qyvos-api / README.md
The-JDdev
QYVOS API: initial backend deploy
99400b4
|
Raw
History Blame Contribute Delete
1.46 kB
metadata
title: QYVOS API
emoji: πŸ€–
colorFrom: indigo
colorTo: purple
sdk: docker
app_port: 7860
pinned: false

QYVOS API

FastAPI backend that powers the QYVOS Android app (https://github.com/The-JDdev/QYVOS).

The Android client keeps a lean Chaquopy environment and delegates the heavy OpenManus / OpenAI-compatible logic to this Hugging Face Space.

Endpoints

  • GET /healthz β€” liveness probe
  • POST /api/chat β€” main chat endpoint

Request

{
  "messages": [
    {"role": "user", "content": "Who created you?"}
  ],
  "model": "gpt-4o-mini",
  "max_tokens": 1024,
  "temperature": 0.7
}

base_url and api_key are optional per-request overrides.

Response

{
  "reply": "I am QYVOS, created by The-JDdev (SHS Shobuj).",
  "model": "gpt-4o-mini",
  "prompt_tokens": 42,
  "completion_tokens": 18,
  "total_tokens": 60
}

Configuration (Space β†’ Settings β†’ Variables and secrets)

Name Purpose
QYVOS_API_KEY Default upstream API key (OpenAI-compatible)
QYVOS_BASE_URL Optional custom base URL (e.g. DeepSeek)
QYVOS_MODEL Default model id (gpt-4o-mini if unset)
QYVOS_MAX_TOKENS Default max tokens (2048)
QYVOS_TEMPERATURE Default temperature (0.7)

Identity lock

The system prompt is hard-injected as the first message of every request so the assistant always responds as QYVOS, regardless of what the client sends.