LiteLLM can call a custom OpenAI-compatible endpoint with the openai/ provider prefix.
from openai import OpenAI
client = OpenAI(base_url="https://api.fikraapi.co.ke/v1", api_key="YOUR_FIKRA_KEY")
response = client.chat.completions.create(
model="openai/deepseek-v4.1-flash-1017",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)
Run the Python snippet and print the response content.
Default model: deepseek-v4.1-flash-1017. Prices come from the live catalog.
| Input | $0.05 / 1M |
| Cached input | $0.0075 / 1M |
| Output | $0.15 / 1M |
Repeated prompt prefixes are billed at the cached rate. Cached input can never cost more than fresh input.
While the window is open, model deepseek-v4-1-flash-economy-free is free. Check live availability.
The key is missing, revoked, or has a leading space. Paste it again from the dashboard and use the fk_live_ value.
The base URL usually has an extra /chat/completions. Use exactly https://api.fikraapi.co.ke/v1.
The model ID is case sensitive. Copy it from /models or the snippet on the dashboard.
You hit the account rate limit. Free accounts allow 30 requests per minute and 300 per day. Any top-up raises the limits.
Source used for this guide: https://docs.litellm.ai/docs/providers/openai_compatible. If the tool changes its settings, open an issue and we will update this page.
More guides: all setup guides. Measured reliability: public stats. Prices: pricing. Free model: free DeepSeek API.