chore(gallery): add Clef GGUF variants

Add the 27B Clef decision model with Q4_K_M and Q8_0 builds and a
matching vision projector. Pin artifacts and document installation.

Assisted-by: Codex:gpt-6

docs/content/features/decisions.md

@@ -341,6 +341,26 @@ tokens. LocalAI records supplied usage once; explicit zero counts are distinct
 from missing usage. Missing counts are not estimated, and invalid negative counts
 are rejected rather than billed.
 
+### Clef text and image decisions
+
+Install Cloudflare's 27B Clef decision model with its Q8_0 vision projector:
+
+```sh
+local-ai models install clef --variant clef
+```
+
+This selects the Q4_K_M build. Use `--variant clef-q8` for Q8_0,
+or omit `--variant` to let the gallery choose a build for available memory.
+Both builds use Apache-2.0 weights from
+[ggml-org/Clef-GGUF](https://huggingface.co/ggml-org/Clef-GGUF), pinned to
+revision `63840a1a68cb7084c88610cffc328509356b04cb` with SHA-256 checksums.
+
+Send requests to `/v1/systemone` or `/v1/decisions` with the installed model's
+name. Clef returns typed decisions; it does not generate chat completions.
+Image requests use the bounded base64 image contract described above and require
+the matching projector included in both builds. The gallery sets an 8192-token
+context. Use a current stock `llama-cpp` backend with Clef decision support.
+
 ### Julia-1 CPU example
 
 Install the separate stock llama.cpp entry (existing vllm-cpp entries are unchanged):
@@ -470,7 +490,7 @@ 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. Clef image decisions also require the matching vision projector.
 
 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

@@ -65746,6 +65746,72 @@
     - filename: Kev-4B-Q4_K_M.gguf
       uri: https://huggingface.co/ggml-org/Kev-4B-GGUF/resolve/d924f2e2c3872da8b8aaf3eb4453b4126deceb79/Kev-4B-Q4_K_M.gguf
       sha256: 33ae6b18926502b2209a1bf7d3b61a350d65938441515c686bb87d226a19eff9
+- name: clef
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  variants:
+    - model: clef-q8
+  urls:
+    - https://huggingface.co/Cloudflare/clef
+    - https://huggingface.co/ggml-org/Clef-GGUF
+  license: apache-2.0
+  description: |
+    Clef is Cloudflare's 27B Qwen3.8-based model for typed decisions
+    from text and images. This Q4_K_M GGUF includes the Q8_0 vision projector
+    and uses native llama.cpp decisions through /v1/systemone and /v1/decisions.
+    It answers choice, score, and predicate questions without generating text.
+  tags:
+    - decisions
+    - systemone
+    - gguf
+    - cpu
+    - gpu
+  overrides:
+    backend: llama-cpp
+    known_usecases:
+      - decisions
+    context_size: 8192
+    mmproj: llama-cpp/mmproj/clef/mmproj-Clef-Q8_0.gguf
+    parameters:
+      model: llama-cpp/models/clef/Clef-Q4_K_M.gguf
+  files:
+    - filename: llama-cpp/models/clef/Clef-Q4_K_M.gguf
+      uri: https://huggingface.co/ggml-org/Clef-GGUF/resolve/63840a1a68cb7084c88610cffc328509356b04cb/Clef-Q4_K_M.gguf
+      sha256: bd9b8ae24c5752a2bb5e2733d5168340d5f04457e02b8bf0d1bc9285ec3d4f61
+    - filename: llama-cpp/mmproj/clef/mmproj-Clef-Q8_0.gguf
+      uri: https://huggingface.co/ggml-org/Clef-GGUF/resolve/63840a1a68cb7084c88610cffc328509356b04cb/mmproj-Clef-Q8_0.gguf
+      sha256: 0d901ea999ae122ba0afeb4a3f4c4ac9a90baa07f1d120df9c350082db1190a8
+- name: clef-q8
+  url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  urls:
+    - https://huggingface.co/Cloudflare/clef
+    - https://huggingface.co/ggml-org/Clef-GGUF
+  license: apache-2.0
+  description: |
+    Clef is Cloudflare's 27B Qwen3.8-based model for typed decisions
+    from text and images. This Q8_0 GGUF includes the Q8_0 vision projector
+    and uses native llama.cpp decisions through /v1/systemone and /v1/decisions.
+    It answers choice, score, and predicate questions without generating text.
+  tags:
+    - decisions
+    - systemone
+    - gguf
+    - cpu
+    - gpu
+  overrides:
+    backend: llama-cpp
+    known_usecases:
+      - decisions
+    context_size: 8192
+    mmproj: llama-cpp/mmproj/clef/mmproj-Clef-Q8_0.gguf
+    parameters:
+      model: llama-cpp/models/clef/Clef-Q8_0.gguf
+  files:
+    - filename: llama-cpp/models/clef/Clef-Q8_0.gguf
+      uri: https://huggingface.co/ggml-org/Clef-GGUF/resolve/63840a1a68cb7084c88610cffc328509356b04cb/Clef-Q8_0.gguf
+      sha256: 07c6410af7011e0e56873a3b0b3f4ad9e31fb176f0ca80d6fd1525fe6f036548
+    - filename: llama-cpp/mmproj/clef/mmproj-Clef-Q8_0.gguf
+      uri: https://huggingface.co/ggml-org/Clef-GGUF/resolve/63840a1a68cb7084c88610cffc328509356b04cb/mmproj-Clef-Q8_0.gguf
+      sha256: 0d901ea999ae122ba0afeb4a3f4c4ac9a90baa07f1d120df9c350082db1190a8
 - name: lev-llama-cpp
   urls:
     - https://huggingface.co/interfaze-ai/lev