DAppNodePackage-qvac/dappnode_package.json at main · dappnode/DAppNodePackage-qvac · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
{
  "upstream": [
    {
      "repo": "tetherto/qvac",
      "version": "0.12.0",
      "arg": "UPSTREAM_VERSION"
    }
  ],
  "shortDescription": "Private multimodal AI with OpenAI-compatible APIs",
  "description": "Run the complete [QVAC](https://qvac.tether.io/) CLI locally on DAppNode as an OpenAI-compatible AI provider. QVAC keeps prompts and inference on your own hardware. Its HTTP server supports text generation, embeddings, RAG, image and video generation, transcription, text-to-speech, and audio translation when the corresponding models are configured.\n\nA built-in configuration UI includes a searchable model library, validates and backs up settings, and applies changes without shell access. The QVAC worker reloads independently, so the UI remains available while models download or start. The default configuration serves the compact `QWEN3_600M_INST_Q4` text-generation model as `qwen3-600m`. Downloaded models, custom configuration, and backups are stored in a persistent volume.\n\nThis image includes QVAC's upstream worker and all native Linux addons shipped by the CLI. The default package does not reserve a GPU, so it installs and runs on CPU-only DAppNodes.\n\n- Configuration UI: `http://qvac.dappnode:8080`\n- API: `http://qvac.dappnode:11434/v1`\n- API documentation: `http://qvac.dappnode:11434/docs`\n- At least 4 GB RAM and 5 GB free disk are required for the default model; larger models require substantially more RAM and storage.\n- Set an API key during installation if clients outside your trusted DAppNode network can reach the service.",
  "type": "service",
  "mainService": "qvac",
  "author": "DAppNode Association <admin@dappnode.io> (https://github.com/dappnode)",
  "license": "Apache-2.0",
  "categories": ["AI"],
  "keywords": ["ai", "embeddings", "inference", "llm", "multimodal", "openai", "qvac", "rag", "speech", "tether"],
  "architectures": ["linux/amd64", "linux/arm64"],
  "dockerTimeout": "20min",
  "requirements": {
    "minimumDappnodeVersion": "0.2.106"
  },
  "links": {
    "ui": "http://qvac.dappnode:8080",
    "api": "http://qvac.dappnode:11434/v1",
    "docs": "http://qvac.dappnode:11434/docs",
    "homepage": "https://qvac.tether.io/"
  },
  "repository": {
    "type": "git",
    "url": "https://github.com/dappnode/DAppNodePackage-qvac.git"
  },
  "bugs": {
    "url": "https://github.com/dappnode/DAppNodePackage-qvac/issues"
  },
  "backup": [
    {
      "name": "QVAC configuration",
      "path": "/data/config"
    }
  ],
  "warnings": {
    "onInstall": "This package runs on CPU-only DAppNodes. CPU inference can be slow, so start with the default compact model before selecting larger models. Without an API key, QVAC is available without authentication to clients that can reach the DAppNode network. The configuration UI on port 8080 is administrative and should not be exposed publicly. The complete runtime image and configured model downloads can use many GB of storage. Large models may also require substantial RAM and take several minutes to load."
  }
}