chore(gallery): add LiquidAI d1-3B variants

Add Q4_K_M and Q8_0 builds with the Q8 vision projector for native
llama.cpp decisions. Pin artifacts and document installation.

Assisted-by: Codex:gpt-6

docs/content/features/decisions.md

@@ -450,6 +450,38 @@ probes also check serialized escaping before allocation. The separate 64 KiB tex
 request limit is unchanged. Anthropic conversion preserves typed content blocks
 through both native selection and fallback, including ordered text and images.
 
+### LiquidAI d1-3B
+
+The `d1-3b-q4` gallery entry installs LiquidAI d1-3B Q4_K_M with its Q8_0
+vision projector. It offers `d1-3b-q8` as a higher-precision variant.
+Both use native `llama-cpp` decisions for text, JSON, and images.
+The weights use the [Liquid AI LFM Open License v1.0](https://huggingface.co/LiquidAI/d1-3B-GGUF/blob/bb1e436ea78eb96a3f1acb6da865f70c2fbeb563/LICENSE).
+
+Install a specific build:
+
+```bash
+local-ai models install d1-3b-q4 --variant d1-3b-q4
+```
+
+Use `d1-3b-q4` as the model in the `/v1/decisions` example above, or send
+a SystemOne request:
+
+```bash
+curl http://localhost:8080/v1/systemone \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "model": "d1-3b-q4",
+    "state": "I was charged twice. Please refund the duplicate charge.",
+    "questions": {
+      "refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"}
+    }
+  }'
+```
+
+Both builds use an 8192-token context. Image requests follow the bounds above
+and require the installed projector. These entries declare the `decisions`
+usecase and do not appear as chat models.
+
 ### OpenJev image decisions
 
 The `openjev-llama-cpp` gallery entry installs OpenJev Q4_K_M and its pinned
@@ -470,7 +502,7 @@ fit; context overflow remains an error. Size the context for the actual workload
 Native image decisions require both a decision format that accepts images and a
 loaded projector that supports **vision input**. A missing projector, an
 audio-only projector, or a text-only decision model does not silently fall back
-to a text decision. Other decision gallery entries remain text-only.
+to a text decision. Gallery entries without a vision projector remain text-only.
 
 Direct native RPC callers receive the same image count, encoded/decoded byte,
 dimension and aggregate pixel bounds as public callers. Validation precedes

gallery/index.yaml

@@ -65555,6 +65555,74 @@
     - filename: lev-Q4_K_M.gguf
       uri: https://huggingface.co/ggml-org/lev-GGUF/resolve/3e9286a79ae857b4e1de051c92fab6dc574581ce/lev-Q4_K_M.gguf
       sha256: 3f61b27c00a098dbc79ed099cca3985cbe0b63d17ac7fedebf2dca1d84f7f3a8
+- name: d1-3b-q4
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/LiquidAI/d1-3B
+    - https://huggingface.co/LiquidAI/d1-3B-GGUF
+  description: |
+    LiquidAI d1-3B is a 3B decision model based on LFM2.5-VL.
+    Answers typed choice, score, and noul questions over text, JSON, or images
+    through the native llama.cpp Decisions API. Includes Q4_K_M weights and
+    the Q8_0 vision projector. Licensed under the Liquid AI LFM Open License v1.0.
+  license: other
+  tags:
+    - decisions
+    - systemone
+    - gguf
+    - cpu
+    - gpu
+  last_checked: "2026-10-10"
+  variants:
+    - model: d1-3b-q8
+  overrides:
+    backend: llama-cpp
+    known_usecases:
+      - decisions
+    context_size: 8192
+    mmproj: mmproj-d1-3B-Q8_0.gguf
+    parameters:
+      model: d1-3B-Q4_K_M.gguf
+  files:
+    - filename: d1-3B-Q4_K_M.gguf
+      uri: https://huggingface.co/LiquidAI/d1-3B-GGUF/resolve/bb1e436ea78eb96a3f1acb6da865f70c2fbeb563/d1-3B-Q4_K_M.gguf
+      sha256: 16aff27ea2eefdc32b9897f43854a5d3170c1dc8dccb9c756905af30a4e22402
+    - filename: mmproj-d1-3B-Q8_0.gguf
+      uri: https://huggingface.co/LiquidAI/d1-3B-GGUF/resolve/bb1e436ea78eb96a3f1acb6da865f70c2fbeb563/mmproj-d1-3B-Q8_0.gguf
+      sha256: 2505920ce464b7c54acd8b7b74a931c0c671f06d0c8833760b2fbceeb253e92a
+- name: d1-3b-q8
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/LiquidAI/d1-3B
+    - https://huggingface.co/LiquidAI/d1-3B-GGUF
+  description: |
+    LiquidAI d1-3B is a 3B decision model based on LFM2.5-VL.
+    Answers typed choice, score, and noul questions over text, JSON, or images
+    through the native llama.cpp Decisions API. Includes Q8_0 weights and
+    the Q8_0 vision projector. Licensed under the Liquid AI LFM Open License v1.0.
+  license: other
+  tags:
+    - decisions
+    - systemone
+    - gguf
+    - cpu
+    - gpu
+  last_checked: "2026-10-10"
+  overrides:
+    backend: llama-cpp
+    known_usecases:
+      - decisions
+    context_size: 8192
+    mmproj: mmproj-d1-3B-Q8_0.gguf
+    parameters:
+      model: d1-3B-Q8_0.gguf
+  files:
+    - filename: d1-3B-Q8_0.gguf
+      uri: https://huggingface.co/LiquidAI/d1-3B-GGUF/resolve/bb1e436ea78eb96a3f1acb6da865f70c2fbeb563/d1-3B-Q8_0.gguf
+      sha256: 2f0942d5a5f64cf69c3356d2be439b71644b1f0eb3a580912a5a0179eeaba77d
+    - filename: mmproj-d1-3B-Q8_0.gguf
+      uri: https://huggingface.co/LiquidAI/d1-3B-GGUF/resolve/bb1e436ea78eb96a3f1acb6da865f70c2fbeb563/mmproj-d1-3B-Q8_0.gguf
+      sha256: 2505920ce464b7c54acd8b7b74a931c0c671f06d0c8833760b2fbceeb253e92a
 - name: openjev-llama-cpp
   urls:
     - https://huggingface.co/openjev/openjev