feat(gallery): add Ternary Bonsai 2 9B variants

Add three GGUF packings with the Q8_0 vision projector and document
installation. Pin asset revisions and verify hashes against HF metadata
and download headers.

Assisted-by: Codex:gpt-6

docs/content/features/model-gallery.md

@@ -44,6 +44,26 @@ Both views store the view, search, filter, and selection in the URL. Installing
 from Explore does not move you away from the catalog; the entry updates in place
 when the operation finishes.
 
+### Ternary Bonsai 2 9B
+
+[Ternary Bonsai 2 9B](https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf)
+supports chat and image understanding through the Bonsai backend.
+The PQ2_0, PTQ1_0, and Q2_0 builds contain the same ternary weights in
+different storage layouts. Each includes the Q8_0 vision projector and
+uses a 32,768-token default context.
+
+Install with automatic variant selection:
+
+```bash
+local-ai models install ternary-bonsai-2-9b
+```
+
+Use `--variant ternary-bonsai-2-9b` to select PQ2_0,
+`--variant ternary-bonsai-2-9b-ptq1` for PTQ1_0, or
+`--variant ternary-bonsai-2-9b-q2` for Q2_0.
+Automatic selection can choose the larger Q2_0 packing when it fits;
+a larger packing does not imply higher precision for these ternary weights.
+
 ### MiniCPM-V 4.7 1B
 
 [MiniCPM-V 4.7 1B](https://huggingface.co/openbmb/MiniCPM-V-4.7-1B)

gallery/index.yaml

@@ -3,6 +3,150 @@
 # written by scripts/models/cmd/modelcatalog and stays only so releases before
 # the cutover keep working. CI fails when it differs from the generated copy:
 #   go run ./scripts/models/cmd/modelcatalog --library gallery/models/index.yaml --out gallery/index.yaml
+- url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  name: ternary-bonsai-2-9b
+  description: |
+    Ternary Bonsai 2 9B is a vision-language model derived from Qwen3.5-9B
+    with ternary transformer weights. This PQ2_0 build uses the Bonsai
+    backend and includes the Q8_0 vision projector. The three standard
+    packings contain the same weights in different storage layouts.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf
+    - https://huggingface.co/Qwen/Qwen3.5-9B
+  tags:
+    - llm
+    - gguf
+    - reasoning
+    - vision
+    - multimodal
+  files:
+    - filename: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PQ2_0.gguf
+      sha256: 7a89ab48f5038c704480f60781f9b8d4035351f5f6beb7d11fad80001fde23f9
+      uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-PQ2_0.gguf
+    - filename: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+      sha256: e0cb6f6988cea323ef798cd77d019bbb819af75582f034eaab0e7232314953db
+      uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+  overrides:
+    backend: bonsai
+    context_size: 32768
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+      - vision
+    mmproj: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+    options:
+      - use_jinja:true
+    parameters:
+      min_p: 0.0
+      model: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PQ2_0.gguf
+      presence_penalty: 1.0
+      repeat_penalty: 1.0
+      temperature: 1.0
+      top_k: 20
+      top_p: 0.95
+    template:
+      use_tokenizer_template: true
+  variants:
+    - model: ternary-bonsai-2-9b-ptq1
