Models

The Models page enables model alias mapping and provider decoupling in Aegis.

Instead of hardcoding vendor-specific model identifiers (such as publishers/google/models/gemini-1.5-pro-002 or gpt-4o-2024-08-06) directly into your application codebase, developers invoke clean, platform-managed Model Aliases (such as gemini-flash or gemini-1.5-pro).

---

Why Use Model Aliases?

• Zero-Downtime Hot Swapping: Switch an alias from one provider or model version to another in the Aegis Console instantly without deploying new code.
• Fallbacks & Redundancy: Map aliases to primary and fallback providers so traffic automatically reroutes during upstream provider outages.
• Cost Optimization: Direct non-critical tasks to lightweight aliases (gemini-flash) while routing complex reasoning requests to high-capacity aliases (gemini-pro).

---

Model Alias Configuration

Navigate to Gateway > Models in the Aegis Console to inspect active model aliases, provider bindings, and model IDs.

---

Adding a Model to a Provider

1. Navigate to Gateway > Models in the Aegis Console.
2. Click + Add Model in the top right.
3. In the modal dialog, configure the model properties:

• Provider: Select the linked cloud provider connection (e.g., gcp).
• Display Name: Enter a human-readable display label (e.g., Gemini-Flash).
• Model ID (Provider-native ID): Enter the exact model identifier required by the backend API (e.g., gemini-3.6-flash).
• Alias ID (Optional): Enter the short alias string used by client applications to request this model (e.g., gemini-flash).

4. Click Add Model.

---

Model Mapping Table

The Models table lists all registered aliases across your active workspace:

Field	Description	Example
Display Name	UI display label	Gemini-Flash
Alias ID	Short alias string used by clients	gemini-flash
Model ID	Provider-native model identifier	gemini-3.6-flash
Provider	Linked cloud provider credential	gcp (Google Cloud)
Actions	Edit mapping or delete model alias	—

---

How Applications Resolve Model Aliases

When a client app or agent runtime sends an LLM completion request to Aegis:

POST /v1/chat/completions
{
  "model": "gemini-flash",
  "messages": [{"role": "user", "content": "Analyze transaction risk"}]
}

1. Alias Lookup: Aegis intercepts "model": "gemini-flash".
2. Provider Resolution: Resolves gemini-flash to the linked gcp provider credential and upstream model gemini-3.6-flash.
3. Governed Dispatch: Validates rate limits, injects Service Account OAuth tokens, logs trace spans, and forwards the request to Vertex AI.