeholder: GCP wizard Phase 3 showing validation results]

Phase 4: Complete

• Summary of your configuration
• Click “Done”

[Screenshot placeholder: GCP wizard Phase 4 completion screen]

---

Costs

With Vertex AI, costs appear on your Google Cloud bill:

Service	Cost
Vertex AI embeddings	~$0.20 per million tokens
Cloud Storage	~$0.023/GB/month
Vertex AI Q&A (Gemini 2.5 Flash)	~$0.15 per million input tokens

For a typical personal knowledge base (a few hundred documents), expect under $1/month.

---

Privacy & Data Residency

• Your files stay on your Mac. Documents are not uploaded to GCS by default (the upload step is skipped).
• Embeddings are processed via Vertex AI. Content is sent to Google for embedding through your GCP project.
• Region pinning: You choose the processing region. However, gemini-embedding-2-preview is currently only available in us-central1. When Google adds EU regions, your Vertex AI setup will automatically use them.
• GCP data controls: You get full IAM, VPC-SC, and organization-level controls through your GCP project.

---

Troubleshooting

Problem	Solution
”Vertex AI API not enabled”	Go to API Library and enable it (Step 3)
“Bucket not found”	Check the bucket name — it must match exactly
”OAuth token invalid”	Click “Sign Out” in Settings → Cloud, then re-authenticate
”Permission denied”	Your Google account may not have the right IAM roles. You need at minimum: Vertex AI User + Storage Object Admin
3 Keychain prompts on launch	Known issue (#34) — click “Always Allow” for each prompt

---

Switching to Gemini API Key

If you want to simplify your setup:

1. Open Settings → Cloud tab
2. Click “Switch to Gemini API Key”
3. Follow the Gemini API Key setup guide

Your existing embeddings and search data will continue to work — both paths use the same model.