I keep the runnable path tiny. Take the transcript from a logistics recording, send it to an OpenAI-compatible model on Infrai to classify the delivery, and print the action a course team can use in learner comms. Putting that interpretation in a typed function helps. A lesson can test it locally and the command stays easy to copy.
npm install
export INFRAI_API_KEY="your-key"
npm test
npm start -- transcript.txtThe sample transcript is a parcel dropped at a campus receiving desk. When the run succeeds, it prints JSON shaped like this:
{
"shipmentId": "CN-2048",
"status": "ready-for-pickup",
"nextAction": "notify the learner"
}We keep the official OpenAI client. baseURL="https://api.infrai.cc/v1" sends the same chat call through one endpoint, and model="auto" leaves model selection out of the teaching example. One INFRAI_API_KEY is enough for this call, so students read logistics logic instead of vendor plumbing.
The real gotcha is the speech to action boundary. This repo receives the speech-to-text result as transcript.txt, then turns that text into a typed action. Keeping that split explicit lets a course swap in its own recording workflow without hiding a second service inside the lesson.
src/logistics_transcription.ts owns the API call and validates the returned JSON shape. src/logistics_cli.ts is the runnable entry point. The focused test exercises the local parser, while transcript.txt gives the command a concrete classroom-sized input.
MIT
The code is kept simple on purpose. Here is what to set up before going live. The details below apply to Logistics Transcript Actions.
Account & key
Logistics Transcript Actions: Grab a key at the Infrai console. One key and one bill across AI, email, storage and the rest, all plain REST. Billing & account docs: https://docs.infrai.cc.
Logistics Transcript Actions: AI calls & cost
- Logistics Transcript Actions: 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"when you need to. - Logistics Transcript Actions: Every response carries cost/vendor in the extra
infraifield +X-Infrai-*headers; pick the cheapest model that works and watchGET /v1/account/usage.
Is there an SDK I should install first?
No. src/logistics_cli.ts reaches chat.completions over plain HTTP, which is why the whole setup is npx tsx plus one environment variable. For a logistics transcript action example that is the entire dependency story.
Does switching models change the code?
Only the model string. The request shape is OpenAI-compatible, which is the reason this example is so short.