@bendyline/gilde 0.1.3 → 0.1.4

package/data/chat-models/de/deepseek-v4-flash-284b-q2/manifest.json

@@ -11,7 +11,7 @@
     "url": "https://huggingface.co/antirez/deepseek-v4-gguf"
   },
   "version": "1.0.0",
-  "updatedAt": "2026-06-02T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "MIT",
   "licenseClass": "open",
   "licenseShortName": "MIT",
@@ -30,12 +30,11 @@
     "prompt.private-reasoning-guidance",
     "prompt.tool-cookbook-condensed",
     "mcp.compact-tool-schemas",
-    "fabrication.detect-past-tense-no-tools",
-    "turn.preamble-folding"
+    "fabrication.detect-past-tense-no-tools"
   ],
   "tuning": {
     "sampling": {
-      "maxTokens": 8192
+      "maxTokens": 16384
     },
     "reasoning": {
       "enableThinking": true,

package/data/chat-models/de/deepseek-v4-flash-284b-q4/manifest.json

@@ -11,7 +11,7 @@
     "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash"
   },
   "version": "1.0.0",
-  "updatedAt": "2026-05-17T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "MIT",
   "licenseClass": "open",
   "licenseShortName": "MIT",
@@ -30,12 +30,11 @@
     "prompt.private-reasoning-guidance",
     "prompt.tool-cookbook-condensed",
     "mcp.compact-tool-schemas",
-    "fabrication.detect-past-tense-no-tools",
-    "turn.preamble-folding"
+    "fabrication.detect-past-tense-no-tools"
   ],
   "tuning": {
     "sampling": {
-      "maxTokens": 8192
+      "maxTokens": 16384
     },
     "reasoning": {
       "enableThinking": true,

package/data/chat-models/ge/gemma4-12b-q4/manifest.json

@@ -19,7 +19,7 @@
     "url": "https://huggingface.co/google/gemma-4-12B-it"
   },
   "version": "1.1.1",
-  "updatedAt": "2026-06-06T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "Apache-2.0",
   "licenseClass": "open",
   "licenseShortName": "Apache 2.0",
@@ -177,7 +177,12 @@
     "filename": "gemma-4-12B-it-qat-UD-Q4_K_XL.gguf",
     "sha256": "cc9ff072e0a8203429ed854e6662c17a6c2bc1e5dca5b475dd4736caaacbc165",
     "approxSizeBytes": 6716355328,
-    "quantization": "UD-Q4_K_XL"
+    "quantization": "UD-Q4_K_XL",
+    "mmproj": {
+      "filename": "mmproj-BF16.gguf",
+      "sha256": "dcb8103adad042b1bf99df767aaf34eb37c5a73a4a2f0417e4d7ba557e91664f",
+      "sizeBytes": 175115840
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-12B-it-qat-4bit",

package/data/chat-models/ge/gemma4-12b-q4/versions/1.1.0/manifest.json

@@ -9,7 +9,12 @@
     "filename": "gemma-4-12B-it-qat-UD-Q4_K_XL.gguf",
     "sha256": "cc9ff072e0a8203429ed854e6662c17a6c2bc1e5dca5b475dd4736caaacbc165",
     "approxSizeBytes": 6716355328,
-    "quantization": "UD-Q4_K_XL"
+    "quantization": "UD-Q4_K_XL",
+    "mmproj": {
+      "filename": "mmproj-BF16.gguf",
+      "sha256": "dcb8103adad042b1bf99df767aaf34eb37c5a73a4a2f0417e4d7ba557e91664f",
+      "sizeBytes": 175115840
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-12B-it-qat-4bit",
@@ -68,5 +73,6 @@
         "sizeBytes": 2747
       }
     ]
-  }
+  },
+  "updatedAt": "2026-07-25T00:00:00Z"
 }

