Inference API (OpenAI-Compatible)
ds4-server exposes an OpenAI-compatible API at http://127.0.0.1:8000/v1.
Authorization: Bearer *** (if configured)
Chat Completions
POST /v1/chat/completions
Available models:
| Model | Description |
|---|---|
deepseek-v4-flash |
Primary inference model |
deepseek-v4-pro |
Alias served by the same loaded model |
Endpoints
| Endpoint | Method | Description |
|---|---|---|
/v1/chat/completions |
POST | Chat completion (streaming or non-streaming) |
/v1/models |
GET | List available models with metadata |
Hardware Bridge (JetKVM) — Retired
The standalone bridge that exposed http://127.0.0.1:8005 is retired; desktop automation now runs through Hermes Agent's computer-use tools driving the JetKVM directly. The endpoint table below is retained for historical reference.
| Endpoint | Method | Description |
|---|---|---|
/stats |
GET | Bridge status: frame count, resolution, frame age |
/screenshot |
GET | Capture desktop screenshot — returns base64 JPEG + metadata |
/record |
POST | Start video recording (output_path, duration, fps, scale) |
/record/status |
GET | Recording progress |
/record/stop |
POST | Stop current recording |
/edid |
POST | Set EDID profile on JetKVM (edid_name) |
/mouse/move |
POST | Move mouse to (x, y) in real screen coordinates |
/mouse/click |
POST | Click at (x, y) with button |
/keyboard/type |
POST | Type a text string |
/keyboard/press |
POST | Press one or more keys |
Screenshot response schema:
{
"image_base64": "...",
"width": 1920,
"height": 1080,
"timestamp": "2026-02-22T02:00:00Z",
"letterbox_scale": 0.74375,
"letterbox_padding": [0, 46],
"mouse_position": {
"real": {"x": 960, "y": 540},
"fara": {"x": 713, "y": 447},
"confidence": 1.0
}
}
Hermes Agent Integration
Connect Hermes Agent to ds4-server as a custom LLM provider:
Provider type: openai-compatible
Base URL: http://127.0.0.1:8000/v1
API Key: (optional, ds4-server defaults to no-auth)
Default model: deepseek-v4-flash
Hermes Agent handles channel routing (Telegram, Signal, WhatsApp), scheduling, memory indexing, and skill execution — ds4-server provides the inference layer.
Integration Notes
- All services communicate over localhost; no public ports are exposed by default
- ds4-server handles inference; Hermes Agent handles tool dispatch, memory, and channel routing
- Desktop automation runs through Hermes Agent's computer-use tools (JetKVM WebRTC + USB HID); the legacy port-8005 bridge is retired