Hugging Face's logo

Mr-J-369
/
Fancy-AI

Model card Files
xet
 Community
2
Mr-J-369 commited on
Commit
403d6b8
·
verified ·
1 Parent(s): 137b759

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +55 -65
README.md CHANGED
@@ -1,65 +1,55 @@
1
- ---
2
- license: other
3
- language:
4
- - en
5
- tags:
6
- - gguf
7
- - llama.cpp
8
- - android
9
- - fancy-ai
10
- - on-device
11
- pretty_name: Fancy AI — Add-on Catalog
12
- ---
13
-
14
- # Fancy AI — Add-on Catalog
15
-
16
- Downloadable add-ons for the **Fancy AI** Android app — an on-device AI companion app
17
- powered by `llama.cpp`. The app ships small and streams optional components (chat models
18
- and the background "Root" agent model) from this repo on demand.
19
-
20
- - App & issues: https://github.com/Mr-J-369/Fancy-Ai
21
- - This repo hosts: `manifest.json` (the catalog the app reads) + the `.gguf` weight files.
22
-
23
- ## How it works
24
-
25
- In the app: **Settings → On-Device Model → Browse Add-ons**. The app reads `manifest.json`
26
- from this repo, lists each component, and lets you **download + activate** in one tap:
27
-
28
- - `type: "agent"` → **Use as Root's brain** (tiny model for background tasks).
29
- - `type: "chat"` → **Use as chat model** (your main conversational model).
30
-
31
- Only model weights are downloaded — never executable code.
32
-
33
- ## Catalog format (`manifest.json`)
34
-
35
- A single JSON file at the repo root. Each component is single-file (`url`) or multi-file (`files[]`):
36
- ```
37
- {
38
- "components": [
39
- {
40
- "id": "root-qwen2.5-0.5b",
41
- "type": "agent",
42
- "name": "Root Agent — Qwen2.5 0.5B",
43
- "description": "Tiny, fast brain for background tasks.",
44
- "quant": "Q4_K_M",
45
- "minRamMb": 1500,
46
- "url": "https://huggingface.co/Mr-J-369/Fancy-AI/resolve/main/qwen2.5-0.5b-instruct-q4_k_m.gguf",
47
- "sizeBytes": 398000000,
48
- "sha256": "<sha256-of-the-file>"
49
- },
50
- {
51
- "id": "chat-llama3.2-3b",
52
- "type": "chat",
53
- "name": "Chat — Llama 3.2 3B Instruct",
54
- "description": "Balanced on-device chat brain for mid-range phones.",
55
- "quant": "Q4_K_M",
56
- "minRamMb": 4000,
57
- "url": "https://huggingface.co/Mr-J-369/Fancy-AI/resolve/main/llama-3.2-3b-instruct-q4_k_m.gguf",
58
- "sizeBytes": 2020000000,
59
- "sha256": "<sha256-of-the-file>"
60
- }
61
- ]
62
- }
63
- ```
64
- These are quantized redistributions of upstream models; each retains its original license
65
- (e.g. the Qwen and Llama community licenses). See each base model's card for terms.
 
1
+ ---
2
+ license: other
3
+ language:
4
+ - en
5
+ tags:
6
+ - gguf
7
+ - llama.cpp
8
+ - android
9
+ - fancy-ai
10
+ - on-device
11
+ pretty_name: Fancy AI — Add-on Catalog
12
+ ---
13
+
14
+ # Fancy AI — Add-on Catalog
15
+
16
+ Add-on catalog for the **Fancy AI** Android app — an on-device AI companion app powered by
17
+ `llama.cpp`. The app ships small and streams optional models on demand from this catalog.
18
+
19
+ - App & issues: https://github.com/Mr-J-369/Fancy-Ai
20
+ - This repo hosts **`manifest.json`** — the catalog the app reads. The model weights themselves
21
+ are **not re-hosted here**; the catalog links directly to the original creators' repos
22
+ (Qwen, bartowski), so downloads come straight from the source.
23
+
24
+ ## How to use it (in the app)
25
+
26
+ **Settings → On-Device Model → Browse Add-ons.** The app reads `manifest.json`, lists each
27
+ component, and lets you **download + activate** in one tap:
28
+
29
+ - `agent` → **Use as Root's brain** (the tiny model that handles background housekeeping).
30
+ - `chat` → **Use as chat model** (your main conversational model).
31
+
32
+
33
+
34
+ ## Current catalog
35
+
36
+ | Model | Type | Quant | Size | Min RAM | License | Source |
37
+ |---|---|---|---|---|---|---|
38
+ | Qwen2.5 0.5B Instruct | agent | Q4_K_M | ~491 MB | 1.5 GB | Apache-2.0 | [Qwen](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF) |
39
+ | Llama 3.2 3B Instruct | chat | Q4_K_M | ~2.0 GB | 6 GB | Llama 3.2 | [bartowski](https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-GGUF) |
40
+ | Qwen2.5 3B Instruct | chat | Q4_K_M | ~2.1 GB | 6 GB | Qwen | [Qwen](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct-GGUF) |
41
+
42
+
43
+ ## Manifest format
44
+
45
+ `manifest.json` is a single JSON file with a `components` array. Each component has:
46
+ `id`, `type` (`agent` | `chat` | …), `name`, `description`, `quant`, `minRamMb`, `url`
47
+ (a direct `…/resolve/main/<file>.gguf` link), `sizeBytes`, and `sha256`. Multi-file components
48
+ (e.g. a vision model + projector) use a `files` array instead of a single `url`.
49
+
50
+ ## Licenses
51
+
52
+ Each model retains its **original license** (Apache-2.0 for Qwen 0.5B; the Llama 3.2 and Qwen
53
+ community licenses for the 3B models). Weights are downloaded directly from the creators' repos;
54
+ this repo only distributes the catalog metadata. See each linked model card for full terms.
55
+