package/data/chat-models/ge/gemma4-12b-q8/manifest.json

@@ -18,7 +18,7 @@
     "url": "https://huggingface.co/google/gemma-4-12B-it"
   },
   "version": "1.0.1",
-  "updatedAt": "2026-06-03T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "Apache-2.0",
   "licenseClass": "open",
   "licenseShortName": "Apache 2.0",
@@ -151,7 +151,12 @@
     "filename": "gemma-4-12B-it-Q8_0.gguf",
     "sha256": "047dae1d7894b9de8f08141e841544e007243290c02df8b39872991d1940c795",
     "approxSizeBytes": 12669645728,
-    "quantization": "Q8_0"
+    "quantization": "Q8_0",
+    "mmproj": {
+      "filename": "mmproj-gemma-4-12B-it-bf16.gguf",
+      "sha256": "675ad6e68101ca9413ec806855c452362f0213f2dfc5800996b086fdb8119842",
+      "sizeBytes": 175115584
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-12b-it-8bit",

package/data/chat-models/ge/gemma4-12b-q8/versions/1.0.0/manifest.json

@@ -9,7 +9,12 @@
     "filename": "gemma-4-12B-it-Q8_0.gguf",
     "sha256": "047dae1d7894b9de8f08141e841544e007243290c02df8b39872991d1940c795",
     "approxSizeBytes": 12669645728,
-    "quantization": "Q8_0"
+    "quantization": "Q8_0",
+    "mmproj": {
+      "filename": "mmproj-gemma-4-12B-it-bf16.gguf",
+      "sha256": "675ad6e68101ca9413ec806855c452362f0213f2dfc5800996b086fdb8119842",
+      "sizeBytes": 175115584
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-12b-it-8bit",
@@ -68,5 +73,6 @@
         "sizeBytes": 2719
       }
     ]
-  }
+  },
+  "updatedAt": "2026-07-25T00:00:00Z"
 }

package/data/chat-models/ge/gemma4-26b-q4/manifest.json

@@ -4,14 +4,22 @@
   "id": "gemma4-26b-q4",
   "name": "Gemma 4 (26B, MoE, Q4)",
   "description": "Google's Gemma 4 mixture-of-experts model \u2014 25.2B total parameters with 3.8B active per token. Multimodal with long-context agentic workflows. The full 26B has to fit in memory but only ~4B parameters fire per token, so on capable hardware it runs noticeably faster than the dense 31B at a similar memory cost \u2014 the recommended high-end on-device default.",
-  "tags": ["google", "multimodal", "vision", "tools", "mix of experts", "agentic", "qat"],
+  "tags": [
+    "google",
+    "multimodal",
+    "vision",
+    "tools",
+    "mix of experts",
+    "agentic",
+    "qat"
+  ],
   "category": "general",
   "maintainer": {
     "name": "Google",
     "url": "https://huggingface.co/google/gemma-4-26B-A4B-it"
   },
   "version": "1.2.0",
-  "updatedAt": "2026-06-06T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "Apache-2.0",
   "licenseClass": "open",
   "licenseShortName": "Apache 2.0",
@@ -134,7 +142,12 @@
     "filename": "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf",
     "sha256": "dcf179a91153e3a7ece792e48ef872180d9d6ef9b7677f0a0bd3e83cfe624d5e",
     "approxSizeBytes": 14249045120,
-    "quantization": "UD-Q4_K_XL"
+    "quantization": "UD-Q4_K_XL",
+    "mmproj": {
+      "filename": "mmproj-BF16.gguf",
+      "sha256": "7b06953ccdbe8cf363f47841a7afaacd2b1c2ff9a8d6b426fdec7521a6878744",
+      "sizeBytes": 1194828256
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-26B-A4B-it-qat-nvfp4",