+    - model: ternary-bonsai-2-9b-q2
+- url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  name: ternary-bonsai-2-9b-ptq1
+  description: |
+    Ternary Bonsai 2 9B is a vision-language model derived from Qwen3.5-9B
+    with ternary transformer weights. This PTQ1_0 build uses the Bonsai
+    backend and includes the Q8_0 vision projector. The three standard
+    packings contain the same weights in different storage layouts.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf
+    - https://huggingface.co/Qwen/Qwen3.5-9B
+  tags:
+    - llm
+    - gguf
+    - reasoning
+    - vision
+    - multimodal
+  files:
+    - filename: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PTQ1_0.gguf
+      sha256: 3561df7ac7d26a71ffa991ac2db740b10841c3a350aa2676ccd3ff5d428b4e8b
+      uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-PTQ1_0.gguf
+    - filename: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+      sha256: e0cb6f6988cea323ef798cd77d019bbb819af75582f034eaab0e7232314953db
+      uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+  overrides:
+    backend: bonsai
+    context_size: 32768
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+      - vision
+    mmproj: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+    options:
+      - use_jinja:true
+    parameters:
+      min_p: 0.0
+      model: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PTQ1_0.gguf
+      presence_penalty: 1.0
+      repeat_penalty: 1.0
+      temperature: 1.0
+      top_k: 20
+      top_p: 0.95
+    template:
+      use_tokenizer_template: true
+- url: github:mudler/LocalAI/gallery/virtual.yaml@master
+  name: ternary-bonsai-2-9b-q2
+  description: |
+    Ternary Bonsai 2 9B is a vision-language model derived from Qwen3.5-9B
+    with ternary transformer weights. This Q2_0 build uses the Bonsai
+    backend and includes the Q8_0 vision projector. The three standard
+    packings contain the same weights in different storage layouts.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf
+    - https://huggingface.co/Qwen/Qwen3.5-9B
+  tags:
+    - llm
+    - gguf
+    - reasoning
+    - vision
+    - multimodal
+  files:
+    - filename: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-Q2_0.gguf
+      sha256: 1f884aa56aacd2bbbb2e4b7abaead2b1102b74a9c75a23fc33dab916813a384d
+      uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-Q2_0.gguf
+    - filename: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+      sha256: e0cb6f6988cea323ef798cd77d019bbb819af75582f034eaab0e7232314953db
+      uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+  overrides:
+    backend: bonsai
+    context_size: 32768
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+      - vision
+    mmproj: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+    options:
+      - use_jinja:true
+    parameters:
+      min_p: 0.0
+      model: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-Q2_0.gguf
+      presence_penalty: 1.0
+      repeat_penalty: 1.0
+      temperature: 1.0
+      top_k: 20
+      top_p: 0.95
+    template:
+      use_tokenizer_template: true
 - url: github:mudler/LocalAI/gallery/virtual.yaml@master
   name: adavar-3b-q4
   description: |

gallery/models/index.yaml

@@ -4,6 +4,174 @@
 # changes and never edit it by hand. CI fails when it is stale. A build from
 # master reads it as the default model source; a tagged build reads the gallery
 # until the release cutover.
+- kind: model
+  id: ternary-bonsai-2-9b
+  name: ternary-bonsai-2-9b
+  version: 0.0.0-gallery.275b12980f1b
+  description: |
+    Ternary Bonsai 2 9B is a vision-language model derived from Qwen3.5-9B
+    with ternary transformer weights. This PQ2_0 build uses the Bonsai
+    backend and includes the Q8_0 vision projector. The three standard
+    packings contain the same weights in different storage layouts.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf
+    - https://huggingface.co/Qwen/Qwen3.5-9B
+  tags:
+    - llm
+    - gguf
+    - reasoning
+    - vision
+    - multimodal
+  channel: gallery
+  policy:
+    requireSignature: false
+  schemaVersion: 1
+  model:
+    url: github:mudler/LocalAI/gallery/virtual.yaml@master
+    files:
+      - filename: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PQ2_0.gguf
+        sha256: 7a89ab48f5038c704480f60781f9b8d4035351f5f6beb7d11fad80001fde23f9
+        uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-PQ2_0.gguf
+      - filename: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+        sha256: e0cb6f6988cea323ef798cd77d019bbb819af75582f034eaab0e7232314953db
+        uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+    overrides:
+      backend: bonsai
+      context_size: 32768
+      function:
+        automatic_tool_parsing_fallback: true
+        grammar:
+          disable: true
+      known_usecases:
+        - chat
+        - vision
+      mmproj: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+      options:
+        - use_jinja:true
+      parameters:
+        min_p: 0.0
+        model: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PQ2_0.gguf
+        presence_penalty: 1.0
+        repeat_penalty: 1.0
+        temperature: 1.0
+        top_k: 20
+        top_p: 0.95
+      template:
+        use_tokenizer_template: true
+    variants:
+      - model: ternary-bonsai-2-9b-ptq1
+      - model: ternary-bonsai-2-9b-q2
+- kind: model
+  id: ternary-bonsai-2-9b-ptq1
+  name: ternary-bonsai-2-9b-ptq1
+  version: 0.0.0-gallery.fd060a8d2e1e
+  description: |
+    Ternary Bonsai 2 9B is a vision-language model derived from Qwen3.5-9B
+    with ternary transformer weights. This PTQ1_0 build uses the Bonsai
+    backend and includes the Q8_0 vision projector. The three standard
+    packings contain the same weights in different storage layouts.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf
+    - https://huggingface.co/Qwen/Qwen3.5-9B
+  tags:
+    - llm
+    - gguf
+    - reasoning
+    - vision
+    - multimodal
+  channel: gallery
+  policy:
+    requireSignature: false
+  schemaVersion: 1
+  model:
+    url: github:mudler/LocalAI/gallery/virtual.yaml@master
+    files:
+      - filename: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PTQ1_0.gguf
+        sha256: 3561df7ac7d26a71ffa991ac2db740b10841c3a350aa2676ccd3ff5d428b4e8b
+        uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-PTQ1_0.gguf
+      - filename: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+        sha256: e0cb6f6988cea323ef798cd77d019bbb819af75582f034eaab0e7232314953db
+        uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+    overrides:
+      backend: bonsai
+      context_size: 32768
+      function:
+        automatic_tool_parsing_fallback: true
+        grammar:
+          disable: true
+      known_usecases:
+        - chat
+        - vision
+      mmproj: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+      options:
+        - use_jinja:true
+      parameters:
+        min_p: 0.0
+        model: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-PTQ1_0.gguf
+        presence_penalty: 1.0
+        repeat_penalty: 1.0
+        temperature: 1.0
+        top_k: 20
+        top_p: 0.95
+      template:
+        use_tokenizer_template: true
+- kind: model
+  id: ternary-bonsai-2-9b-q2
+  name: ternary-bonsai-2-9b-q2
+  version: 0.0.0-gallery.ba7c9dff50f8
+  description: |
+    Ternary Bonsai 2 9B is a vision-language model derived from Qwen3.5-9B
+    with ternary transformer weights. This Q2_0 build uses the Bonsai
+    backend and includes the Q8_0 vision projector. The three standard
+    packings contain the same weights in different storage layouts.
+  license: apache-2.0
+  urls:
+    - https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf
+    - https://huggingface.co/Qwen/Qwen3.5-9B
+  tags:
+    - llm
+    - gguf
+    - reasoning
+    - vision
+    - multimodal
+  channel: gallery
+  policy:
+    requireSignature: false
+  schemaVersion: 1
+  model:
+    url: github:mudler/LocalAI/gallery/virtual.yaml@master
+    files:
+      - filename: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-Q2_0.gguf
+        sha256: 1f884aa56aacd2bbbb2e4b7abaead2b1102b74a9c75a23fc33dab916813a384d
+        uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-Q2_0.gguf
+      - filename: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+        sha256: e0cb6f6988cea323ef798cd77d019bbb819af75582f034eaab0e7232314953db
+        uri: https://huggingface.co/prism-ml/Ternary-Bonsai-2-9B-gguf/resolve/5b2abb02362efbed46bbe204014b02837778b2c2/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+    overrides:
+      backend: bonsai
+      context_size: 32768
+      function:
+        automatic_tool_parsing_fallback: true
+        grammar:
+          disable: true
+      known_usecases:
+        - chat
+        - vision
+      mmproj: bonsai/mmproj/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-mmproj-Q8_0.gguf
+      options:
+        - use_jinja:true
+      parameters:
+        min_p: 0.0
+        model: bonsai/models/ternary-bonsai-2-9b/Ternary-Bonsai-2-9B-Q2_0.gguf
+        presence_penalty: 1.0
+        repeat_penalty: 1.0
+        temperature: 1.0
+        top_k: 20
+        top_p: 0.95
+      template:
+        use_tokenizer_template: true
 - kind: model
   id: adavar-3b-q4
   name: adavar-3b-q4