Check how your AI route handles failures.

Send one test request

This sends a normal request and shows its response headers. To test recovery from a 429, use two authorized routes and a controlled rate-limit case in your own setup.

# One routed request. Replace the key placeholder with your generated key.
# A normal response shows the route; this command does not trigger a 429.

export OPENAI_BASE_URL=https://api.tekiz.ai/v1
export OPENAI_API_KEY=sk_tekiz_your_key_here

curl -i "$OPENAI_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tekizai/frontier",
    "messages": [{"role": "user", "content": "Return one sentence about provider failover."}]
  }'

# Inspect response headers for route/debug signals:
# X-TEKIZAI-Router-Provider
# X-TEKIZAI-Router-Model
# X-TEKIZAI-Router-Modalities
# Retry-After on hosted quota/rate-limit responses

# Public edge health:
# https://api.tekiz.ai/edge/health

Example provider key setup

This illustrates two authorized keys on a router you run. Check the local router guide for the setup steps that fit your provider.

# Provider credential pool shape
# Configure multiple authorized keys or subscription paths for the same provider.

providers:
  openai:
    loadBalance: round-robin
    apiKeys:
      - env: OPENAI_API_KEY_PRIMARY
      - env: OPENAI_API_KEY_SECONDARY
  ollama:
    oauthPaths:
      - ~/.ollama/primary
      - ~/.ollama/backup

# Behavior:
# - pick a working key using the configured order
# - fail over on retryable 429/5xx responses
# - keep provider credentials on the trusted router host
# - expose one generated sk_tekiz key to apps and teammates

Send an image request

Use an account with an image-capable model. This example uses a public TEKIZ.AI image; check the response and served model before trying your own files.

# Image routing example. Set OPENAI_BASE_URL and OPENAI_API_KEY first.

curl "$OPENAI_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tekizai/frontier",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "Describe the image safely."},
        {"type": "image_url", "image_url": {"url": "https://tekiz.ai/visuals/community-marketplace-network.png"}}
      ]
    }]
  }'

# Router invariant:
# - route image requests only to image-capable models
# - route audio/video/document inputs only to capable models
# - exclude text-only GLM-style routes unless that model is declared or learned capable
# - record X-TEKIZAI-Router-Modalities for the served request

Copyable Reddit reply

Before wiring TEKIZ.AI into an agent, I would check three things:

1. 429/provider failure behavior through one OpenAI-compatible endpoint.
2. Multiple authorized provider keys or subscription paths behind one provider profile.
3. Multimodal routing constraints so image/audio/video/document input only reaches capable models.

Reliability guide:
https://tekiz.ai/reliability-proof?utm_source=reddit&utm_medium=community&utm_campaign=tekizai-launch&utm_content=proof-reply

Public edge health:
https://api.tekiz.ai/edge/health

Quickstart:
https://tekiz.ai/quickstart?utm_source=reddit&utm_medium=community&utm_campaign=tekizai-launch&utm_content=proof-reply

TEKIZ.AI provides routing, API keys, usage limits, request history, and retries when an eligible route is available. Use provider accounts only with permission. Keep credentials and customer data out of public posts.