package/data/chat-models/ge/gemma4-26b-q4/versions/1.2.0/manifest.json

@@ -12,7 +12,12 @@
     "filename": "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf",
     "sha256": "dcf179a91153e3a7ece792e48ef872180d9d6ef9b7677f0a0bd3e83cfe624d5e",
     "approxSizeBytes": 14249045120,
-    "quantization": "UD-Q4_K_XL"
+    "quantization": "UD-Q4_K_XL",
+    "mmproj": {
+      "filename": "mmproj-BF16.gguf",
+      "sha256": "7b06953ccdbe8cf363f47841a7afaacd2b1c2ff9a8d6b426fdec7521a6878744",
+      "sizeBytes": 1194828256
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-26B-A4B-it-qat-nvfp4",
@@ -71,5 +76,6 @@
         "sizeBytes": 2740
       }
     ]
-  }
+  },
+  "updatedAt": "2026-07-25T00:00:00Z"
 }

package/data/chat-models/ge/gemma4-31b-q4/manifest.json

@@ -4,14 +4,23 @@
   "id": "gemma4-31b-q4",
   "name": "Gemma 4 (31B)",
   "description": "Google's flagship Gemma 4 at 30.7B parameters. Multimodal (text + image + video), 256K context, configurable thinking mode, native function-calling, 140+ languages. Frontier-tier instruction-following on device \u2014 needs 32GB+ of usable memory.",
-  "tags": ["google", "multimodal", "vision", "video", "tools", "reasoning", "large", "qat"],
+  "tags": [
+    "google",
+    "multimodal",
+    "vision",
+    "video",
+    "tools",
+    "reasoning",
+    "large",
+    "qat"
+  ],
   "category": "general",
   "maintainer": {
     "name": "Google",
     "url": "https://huggingface.co/google/gemma-4-31B-it"
   },
   "version": "1.2.0",
-  "updatedAt": "2026-06-06T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "Apache-2.0",
   "licenseClass": "open",
   "licenseShortName": "Apache 2.0",
@@ -135,7 +144,12 @@
     "sha256": "9188a71055550f1e60b875d02b7abb63625ac11b4a6f148d6b22b3b28ba3d335",
     "approxSizeBytes": 17287668064,
     "quantization": "UD-Q4_K_XL",
-    "residentBytes": 21030000000
+    "residentBytes": 21030000000,
+    "mmproj": {
+      "filename": "mmproj-BF16.gguf",
+      "sha256": "d904b3579a9fbfbd50bc9bf40cb7384909edbd69fa9276db5ddc853e80f0edca",
+      "sizeBytes": 1200726496
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-31b-it-4bit",

package/data/chat-models/ge/gemma4-31b-q4/versions/1.2.0/manifest.json

@@ -13,7 +13,12 @@
     "sha256": "9188a71055550f1e60b875d02b7abb63625ac11b4a6f148d6b22b3b28ba3d335",
     "approxSizeBytes": 17287668064,
     "quantization": "UD-Q4_K_XL",
-    "residentBytes": 21030000000
+    "residentBytes": 21030000000,
+    "mmproj": {
+      "filename": "mmproj-BF16.gguf",
+      "sha256": "d904b3579a9fbfbd50bc9bf40cb7384909edbd69fa9276db5ddc853e80f0edca",
+      "sizeBytes": 1200726496
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-31b-it-4bit",
@@ -78,5 +83,6 @@
       }
     ],
     "residentBytes": 23977741909
-  }
+  },
+  "updatedAt": "2026-07-25T00:00:00Z"
 }

package/data/chat-models/ge/gemma4-e2b-q8/manifest.json

@@ -18,7 +18,7 @@
     "url": "https://huggingface.co/google/gemma-4-E2B-it"
   },
   "version": "1.1.1",
-  "updatedAt": "2026-04-26T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "Apache-2.0",
   "licenseClass": "open",
   "licenseShortName": "Apache 2.0",
@@ -65,7 +65,12 @@
     "sha256": "e049411c01fb7a81161768c52e38828970e55a64e22738957adcbe51d20f1c8e",
     "approxSizeBytes": 4967494592,
     "quantization": "Q8_0",
-    "residentBytes": 5960993510
+    "residentBytes": 5960993510,
+    "mmproj": {
+      "filename": "mmproj-gemma-4-E2B-it-bf16.gguf",
+      "sha256": "e42083b71a9e31e0f722171d551f6d92b101544001c4dde040306a8f2160fe8c",
+      "sizeBytes": 986833376
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-e2b-it-4bit",

