Quickstart

Upfyn gives you three models behind one endpoint. The API speaks the OpenAI Chat Completions dialect, so if you have written against that before, you already know this one — point your existing client at a different base URL and change the model id.

Get a key#

1. 1

   Create an API key

   Go to API Keys in your dashboard and create one. Keys look like upfynai-sk- followed by 32 characters.
2. 2

   Copy it now

   The key is shown once. Afterwards the dashboard only ever displays the last four characters, because we store a hash rather than the key itself. If you lose it, revoke it and make another.
3. 3

   Keep it server-side

   A key carries your credits. Put it in an environment variable and call the API from your own backend — never from a browser or a mobile app, where anyone can read it.

Make your first call#

The base URL is https://ai.upfyn.com/v1. Authentication is a bearer token.

1curl https://ai.upfyn.com/v1/chat/completions \
2  -H "Authorization: Bearer $UPFYN_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d {
5    "model": "upfyn-yuva",
6    "messages": [
7      { "role": "user", "content": "Explain what a gateway does, in two sentences." }
8    ]
9  }

What comes back#

A standard chat.completion object, plus a few fields of our own so you never have to make a second call to find out what a request cost.

1{
2  "id": "chatcmpl-4f1c...",
3  "object": "chat.completion",
4  "created": 1787950410,
5  "model": "upfyn-yuva",
6  "choices": [
7    {
8      "index": 0,
9      "message": { "role": "assistant", "content": "A gateway sits between..." },
10      "finish_reason": "stop"
11    }
12  ],
13  "usage": {
14    "prompt_tokens": 18,
15    "completion_tokens": 42,
16    "total_tokens": 60
17  },
18  "credits_used": 0.31,
19  "credits_remaining": 4821.44,
20  "request_id": "req_9a2f...",
21  "session_id": null
22}

Pick a model#

There are three, and the id is the whole choice — you never name a provider.

Models covers what each one accepts and produces in full.

Stream the answer#

Set stream: true. Add stream_options.include_usage if you want the final cost — without it, a streaming call never tells you what it settled at.

1stream = client.chat.completions.create(
2    model="upfyn-yuva",
3    messages=[{"role": "user", "content": "Write a haiku about latency."}],
4    stream=True,
5    stream_options={"include_usage": True},   # ask for the final usage + credits chunk
6)
7 
8for chunk in stream:
9    if chunk.choices and chunk.choices[0].delta.content:
10        print(chunk.choices[0].delta.content, end="", flush=True)

Streaming explains the chunk format, how the stream terminates, and how to resume one that dropped.

Where to go next#

• Tool calling — let the model call your functions.
• Structured outputs — get JSON that matches a schema.
• Errors — every status code and what to do about it.
• Rate limits — the actual numbers, and the headers that report them.