Skip to content

Commit a7cd9f3

Browse files
l3utterflyl3utterfly
committed
added readme
1 parent c24d133 commit a7cd9f3

1 file changed

Lines changed: 131 additions & 0 deletions

File tree

‎ReadMe.md‎

Lines changed: 131 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,131 @@
1+
# Layla Server
2+
3+
**Run powerful LLM models on your PC. Chat from anywhere on your phone.**
4+
5+
Layla Server is a lightweight wrapper around `llama-server` (or any OpenAI-compatible inference engine) that exposes your local LLM over WebRTC — so you can connect to it from anywhere, just by scanning a QR code with the [Layla app](https://www.layla-network.ai).
6+
7+
---
8+
9+
## Why Layla Server?
10+
11+
Your PC has the horsepower to run large, capable models. Your phone has the convenience. Layla Server bridges the two — no port forwarding, no static IP, no cloud subscription required.
12+
13+
- ✅ **One QR code** to connect your phone to your PC's LLM
14+
- ✅ **Works anywhere** — WebRTC punches through NAT and firewalls
15+
- ✅ **Long-lived sessions** — polling-based signalling keeps the server up indefinitely
16+
- ✅ **One-click installer** — ships with a bundled `llama-server` snapshot, no setup needed
17+
- ✅ **BYOM** — bring your own GGUF models
18+
- ✅ **Swap backends** — switch to any OpenAI-compatible inference engine in settings
19+
20+
---
21+
22+
## How It Works
23+
24+
```
25+
┌─────────────────────────────┐ WebRTC ┌──────────────┐
26+
│ Your PC │ ◄──────────────────► │ Layla App │
27+
│ │ │ (iPhone / │
28+
│ llama-server ←→ Layla │ QR code │ Android) │
29+
│ (or any OpenAI-compat API) │ └──────────────┘
30+
└─────────────────────────────┘
31+
```
32+
33+
1. Layla Server starts `llama-server` (or connects to your preferred backend)
34+
2. It generates a QR code encoding the WebRTC connection offer
35+
3. Scan the QR code with the Layla app — a peer-to-peer connection is established
36+
4. All OpenAI-compatible HTTP calls from the app are proxied over WebRTC to your local model
37+
5. Reconnect any time by scanning a new QR code — the server stays alive via polling-based signalling
38+
39+
---
40+
41+
## Quick Start
42+
43+
### Option 1 — One-Click Install (Recommended for non-technical users)
44+
45+
Download the latest release for your platform from the [Releases page](https://github.com/your-org/layla-server/releases).
46+
47+
Layla Server ships with a bundled `llama-server` snapshot — no separate installation required (you can easily swap it out with newer versions or even different servers).
48+
49+
1. Download and unzip the release
50+
2. Drop your `.gguf` model file into the `models/` folder
51+
3. Launch `layla-server`
52+
4. Open the Layla app on your phone and scan the QR code
53+
54+
That's it. You're now chatting with your local model from your phone.
55+
56+
---
57+
58+
### Option 2 — From Source
59+
60+
**Prerequisites:** Node.js 18+ (or your runtime of choice)
61+
62+
```bash
63+
git clone https://github.com/your-org/Layla-Server.git
64+
cd Layla-Server
65+
npm install
66+
npm start
67+
```
68+
69+
Build an `.exe` for Windows:
70+
71+
```bash
72+
npm run dist
73+
```
74+
---
75+
76+
## Configuration
77+
78+
All settings are available in the **Settings** panel of the Layla Server UI.
79+
80+
| Setting | Description |
81+
|---|---|
82+
| **Model path** | Path to your `.gguf` model file |
83+
| **Inference backend** | URL of any OpenAI-compatible server (default: bundled `llama-server`) |
84+
| **Additional Cmd Args** | Additional command-line arguments to pass to the server executable |
85+
86+
### Using a Custom Backend
87+
88+
Layla Server proxies to any OpenAI-compatible endpoint. To use a different inference engine (LM Studio, Ollama, vLLM, etc.), just point the backend URL to it in Settings — no other changes needed.
89+
90+
---
91+
92+
## Using Your Own Models
93+
94+
Layla Server supports any model in **GGUF format**.
95+
96+
1. Download a GGUF model (e.g. from [Hugging Face](https://huggingface.co/models?library=gguf))
97+
2. Select your model in the settings page
98+
3. Restart the server — the new model is loaded automatically
99+
100+
---
101+
102+
## Connecting the Layla App
103+
104+
1. Launch Layla Server on your PC
105+
2. A QR code will appear in the UI
106+
3. Open the **Layla app** on your phone
107+
4. Tap **Connect** → **Scan QR code**
108+
5. Point your camera at the code — connection is established instantly
109+
110+
The server uses **polling-based WebRTC signalling**, so it stays alive and connectable for as long as it's running. You can close the app, leave for hours, and reconnect whenever you like (no need to scan QR code again).
111+
112+
---
113+
114+
## Architecture Notes
115+
116+
- **WebRTC transport** — peer-to-peer data channel; traffic goes directly between your devices once connected
117+
- **Polling signalling** — no persistent WebSocket required; the server polls for new connection offers at a configurable interval, making it robust for long-running sessions
118+
- **OpenAI-compatible proxy** — the WebRTC channel transparently forwards HTTP requests and responses, so the Layla app doesn't need to know anything about the underlying transport
119+
120+
---
121+
122+
## Related
123+
124+
- [Layla App](https://www.layla-network.ai) — the offline LLM mobile client
125+
- [llama.cpp](https://github.com/ggml-org/llama.cpp) — the inference engine powering the bundled backend
126+
127+
---
128+
129+
## License
130+
131+
MIT © Layla Network Pty Ltd

0 commit comments

Comments
 (0)