package/data/chat-models/ge/gemma4-e2b-q8/versions/1.1.0/manifest.json

@@ -13,7 +13,12 @@
     "sha256": "e049411c01fb7a81161768c52e38828970e55a64e22738957adcbe51d20f1c8e",
     "approxSizeBytes": 4967494592,
     "quantization": "Q8_0",
-    "residentBytes": 5960993510
+    "residentBytes": 5960993510,
+    "mmproj": {
+      "filename": "mmproj-gemma-4-E2B-it-bf16.gguf",
+      "sha256": "e42083b71a9e31e0f722171d551f6d92b101544001c4dde040306a8f2160fe8c",
+      "sizeBytes": 986833376
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-e2b-it-4bit",
@@ -63,5 +68,6 @@
       }
     ],
     "residentBytes": 4697585580
-  }
+  },
+  "updatedAt": "2026-07-25T00:00:00Z"
 }

package/data/chat-models/ge/gemma4-e4b-q8/manifest.json

@@ -17,7 +17,7 @@
     "url": "https://huggingface.co/google/gemma-4-E4B-it"
   },
   "version": "1.1.1",
-  "updatedAt": "2026-06-07T00:00:00Z",
+  "updatedAt": "2026-07-25T00:00:00Z",
   "license": "Apache-2.0",
   "licenseClass": "open",
   "licenseShortName": "Apache 2.0",
@@ -60,7 +60,12 @@
     "sha256": "fb8f0c032de00b18c710824af3c7e5777c71e5fb60b13f13575f0a9e92ddecd0",
     "approxSizeBytes": 8031240160,
     "quantization": "Q8_0",
-    "residentBytes": 9637488192
+    "residentBytes": 9637488192,
+    "mmproj": {
+      "filename": "mmproj-gemma-4-E4B-it-bf16.gguf",
+      "sha256": "4c199e460410ba219a8c63930a7121154e1c70cdf66044858f767966332e5a54",
+      "sizeBytes": 991551968
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-E4B-it-qat-4bit",
@@ -114,7 +119,7 @@
     ],
     "approxSizeBytes": 6830817013,
     "quantization": "4bit-qat",
-    "disabledReason": "mlx-vlm cannot load Gemma-4 E4B (elastic MatFormer + audio tower) — fails on every published build across mlx-vlm 0.6.1-0.6.4."
+    "disabledReason": "mlx-vlm cannot load Gemma-4 E4B (elastic MatFormer + audio tower) \u2014 fails on every published build across mlx-vlm 0.6.1-0.6.4."
   },
   "tuning": {
     "sampling": {

package/data/chat-models/ge/gemma4-e4b-q8/versions/1.1.0/manifest.json

@@ -13,7 +13,12 @@
     "sha256": "fb8f0c032de00b18c710824af3c7e5777c71e5fb60b13f13575f0a9e92ddecd0",
     "approxSizeBytes": 8031240160,
     "quantization": "Q8_0",
-    "residentBytes": 9637488192
+    "residentBytes": 9637488192,
+    "mmproj": {
+      "filename": "mmproj-gemma-4-E4B-it-bf16.gguf",
+      "sha256": "4c199e460410ba219a8c63930a7121154e1c70cdf66044858f767966332e5a54",
+      "sizeBytes": 991551968
+    }
   },
   "mlx": {
     "huggingfaceRepo": "mlx-community/gemma-4-E4B-it-qat-4bit",
@@ -67,6 +72,7 @@
     ],
     "approxSizeBytes": 6830817013,
     "quantization": "4bit-qat",
-    "disabledReason": "mlx-vlm cannot load Gemma-4 E4B (elastic MatFormer + audio tower) — fails on every published build across mlx-vlm 0.6.1-0.6.4."
-  }
+    "disabledReason": "mlx-vlm cannot load Gemma-4 E4B (elastic MatFormer + audio tower) \u2014 fails on every published build across mlx-vlm 0.6.1-0.6.4."
+  },
+  "updatedAt": "2026-07-25T00:00:00Z"
 }

