p-typed Context API

Quick Start — Python & JS SDKs

Any OpenAI-compatible SDK can be re-pointed to p-typed Context API seamlessly. Create an API key in the Developer Portal, then use the following patterns:

Python SDK Setup

from openai import OpenAI

# Just point Base URL and provide your p-typed Key:
client = OpenAI(
    api_key="alfie_sk_live_xxxxxxxxxx...",
    base_url="https://api.157.230.94.55.sslip.io/v1",
)

response = client.chat.completions.create(
    model="grok-4.3",  # or gemini-3-flash-preview, deepseek/deepseek-v4-pro
    messages=[
        {"role": "system", "content": "Reply in one short sentence."},
        {"role": "user", "content": "What is the capital of Uganda?"},
    ],
    temperature=0.3,
)

print(response.choices[0].message.content)

Node.js SDK Setup

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "alfie_sk_live_xxxxxxxxxx...",
  baseURL: "https://api.157.230.94.55.sslip.io/v1",
});

const completion = await client.chat.completions.create({
  model: "grok-4.3",
  messages: [{ role: "user", content: "Say hello!" }],
});

Authentication

Authenticated calls require a Bearer key in the HTTP Authorization header:

Authorization: Bearer your_p_typed_api_key

API keys are formatted as alfie_sk_live_[32-hex-characters]. Programmatic operations count strictly toward your primary owner's subscription tier.

OpenAI-Compatible Embeddings

Create vectors in the real OpenAI text-embedding-3 semantic spaces through the same p-typed API key. Existing OpenAI SDK code works unchanged after setting the p-typed base URL.

from openai import OpenAI

client = OpenAI(
    api_key="alfie_sk_live_xxxxxxxxxx...",
    base_url="https://api.157.230.94.55.sslip.io/v1",
)

response = client.embeddings.create(
    model="text-embedding-3-small",
    input=["Ugandan employment law", "Constitutional land rights"],
    encoding_format="float",
)

print(len(response.data[0].embedding))  # 1536

text-embedding-3-small returns 1,536 values by default; text-embedding-3-large returns 3,072. Provider-prefixed aliases are accepted, but responses use the standard OpenAI model IDs.

Server-Sent Events (SSE) Streaming

Setting stream: true causes tokens to stream dynamically. In the Python SDK, simply loop over chunks:

stream = client.chat.completions.create(
    model="grok-4.3",
    messages=[{"role": "user", "content": "Explain quantum entanglement briefly."}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="", flush=True)

Totals are re-calculated and provided at the final chunk. (Note: server-side tool groups cannot be streamed).

p-typed Extension — Server-Side Tool Groups

The ukg_groups extension hooks the model directly into state-of-the-art server-side tools (e.g. Uganda legal archives), bypassing client-side processing roundtrips. Specify desired groups in `extra_body`:

response = client.chat.completions.create(
    model="grok-4.3",
    messages=[{"role": "user", "content": "What is the penalty for directorship breach?"}],
    extra_body={"ukg_groups": ["legal"]},  # Enable Uganda Knowledge Graph
)

Discover valid tool group tags from GET /v1/ukg/groups. This is non-streamable.

AI_OCR — Image Data Extraction

Accurately extract all data from images — receipts, IDs, forms, screenshots, documents — into structured JSON, without hallucination. Powered by Google Gemini flash models, tuned for faithful, verbatim transcription.

• Stateless: no memory or conversation context — each image is read in isolation.
• One image at a time: multiple images are queued and processed sequentially, returning a list with one extracted object per image (in order). This prevents data bleeding between images.
• Honest uncertainty: anything blurry, cut off, or ambiguous is reported in a notes field instead of being guessed.

Available models

Python example

import httpx

resp = httpx.post(
    "https://api.157.230.94.55.sslip.io/api/ai-ocr",
    headers={"Authorization": "Bearer ALFIE_KEY"},
    json={
        "images": [
            "data:image/png;base64,iVBORw0KGgo...",   # data URL
            "https://example.com/receipt.jpg",          # or a public URL
        ],
        "model": "gemini-2.5-flash-lite",
    },
    timeout=180,
)

for item in resp.json()["results"]:
    print(item["index"], item["confidence"], item["full_text"])
    if item["notes"]:
        print("  ⚠ uncertain:", item["notes"])

Each image consumes one unit of your daily quota. Failed images are automatically refunded. Accepted inputs: data: URLs, public http(s) URLs, or raw base64. Up to 20 images per request.

Rate Limits & Quotas

Programmatic API requests are rate-limited per minutes to safeguard downstream endpoints:

• Base rate limit: 60 requests / minute, per key.
• Daily quota: Shared across your subscription tier.

Unsuccessful model attempts do not deplete daily quotas. Limit parameters arrive in the HTTP headers: