fix(gallery): normalize inconsistent tag casing/plurals across gallery models (#9574)

- embeddings → embedding (6 models): aligns with the WebUI filter button
  defined in core/http/views/models.html ({ term: 'embedding', ... }), so
  models like nomic-embed-text-v1.5 now appear under the Embedding filter
- TTS → tts (5 models), ASR → asr (2 models): lowercase, per existing
  convention used by 161+ models
- CPU/Cpu → cpu (17 models), GPU → gpu (17 models): lowercase, per existing
  convention used by 666+ models
- dedupe duplicate tag entries on 3 models that already had repeated tags
  (gpt-oss-20b had gguf x2; arcee-ai/AFM-4.5B had gpu x2; one Qwen model
  had default x2)

Closes #9247

Author: Anai-Guo

gallery/index.yaml

@@ -743,7 +743,6 @@
     - https://huggingface.co/mradermacher/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-heretic-i1-GGUF
   tags:
     - default
-    - default
   overrides:
     parameters:
       model: llama-cpp/models/Qwen3.-27B-Claude-4.6-Opus-Reasoning-Distilled-heretic.i1-Q4_K_M.gguf
@@ -1915,7 +1914,7 @@
     Qwen3-TTS is a high-quality text-to-speech model supporting custom voice, voice design, and voice cloning.
   tags:
     - text-to-speech
-    - TTS
+    - tts
   license: apache-2.0
   icon: https://cdn-avatars.huggingface.co/v1/production/uploads/620760a26e3b7210c2ff1943/-s1gyJfvbE1RgO5iBeNOi.png
   name: "qwen3-tts-1.7b-custom-voice"
@@ -1947,7 +1946,7 @@
     Fish Speech S2-Pro is a high-quality text-to-speech model supporting voice cloning via reference audio. Uses a two-stage pipeline: text to semantic tokens (LLaMA-based) then semantic to audio (DAC decoder).
   tags:
     - text-to-speech
-    - TTS
+    - tts
     - voice-cloning
   license: apache-2.0
   icon: https://huggingface.co/fishaudio/s2-pro/resolve/main/overview.png
@@ -1966,7 +1965,7 @@
     Qwen3-ASR is an automatic speech recognition model supporting multiple languages and batch inference.
   tags:
     - speech-recognition
-    - ASR
+    - asr
   license: apache-2.0
   icon: https://cdn-avatars.huggingface.co/v1/production/uploads/620760a26e3b7210c2ff1943/-s1gyJfvbE1RgO5iBeNOi.png
   name: "qwen3-asr-1.7b"
@@ -2575,7 +2574,7 @@
   license: mit
   tags:
     - text-to-speech
-    - TTS
+    - tts
   name: "vibevoice"
   urls:
     - https://github.com/microsoft/VibeVoice
@@ -2609,7 +2608,7 @@
   license: mit
   tags:
     - text-to-speech
-    - TTS
+    - tts
   name: "pocket-tts"
   urls:
     - https://github.com/kyutai-labs/pocket-tts
@@ -3057,8 +3056,8 @@
   license: apache-2.0
   tags:
     - gguf
-    - GPU
-    - CPU
+    - gpu
+    - cpu
     - text-to-text
     - jamba
     - mamba
@@ -3082,8 +3081,8 @@
   icon: https://cdn-avatars.huggingface.co/v1/production/uploads/639bcaa2445b133a4e942436/CEW-OjXkRkDNmTxSu8Egh.png
   tags:
     - gguf
-    - GPU
-    - CPU
+    - gpu
+    - cpu
     - text-to-text
   urls:
     - https://huggingface.co/ibm-granite/granite-4.0-h-small
@@ -3145,8 +3144,8 @@
   license: apache-2.0
   tags:
     - gguf
-    - GPU
-    - CPU
+    - gpu
+    - cpu
     - text-to-text
   icon: https://cdn-avatars.huggingface.co/v1/production/uploads/64f187a2cc1c03340ac30498/TYYUxK8xD1AxExFMWqbZD.png
   urls:
@@ -3169,8 +3168,8 @@
   license: mit
   tags:
     - gguf
-    - GPU
-    - CPU
+    - gpu
+    - cpu
     - text-to-text
   icon: https://cdn-uploads.huggingface.co/production/uploads/634262af8d8089ebaefd410e/9Bnn2AnIjfQFWBGkhDNmI.png
   name: "aurore-reveil_koto-small-7b-it"
@@ -3197,8 +3196,8 @@
   tags:
     - multimodal
     - gguf
-    - GPU
-    - Cpu
+    - gpu
+    - cpu
     - image-to-text
     - text-to-text
   description: |
@@ -3819,7 +3818,6 @@
     - gguf
     - gpu
     - cpu
-    - gguf
     - openai
   icon: https://raw.githubusercontent.com/openai/gpt-oss/main/docs/gpt-oss-20b.svg
   urls:
@@ -4005,7 +4003,6 @@
   tags:
     - gguf
     - gpu
-    - gpu
     - text-generation
   description: |
     AFM-4.5B is a 4.5 billion parameter instruction-tuned model developed by Arcee.ai, designed for enterprise-grade performance across diverse deployment environments from cloud to edge. The base model was trained on a dataset of 8 trillion tokens, comprising 6.5 trillion tokens of general pretraining data followed by 1.5 trillion tokens of midtraining data with enhanced focus on mathematical reasoning and code generation. Following pretraining, the model underwent supervised fine-tuning on high-quality instruction datasets. The instruction-tuned model was further refined through reinforcement learning on verifiable rewards as well as for human preference. We use a modified version of TorchTitan for pretraining, Axolotl for supervised fine-tuning, and a modified version of Verifiers for reinforcement learning.
@@ -9112,7 +9109,7 @@
   description: |
     Granite-Embedding-107M-Multilingual is a 107M parameter dense biencoder embedding model from the Granite Embeddings suite that can be used to generate high quality text embeddings. This model produces embedding vectors of size 384 and is trained using a combination of open source relevance-pair datasets with permissive, enterprise-friendly license, and IBM collected and generated datasets. This model is developed using contrastive finetuning, knowledge distillation and model merging for improved performance.
   tags:
-    - embeddings
+    - embedding
   overrides:
     backend: llama-cpp
     embeddings: true
@@ -9130,7 +9127,7 @@
   description: |
     Granite-Embedding-125m-English is a 125M parameter dense biencoder embedding model from the Granite Embeddings suite that can be used to generate high quality text embeddings. This model produces embedding vectors of size 768. Compared to most other open-source models, this model was only trained using open-source relevance-pair datasets with permissive, enterprise-friendly license, plus IBM collected and generated datasets. While maintaining competitive scores on academic benchmarks such as BEIR, this model also performs well on many enterprise use cases. This model is developed using retrieval oriented pretraining, contrastive finetuning and knowledge distillation.
   tags:
-    - embeddings
+    - embedding
   overrides:
     embeddings: true
     parameters:
@@ -9147,7 +9144,7 @@
   description: |
     EmbeddingGemma 300M is a lightweight, high-quality embedding model from Google, based on the Gemma architecture. It produces 1024-dimensional embeddings optimized for retrieval and semantic similarity tasks. This GGUF version uses QAT (Quantization-Aware Training) Q8_0 quantization for efficient inference.
   tags:
-    - embeddings
+    - embedding
   overrides:
     backend: llama-cpp
     embeddings: true
@@ -15923,7 +15920,7 @@
   tags:
     - gpu
     - cpu
-    - embeddings
+    - embedding
     - python
   name: "all-MiniLM-L6-v2"
   url: "github:mudler/LocalAI/gallery/sentencetransformers.yaml@master"
@@ -16776,7 +16773,7 @@
   description: |
     llama3.2 embeddings model. Using as drop-in replacement for bert-embeddings
   tags:
-    - embeddings
+    - embedding
   overrides:
     embeddings: true
     parameters:
@@ -18499,7 +18496,7 @@
   description: |
     Resizable Production Embeddings with Matryoshka Representation Learning
   tags:
-    - embeddings
+    - embedding
   overrides:
     embeddings: true
     parameters: