cydeai-client Wheel Release Usage Notes

Replace 0.1.9 with the released version.

Install

python3 -m venv ~/cydeai-client-env
. ~/cydeai-client-env/bin/activate
python -m pip install -U pip
python -m pip install cydeai-client==0.1.9
cydeai --help

Configure

Create ~/.cydeai/config.json:

mkdir -p ~/.cydeai
cat > ~/.cydeai/config.json <<EOF
{
  "server_url": "https://<server-host>/<base-path>",
  "api_key": "<provider_api_key>",
  "worker_name": "<worker-display-name>",
  "model_name": "<served-model-name-from-vllm>",
  "model_port": 8000,
  "vllm_health_interval_sec": 30,
  "vllm_health_timeout_sec": 5,
  "vllm_health_failure_threshold": 2,
  "debug": false
}
EOF
chmod 600 ~/.cydeai/config.json

The client reads values in this order:

CLI args > environment variables > config file > defaults

Check Local vLLM

cydeai check sends a tiny OpenAI-compatible chat request to the local model server. It does not inspect GPU utilization or require nvidia-smi.

cydeai check

Expected success:

[INFO][2026-07-09 10:00:00] vLLM normal

If it fails, confirm that vLLM is running and that model_name matches the served model name from:

curl -sS http://127.0.0.1:8000/v1/models | python3 -m json.tool

Dry Run

cydeai connect --dry-run

Dry run prints the derived WebSocket URL, masked PROVIDER_SIGN header, register payload, and local vLLM check result. It does not connect to the server and does not print the raw API key.

One-Shot Register Test

cydeai connect --exit-after-register

This validates WebSocket handshake and register_ok, then exits.

Start Worker

cydeai connect

Keep this process running. Logs are printed to the terminal and appended to:

~/.cydeai/cydeai.log

Output format:

[INFO][2026-07-09 10:00:00] connect attempt 1: wss://...
[INFO][2026-07-09 10:00:00] registered: attempt=1 worker_id=w-... status=available

After registration, the client checks local GET /v1/models every 30 seconds by default. After two consecutive failures, it closes the current gateway connection, keeps the process alive, and reconnects after vLLM recovers. Failures include host, port, URL, request body, response body, status code, response time, and reason in the log.

Debug Server Errors

Normal mode logs only stable error categories such as http_403 or websocket_closed_1008. To include the error description returned by the server, run:

cydeai connect --debug

Debug mode can also be enabled with "debug": true in the config file or CYDEAI_DEBUG=true. Server details are normalized to one line, truncated, and the active provider API key is masked before logging.

Release Notes Checklist

• package name: cydeai-client
• CLI command: cydeai
• version: 0.1.9
• install command: python -m pip install cydeai-client==0.1.9
• required Python version: Python 3.11+
• vLLM must already be running locally
• config path: ~/.cydeai/config.json