Products
Two ways to put Teselo to work.
The same proven library, two ways in. Use Teselo as your endpoint, or keep your own model and put the gateway in front of it.
Teselo Endpoint
A model endpoint that learns.
A drop-in, OpenAI-compatible chat completions endpoint. Skills it has proven are served instantly; everything else goes to a frontier model, and what works is learned.
Python — OpenAI SDK
from openai import OpenAI
client = OpenAI(base_url="https://api.beyonditsparts.ai/v1", api_key="YOUR_TESELO_KEY")
client.chat.completions.create(model="[MODEL ID — LEARNING]", messages=[...])Learning
Learns from your traffic into your private library.
[MODEL ID — LEARNING]Frozen
Serves your existing library and keeps no request data.
[MODEL ID — FROZEN]Teselo Gateway
Your model. Minus the repeated work.
Keep the model and provider you already use. Put Teselo in front of it: proven skills answer instantly, and only new work reaches your model.
Bring your own model
Connect your provider in the console. Your provider key is stored encrypted on our servers and used only to forward requests you send.
Python — OpenAI SDK
from openai import OpenAI
client = OpenAI(base_url="https://api.beyonditsparts.ai/v1", api_key="YOUR_TESELO_KEY")
client.chat.completions.create(model="[GATEWAY MODEL ROUTE — e.g. gateway/your-upstream]", messages=[...])Side by side
Endpoint or Gateway?
| Aspect | Teselo Endpoint | Teselo Gateway |
|---|---|---|
| Who supplies the model | Teselo — requests that miss your library go to a frontier model. | You — connect your own provider and model. |
| Data retention options | Learning keeps proven skills in your private library; Frozen keeps no request data. | [TBD] |
| Setup time | One-line change: point your SDK at the Teselo base URL. | [TBD] |
| Best for | Teams who want a drop-in endpoint that learns from their traffic. | Teams who keep their own model and provider, and want repeated work to stop costing them. |
Not sure which fits?
Tell us about your workload and we'll help you choose.