Your Name
2.4 KiB
2.4 KiB
Configuring the proxy for Vertex AI (GCP)
The proxy supports GCP models via the /proxy/gcp/claude endpoint. There are a few extra steps necessary to use GCP compared to the other supported APIs.
Setting keys
Use the GCP_CREDENTIALS environment variable to set the GCP API keys.
Like other APIs, you can provide multiple keys separated by commas. Each GCP key, however, is a set of credentials including the project id, client email, region and private key. These are separated by a colon (:).
For example:
GCP_CREDENTIALS=my-first-project:xxx@yyy.com:us-east5:-----BEGIN PRIVATE KEY-----xxx-----END PRIVATE KEY-----,my-first-project2:xxx2@yyy.com:us-east5:-----BEGIN PRIVATE KEY-----xxx-----END PRIVATE KEY-----
Setup Vertex AI
- Go to https://cloud.google.com/vertex-ai and sign up for a GCP account. ($150 free credits without credit card or $300 free credits with credit card, credits expire in 90 days)
- Go to https://console.cloud.google.com/marketplace/product/google/aiplatform.googleapis.com to enable Vertex AI API.
- Go to https://console.cloud.google.com/vertex-ai and navigate to Model Garden to apply for access to the Claude models.
- Create a Service Account , and make sure to grant the role of "Vertex AI User" or "Vertex AI Administrator".
- On the service account page you just created, create a new key and select "JSON". The JSON file will be downloaded automatically.
- The required credential is in the JSON file you just downloaded.
Supported model IDs
Users can send these model IDs to the proxy to invoke the corresponding models.
- Claude
claude-haiku-4-5@20251001claude-sonnet-4-5@20250929claude-opus-4-1@20250805claude-3-5-haiku@20241022claude-sonnet-4@20250514claude-opus-4@20250514
For OpenAI-compatible callers, the proxy will also remap Claude-style aliases like claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, and claude-opus-4-1-20250805 to the corresponding Vertex AI model IDs.