package/data/chat-models/gl/glm-5.2-754b-q2/manifest.json

@@ -0,0 +1,66 @@
+{
+  "schemaVersion": 1,
+  "kind": "chat-model",
+  "id": "glm-5.2-754b-q2",
+  "name": "GLM 5.2 (IQ2_XXS)",
+  "description": "antirez's routed 2-bit community GGUF of Z.ai's GLM 5.2 — a 754B-parameter Mixture-of-Experts (256 experts, 8 active per token, ~40B active) with a 1M-token context and DeepSeek-style sparse attention. The routed experts are crushed to IQ2_XXS (block 78 in Q2_K) while attention, the shared expert, embeddings, and the output head stay at Q8-class precision, landing the file at ~197 GB. Only ~20 GB of that is non-routed weight that must stay resident: DwarfStar streams the remaining ~177 GB of experts from SSD, so it runs on a 128 GB machine — but it is a very large download and a slow first turn. Built for antirez's `ds4` engine; these quants do not load on stock llama.cpp.",
+  "tags": [
+    "glm",
+    "reasoning",
+    "moe",
+    "tools",
+    "2-bit",
+    "iq2",
+    "long-context",
+    "large"
+  ],
+  "category": "reasoning",
+  "maintainer": {
+    "name": "antirez",
+    "url": "https://huggingface.co/antirez/glm-5.2-gguf"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-07-25T00:00:00Z",
+  "license": "MIT",
+  "parameterSize": "754B",
+  "approxSizeBytes": 211075856448,
+  "supportsTools": true,
+  "contextWindow": 1048576,
+  "upstream": "https://huggingface.co/zai-org/GLM-5.2",
+  "style": {
+    "family": "glm",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.private-reasoning-guidance",
+    "prompt.tool-cookbook-condensed",
+    "mcp.compact-tool-schemas",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.preamble-folding"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 1,
+      "topP": 0.95,
+      "maxTokens": 16384
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    }
+  },
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.2-gguf",
+    "revision": "2638b3b878f5c6cc3ae7334b8dbea1275025f52e",
+    "filename": "GLM-5.2-UD-IQ2_XXS_RoutedIQ2XXS_blk78Q2K.gguf",
+    "sha256": "a49de64c5020432bdae23de36a423a9660a5621bc0db8d12b66bd8814b07fea0",
+    "approxSizeBytes": 211075856448,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 61363217408,
+    "cacheExpertsBytes": 34359738368,
+    "ssdStreaming": true,
+    "maxLaunchCtx": 65536
+  }
+}

package/data/chat-models/gl/glm-5.2-754b-q2/versions/1.0.0/manifest.json

@@ -0,0 +1,18 @@
+{
+  "schemaVersion": 1,
+  "version": "1.0.0",
+  "releasedAt": "2026-07-25T00:00:00Z",
+  "approxSizeBytes": 211075856448,
+  "ds4": {
+    "huggingfaceRepo": "antirez/glm-5.2-gguf",
+    "revision": "2638b3b878f5c6cc3ae7334b8dbea1275025f52e",
+    "filename": "GLM-5.2-UD-IQ2_XXS_RoutedIQ2XXS_blk78Q2K.gguf",
+    "sha256": "a49de64c5020432bdae23de36a423a9660a5621bc0db8d12b66bd8814b07fea0",
+    "approxSizeBytes": 211075856448,
+    "quantization": "IQ2_XXS",
+    "residentBytes": 61363217408,
+    "cacheExpertsBytes": 34359738368,
+    "ssdStreaming": true,
+    "maxLaunchCtx": 65536
+  }
+}