What is the Fikra API base URL?

The production base URL for Fikra API is https://api.fikraapi.co.ke/v1. It is a drop-in replacement for OpenAI endpoints.

How does Fikra API handle rate limits?

Fikra API uses a Redis-backed token bucket algorithm. Unverified accounts are limited to 30 Requests Per Minute (RPM). Making your first Paystack top-up permanently upgrades you to the Trusted tier at 100 RPM.

Which AI models are available on Fikra API?

We currently route to three primary models: fikra-fast-8b for low-latency tasks, fikra-pro-20b for balanced production workflows, and fikra-pro-120b for high-reasoning tasks.

Is Fikra API OpenAI compatible?

Yes, it is 100% compatible with the OpenAI API specification. You can use standard Python, Node.js, and Go SDKs by simply changing the base_url and injecting your Fikra API key.

How do I authenticate my API requests?

All API requests must include an Authorization header using the Bearer token scheme. Example: Authorization: Bearer fk_live_YOUR_KEY.

// API Reference v1.0

What is the Fikra API and how do I get started?

Fikra API exists to provide high-performance, OpenAI-compatible AI inference for developers in Africa. By serving as a unified proxy engine, we abstract the complexity of localized NPU infrastructure, allowing you to integrate world-class language models using familiar SDKs, powered entirely by transparent M-Pesa billing.

How does Fikra API infrastructure perform?

Our infrastructure acts as a secure cryptographic gateway to enterprise inference hardware. To ensure seamless migration, the Fikra API was engineered to strictly adhere to the OpenAI v1 REST specification. You only need to update your base URL and authenticate with your Fikra API Key.

Specification Parameter	Production Value	Implementation Detail
Base URL	https://api.fikraapi.co.ke/v1	The global endpoint for all inference routing.
Authentication Scheme	Authorization: Bearer	Requires active `fk_live_...` key generation.
Streaming Protocol	Server-Sent Events (SSE)	Supported via `stream: true` payload boolean.
SDK Compatibility	OpenAI SDK v1.0+	Native support for Python, Node.js, Go, and Rust.

What are the Fikra API production rate limits?

To guarantee high availability and protect our inference nodes from abusive scripts, Fikra API employs a Redis-backed Token Bucket algorithm. Your throughput capacity (RPM) is determined dynamically based on the verification status of your account ledger.

Account Tier	Requirement	Request Limit (RPM)	Token Behavior
Unverified	Default state upon sign-up	30 RPM	Uses initial free token allocation.
Trusted	First successful Paystack top-up	100 RPM	Permanently unlocked for production scale.

Note: Exceeding these limits will result in a standard `429 Too Many Requests` HTTP response.

Which Fikra API model should I choose?

Fikra API routes traffic to a curated selection of highly optimized open-weights models. Ensure you pass the exact model string in your JSON payload to avoid `400 Bad Request` routing errors.

Model String ID	Context Window	Primary Use Case	Reasoning Depth
fikra-fast-8b	8,192 tokens	Real-time chat, basic routing, summarization.	Low / Moderate
fikra-pro-20b	32,768 tokens	Production workflows, agentic tasks, RAG.	High
fikra-pro-120b	128,000 tokens	Complex coding, deep analysis, logical deduction.	Maximum

How do I send my first inference request?

Because we map exactly to the OpenAI spec, you do not need to learn a new library. Below are examples of how to instantiate the client and request a chat completion in popular languages.

Python (OpenAI SDK)

import openai
import os

# 1. Initialize client with Fikra Base URL and API Key
client = openai.OpenAI(
    base_url="https://api.fikraapi.co.ke/v1",
    api_key=os.getenv("FIKRA_API_KEY")
)

# 2. Execute standard Chat Completion
response = client.chat.completions.create(
    model="fikra-fast-8b",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello, World!"}
    ]
)

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

cURL (Raw HTTP)

curl https://api.fikraapi.co.ke/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer fk_live_YOUR_API_KEY" \
  -d '{
    "model": "fikra-pro-20b",
    "messages": [
      {
        "role": "user",
        "content": "Write a python script to reverse a string."
      }
    ],
    "temperature": 0.7
  }'

What does a successful API response look like?

When a request completes successfully, the Fikra API returns a standard `200 OK` JSON object. This object contains the generated text, a unique ID, and the exact token usage that was deducted from your global ledger.

JSON Output

{
  "id": "chatcmpl-9b8d4f21",
  "object": "chat.completion",
  "created": 1780665819,
  "model": "fikra-fast-8b",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help you build today?"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 9,
    "total_tokens": 23
  }
}

What do Fikra API error codes mean?

Our gateway provides explicit error messaging to accelerate debugging. Below is the reference table for HTTP status codes returned by the `/v1` endpoint.

Status Code	Error Definition	Resolution Path
400 Bad Request	Unsupported Model / Invalid JSON	Verify your payload uses a valid Fikra model string.
401 Unauthorized	Invalid API Key	Ensure your Bearer token matches your dashboard key exactly.
402 Payment Required	Insufficient Balance	Your global token wallet is empty. Top up via M-Pesa.
429 Too Many Requests	Rate Limit Exceeded	Implement exponential backoff or upgrade to Trusted Tier.
500 Internal Error	Proxy Connection Failure	Check the Fikra API status page or contact developer support.

Next Steps

Authentication Details →

Deep dive into securing your keys and generating JWTs.

Developer Hub

Go to Dashboard ↗

Generate your keys and monitor your live token usage.