One API.
The right model, every time.
Route chat, reasoning, code, and vision across 14 leading AI providers. Auto picks the model, fails over when providers wobble, and keeps spend visible in one credit balance.
prompt: "Review this auth middleware for vulnerabilities"
Describe the job.
Auto handles the model.
Semantic routing evaluates task type and complexity locally, then chooses a model and a resilient provider path before inference begins.
Powered by lee101/pybed — low-allocation BERT tokenization and optional custom NVIDIA CUDA kernels.
Routing without surrendering control.
Pick an exact model when you need determinism. Choose an Auto profile when you want the router to balance quality, latency, and price.
Read the API docs arrow_forwardProvider failover
Retry transient upstream failures through model-aware fallback chains.
One credit balance
Fund once, use every routed model, and keep usage in one account.
Your keys or ours
Use the managed gateway now and retain a clean path to provider keys.
Privacy first
Local task classification keeps routing fast and avoids an extra AI call.
from openai import OpenAI
client = OpenAI(
base_url="https://api.text-generator.io/v1",
api_key="YOUR_API_KEY",
)
response = client.chat.completions.create(
model="text-generator/auto",
messages=[{"role": "user",
"content": "Ship something great"}],
)Keep the SDK.
Change the destination.
Point the OpenAI client at Text Generator and unlock the provider network without rewriting your application.
- checkChat completions and streaming-compatible responses
- checkAliases, Auto profiles, and exact model IDs
- checkRouting metadata and provider response headers
}Open source all the way down.
The gateway, auto-router, tokenizer, embeddings, and search index are inspectable. Fork it, self-host it, or help us make routing better.
One endpoint. More ways through.
Build against the model network today and keep your options open tomorrow.