Examples

A few worked end-to-end recipes. All examples use environment variables — never paste a real key.

Default auto-router

export SMART_GATEWAY_BASE_URL="https://smart-openrounter.bee1x.one"
export SMART_GATEWAY_API_KEY="$YOUR_GATEWAY_KEY"

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router",
        "messages": [
          {"role": "system", "content": "You are a helpful assistant."},
          {"role": "user",   "content": "Explain HMAC in one paragraph."}
        ],
        "max_tokens": 200
      }'

Force Pro for a hard task

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router-pro",
        "messages": [
          {"role": "user", "content": "[reasoning:high] Design a token-bucket rate limiter with per-tenant quotas."}
        ],
        "max_tokens": 600
      }'

Force Flash for a cheap task

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router-flash",
        "messages": [{"role": "user", "content": "Rename foo to bar in this function."}],
        "max_tokens": 200
      }'

Streaming

curl -sN "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router-flash",
        "stream": true,
        "messages": [{"role": "user", "content": "Walk me through merge sort."}],
        "max_tokens": 400
      }'

Tool calls (OpenAI shape)

The gateway forwards tool definitions to OpenRouter unchanged.

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router",
        "messages": [{"role": "user", "content": "What is the weather in Tokyo?"}],
        "tools": [{
          "type": "function",
          "function": {
            "name": "get_weather",
            "description": "Get current weather",
            "parameters": {
              "type": "object",
              "properties": {"city": {"type": "string"}},
              "required": ["city"]
            }
          }
        }],
        "tool_choice": "auto"
      }'

Multi-turn coding session

Send a real session ID and goal ID so the gateway can keep budget and runtime-evidence counters across requests:

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "X-Smart-Gateway-Session: proj-A-2026-08-23" \
  -H "X-Smart-Gateway-Goal: fix-flaky-test" \
  -H "X-Task-Id: task-17" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router",
        "messages": [
          {"role": "user", "content": "Make this test deterministic."},
          {"role": "assistant", "content": "Let me look at the timestamps."},
          {"role": "tool", "tool_call_id": "t1", "content": "FAIL: timestamps differ by 12 ms"}
        ]
      }'

Dry-run a routing decision

Don't talk to OpenRouter at all — ask the gateway what it would do:

curl -sS "$SMART_GATEWAY_BASE_URL/v1/route/decision" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "task": {
          "task_id": "t1",
          "goal_id": "g1",
          "kind": "AUTH",
          "instruction": "audit the refresh-token rotation logic",
          "sensitive": true,
          "risk_level": "HIGH",
          "blast_radius": "HIGH"
        },
        "runtime": {
          "attempts": 1,
          "same_error_count": 1,
          "files_touched": 6,
          "tool_depth": 4
        }
      }'

Returns:

{
  "task_id": "t1",
  "action": "ROUTE",
  "model": "deepseek_v4_pro",
  "purpose": "EXECUTE",
  "cause": "hard_pro_rule",
  "score": 50.0,
  "confidence": 0.98,
  "hard_rule": true,
  "evidence": ["task_kind=AUTH", "risk=HIGH"],
  "policy_version": "1"
}

Force Scout on a non-sensitive task

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "X-Scout: true" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router",
        "messages": [{"role": "user", "content": "Should I use a tree or a hashmap for this?"}],
        "metadata": {"sensitive": false}
      }'

Private content (the right way)

Use the default router and pass sensitive: true:

curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router",
        "messages": [{"role": "user", "content": "<paste internal code here>"}],
        "metadata": {"sensitive": true}
      }'

The default is sensitive: true; setting it explicitly is a good defensive habit.

Multipart / multimodal (rejected)

# This returns HTTP 400.
curl -sS "$SMART_GATEWAY_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $SMART_GATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "smart-router",
        "messages": [{
          "role": "user",
          "content": [
            {"type": "text", "text": "What is in this image?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/x.png"}}
          ]
        }]
      }'

v1.0.0 does not support multimodal. See Getting started → Concepts.