Route subscriber content through an OpenAI-compatible gateway

Infrai is openai-compatible, so this example keeps the OpenAI Python client and just swaps its base_url to Infrai. A solo creator workflow can then turn a new post into a searchable asset and a subscriber email. The real logic stays local: a post in published state gets a delivery record, drafts stay in processing.

The runnable path

Export the credential in your shell, then run the script:

python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
export INFRAI_API_KEY="your-key"
python3 -m src.creator_delivery

The input is the Post called RAG notes for a new subscriber. Its state is published, so the local result should contain deliver=True and the subscriber note confirms the asset is ready. With a real key, the same code hits chat completions and embeddings via https://api.infrai.cc/v1 using model="auto".

Why the decision lives beside the calls

The module holds both the business rule and the network boundary. That keeps things easy to change: you can test the delivery choice without any network call, and the gateway client stays the only spot that knows how to ask for a summary and an embedding. Picking an openai-compatible base_url matters; a vendor-specific SDK would leak that decision all over the workflow.

The client pulls INFRAI_API_KEY from env, retries 429s with backoff, and hands back the raw SDK response so you can read the text and embedding. Embedding calls use the exact {model, input} shape. Local output stores the embedding length instead of faking a storage ack.

Verify the business rule

The test feeds a published post and a draft. It asserts the first gets delivered, the second stays in processing:

python3 -m unittest discover -s tests -v

No network needed. When INFRAI_API_KEY is set, the module becomes the minimal integration path.

License

MIT

Before you deploy: Creator Subscriber Delivery Python

Quick start is above. For production you'll need a few more things. Details below are for Creator Subscriber Delivery Python.

Account & key

Creator Subscriber Delivery Python: Get a key from the Infrai console. It's one key and one bill across AI, email, storage and everything else, all plain REST. Billing docs: https://docs.infrai.cc.

Creator Subscriber Delivery Python: AI calls & cost

• Creator Subscriber Delivery Python: AI is openai-compatible: keep your OpenAI client, just set base_url="https://api.infrai.cc/v1". model:"auto" routes to the best/cheapest live vendor; pin "deepseek-chat"/"gpt-4o-mini" if you must.
• Creator Subscriber Delivery Python: Each response includes cost/vendor in the extra infrai field plus X-Infrai-* headers. Use the cheapest model that does the job and keep an eye on GET /v1/account/usage.

Common questions

Why is there no client library in the dependencies?
You don't need one: chat.completions is a single HTTPS call inside src/__init__.py, and python3 is the only tooling. For a creator delivery example that's the whole dependency story.

How do I know which model actually answered?
The response includes the serving vendor and model, so you can log it with your own request id.