Base URL: https://api.fikraapi.co.ke. Every route takes Authorization: Bearer fk_live_... and nothing else authenticates unless noted. Every example below is a paste-ready curl. Prices are per model at /models; billing is per 1M input/output tokens for chat and reasoning, per generation for images and video, per minute for speech-to-text, and per 1M characters for text-to-speech. Decision models bill per 1M input tokens, output is free.
OpenAI-compatible chat. Any chat model in the catalog: DeepSeek (chat, reasoning, multimodal), GLM, Kimi, Qwen, MiMo, StepFun, NVIDIA, Gemini. Image inputs work on multimodal models; pass images as base64 data URIs or public URLs.
curl https://api.fikraapi.co.ke/v1/chat/completions \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{ "model": "deepseek-v4-1-flash-economy-free", "messages": [{"role": "user", "content": "Say hi in one word."}] }'
// response (abbreviated) { "id": "chatcmpl-...", "model": "deepseek-v4-1-flash-economy-free", "choices": [{"index": 0, "message": {"role": "assistant", "content": "Hi."}, "finish_reason": "stop"}], "usage": {"prompt_tokens": 12, "completion_tokens": 3, "total_tokens": 15} }
The same engine as chat completions behind the Responses shape: input + instructions instead of messages. Tool calls arrive in output items. Use it if your agent framework targets the Responses API.
curl https://api.fikraapi.co.ke/v1/responses \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{ "model": "deepseek-v4.1-flash", "input": "Return the top 3 risks in this brief.", "instructions": "Be concise." }'
// response (abbreviated) { "id": "resp-...", "object": "response", "status": "completed", "output": [{"type": "message", "content": [{"type": "output_text", "text": "..."}]}] }
A decision model does not write a reply. It picks or scores one of the fixed answers you pass in. Billing is per 1M input tokens; output is free. The model is fixed server-side as gemini-decision-1. Body: input (what to decide over) and questions (the fixed options).
# Example: triage a support ticket to one of three routes curl -X POST https://api.fikraapi.co.ke/v1/decisions \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{"input": "The card swipe failed at checkout and the customer says they were charged twice.", "questions": {"route": ["billing", "fraud", "technical"]}}'
// response (abbreviated) { "candidates": [{"content": {"parts": [{"text": "billing"}]}, "tokenCount": 1}], "usage": {"totalInputTokens": 47, "totalOutputTokens": 0} }
Use a decision model when the answer is one of a known set, latency matters, and a full chat answer is waste. Use chat instead when the task needs reasoning, tool calls, or free-form text. Billed $0.12 / 1M input tokens. See the decision models page for where they fit.
Vector embeddings via qwen-embedding. Pass a string or array of strings.
curl https://api.fikraapi.co.ke/v1/embeddings \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{"model": "qwen-embedding", "input": ["invoice format", "PDF parsing"]}'
// response (abbreviated) { "data": [ {"object": "embedding", "index": 0, "embedding": [0.0123, -0.0483, ...]}, {"object": "embedding", "index": 1, "embedding": [0.0021, -0.0012, ...]} ], "model": "qwen-embedding" }
OpenAI-compatible transcription shape. Multipart upload, model fikra-stt, billed per minute of audio.
curl https://api.fikraapi.co.ke/v1/audio/transcriptions \ -H "Authorization: Bearer fk_live_..." \ -F "[email protected]" \ -F "model=fikra-stt"
// response { "text": "the balance is due on the 30th", "model": "fikra-stt" }
Text to speech, model fikra-tts, billed per 1M characters.
curl https://api.fikraapi.co.ke/v1/audio/speech \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{"model": "fikra-tts", "input": "Your order ships tomorrow.", "voice": "default"}' \ --output speech.mp3
// response: binary audio stream (audio/mpeg), 200 OK
OpenAI-style image generation. Paid models: seedream-5.0-pro, seedream-5.0-flash, seedream-5.0-lite, seedream-4.5, seedream-4.0. seedream-5-5 is limited to 3 generations per account per day and is account-only.
curl https://api.fikraapi.co.ke/v1/images/generations \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{"model": "seedream-5.0-pro", "prompt": "A studio product shot of a blue ceramic mug."}'
// response (abbreviated) { "created": 1763400000, "data": [{"url": "https://.../generated.png"}] }
Text-to-video and image-to-video. Today the account serves minimax-h3 (Fikra retail $0.50 per generation). The endpoint also polls async jobs. Seedance models are not enabled for this account yet and will not work.
curl -X POST https://api.fikraapi.co.ke/v1/videos/generations \ -H "Authorization: Bearer fk_live_..." \ -H "Content-Type: application/json" \ -d '{"model": "minimax-h3", "prompt": "A slow dolly shot over a Nairobi skyline at dusk.", "duration": 6}'
// response: 202 Accepted, poll for the finished asset { "id": "cgt-...", "status": "processing", "model": "minimax-h3", "poll": "/v1/videos/generations/cgt-..." }
# Poll the job until status is completed curl https://api.fikraapi.co.ke/v1/videos/generations/cgt-... \ -H "Authorization: Bearer fk_live_..."
curl https://api.fikraapi.co.ke/v1/models \ -H "Authorization: Bearer fk_live_..."
Returns the live catalog: id, display_name, context_window, status, pricing. Same registry that backs /models.
curl https://api.fikraapi.co.ke/v1/models/deepseek-v4.1-flash \ -H "Authorization: Bearer fk_live_..."
curl https://api.fikraapi.co.ke/v1/balance \ -H "Authorization: Bearer fk_live_..."
Returns the authenticated key's owner balance in USD (and KES where applicable).
curl "https://api.fikraapi.co.ke/v1/usage?days=30" \ -H "Authorization: Bearer fk_live_..."
Per-day aggregates for the authenticated key (key_id optional, days 1..90, default 30): requests, tokens, cost.
Every endpoint bills from the same prepaid wallet. No seat fees. Limits per tier at /pricing. Service-quality telemetry at /stats. Create an account at /account/signup, or read the setup guides.