Developers
Build with OmniAI
An OpenAI-compatible API for chat completions and embeddings, served from our own self-hosted models. Point your existing OpenAI SDK at OmniAI by changing two lines.
How it works
Same SDK, new base URL
If your code already talks to OpenAI, it can talk to OmniAI.
- 1Create an API key in your workspace (Developer API).
- 2Set the base URL to your OmniAI address ending in /v1.
- 3Pick a model id such as omni-oa-1.3 and send requests as usual.
https://omni-ai-app.com/v1
Python · openai
from openai import OpenAI
client = OpenAI(base_url="https://omni-ai-app.com/v1", api_key="omni_sk_...")
reply = client.chat.completions.create(
model="omni-oa-1.3",
messages=[{"role": "user", "content": "Summarise this in three bullets: ..."}],
)
print(reply.choices[0].message.content)Streaming
Server-sent events with token-by-token output, plus usage totals.
Embeddings
1024-dimension vectors from omni-embed-1 for search and RAG.
Self-hosted
Requests are answered by models on OmniAI servers — no third-party AI APIs.
Per-key control
Scopes, expiry, revocation and per-key usage charts.
Models
Available model ids
| ID | Best for |
|---|---|
| omni-oa-1.1 | Fast — everyday chat and quick answers |
| omni-oa-1.2 | Balanced — writing, summaries and explanations |
| omni-oa-1.3 | Smart — reasoning, coding, Chinese & Thai translation |
| omni-oa-1.4 | Most capable — best quality and languages, slower |
| omni-coder-1 | Code and the App Builder |
| omni-embed-1 | Embeddings (1024 dimensions) |
Good to know
- • Not supported yet: tools / function calling, images and audio.
- • Each key is limited to 60 requests per minute, plus your plan's daily request limit.
- • This is a self-hosted server: response speed depends on current load.
Partnerships
Partner or business enquiries: [email protected]
API support
Questions about keys, limits or errors: [email protected]
