AI provider
Choose who serves the models Hamming runs on. Phloem picks the model; new models arrive with a release, never by surprise.
Set it in your Phloem account under AI. Only a workspace admin can change it; members can see it.
Providers
| Provider | Bills | Plans |
|---|---|---|
| Phloem managed | Phloem, counted as AI credits, on a key made for this workspace only | every plan; the only choice on Starter |
| Google Vertex AI | your Google Cloud project | Enterprise |
| Azure AI Foundry | your Azure subscription | Enterprise |
| Amazon Bedrock | your AWS account | Enterprise |
| Anthropic, Google Gemini, OpenAI | that vendor, on your API key | Enterprise |
On a cloud provider, Model family picks whose models run there: Gemini, Claude or OpenAI, as that cloud offers them. A provider you bring does not use up monthly credits.
Access
The page groups providers as Your cloud (named for the cloud your data plane runs in), then Another cloud · federated from your plane (collapsed until opened), then Your API key. Each cloud provider offers these ways to let your plane in:
| Access | Means |
|---|---|
| Your data plane's own identity | no key — allow your plane's service account to call Vertex AI; nothing to store or rotate. Recommended; offered when the data plane is your own |
| Federated from your plane — no keys | AWS or Azure trusts your plane's Google identity; the page shows the exact trust policy or federated credential to create |
| Service account key · Access keys · API key | stored encrypted in your workspace, for accounts that don't allow federation |
A saved key is never shown again, only its last characters or a service account key's email. Where requests go confirms that requests go straight from your data plane; nothing passes through Phloem.
Keyless Vertex AI needs a data plane of your own. On one Phloem runs for you, use a service account key: a project that trusted the shared plane would trust every workspace on it.
Updated