Model Configuration

The agent connects to an AI Gateway that routes requests to 250+ models across multiple providers. The gateway client lives in internal/llm/gateway/client.go and the model catalog is in internal/llm/gateway/models.json.

Architecture

z2e-terminal → AI Gateway → Provider API
                    │
              SSE Stream
                    │
         Text deltas + Tool calls

All communication uses SSE (Server-Sent Events) streaming for real-time output.

Configuration

Environment variables control model selection:

# Required
AI_GATEWAY_API_KEY=sk-...
VERCEL_AI_GATEWAY_API_KEY=sk-...        # Alias

# Optional
AI_GATEWAY_BASE_URL=https://ai-gateway.vercel.sh/v1
AI_GATEWAY_MODEL=openai/gpt-5.2
VERCEL_AI_GATEWAY_MODEL=openai/gpt-5.2  # Alias

Model Catalog

The model catalog (models.json) contains 250+ entries. Models are organized by provider:

Provider	Notable Models
OpenAI	`o3`, `gpt-5.2`, `gpt-4.1`, `gpt-4o-mini`, `o1`, `o1-mini`
Anthropic	`claude-sonnet-4-20250514`, `claude-3.7-sonnet`, `claude-3.5-haiku`
Google	`gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.0-flash`
DeepSeek	`deepseek-chat`, `deepseek-reasoner`
Meta	`llama-4-scout`, `llama-4-maverick`, `llama-3.3-70b`
Mistral	`mistral-large`, `codestral`, `ministral`
xAI	`grok-3`, `grok-3-fast`, `grok-3-mini`

Default: openai/gpt-5.2

Runtime Model Switching

Type /model in the TUI composer to open the interactive model picker. This:

Lists all models from the catalog
Highlights the currently active model
Updates the model on selection
Shows the new model name in the status bar

Retry Logic

Gateway calls retry on failure with exponential backoff and jitter:

Attempt	Delay
1st	Immediate
2nd	Exponential backoff + jitter
3rd	Exponential backoff + jitter
4th	Exponential backoff + jitter

Only transient errors trigger retries: HTTP 429/500/502/503/504, connection resets, refused connections, EOF, broken pipe, DNS failures, and timeouts.

Provider Routing

The AI Gateway handles provider-specific routing:

OpenAI: Chat completions API
Anthropic: Messages API
Google: Gemini API
Others: Unified chat completions interface

Model names use the format <provider>/<model-name> (e.g., openai/gpt-4o-mini).