@bendyline/gilde 0.1.18 → 0.1.19

package/authoring/chat-models/ornith-35b-q8.json

@@ -0,0 +1,145 @@
+{
+  "id": "ornith-35b-q8",
+  "name": "Ornith 1.0 (35B, Q8)",
+  "description": "DeepReinforce's Ornith 1.0 at 35B total parameters (Qwen 3.5 MoE architecture), an agentic-coding model trained with a self-improving reinforcement-learning framework — state-of-the-art on SWE-Bench and Terminal-Bench. Native <think> reasoning, OpenAI-style tool calling, 256K context. Q8_0 quantization for higher fidelity than the Q4 build, at roughly double the memory footprint (~37GB). MIT licensed.",
+  "tags": [
+    "deepreinforce",
+    "coding",
+    "agentic",
+    "tools",
+    "reasoning",
+    "long-context",
+    "moe"
+  ],
+  "category": "coding",
+  "maintainer": {
+    "name": "DeepReinforce",
+    "url": "https://huggingface.co/deepreinforce-ai"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-07-01T00:00:00Z",
+  "license": "MIT",
+  "licenseClass": "open",
+  "licenseShortName": "MIT",
+  "parameterSize": "35B",
+  "approxSizeBytes": 37000000000,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "kvBytesPerTokenF16": 20480,
+  "kvFixedBytesF16": 64389120,
+  "upstream": "https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "turn.ramble-detection"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.6,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "maxTokens": 8192
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.6,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 6144
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.4,
+          "topP": 0.9,
+          "topK": 20,
+          "minP": 0,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.8,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1.1,
+          "topP": 0.95,
+          "topK": 40,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "deepreinforce-ai/Ornith-1.0-35B-GGUF",
+    "filename": "ornith-1.0-35b-Q8_0.gguf",
+    "quantization": "Q8_0",
+    "residentBytes": 44283766656
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Ornith-1.0-35B-8bit",
+    "quantization": "8bit",
+    "residentBytes": 49072879453
+  }
+}

package/authoring/chat-models/ornith-9b-q4.json

@@ -0,0 +1,146 @@
+{
+  "id": "ornith-9b-q4",
+  "name": "Ornith 1.0 (9B)",
+  "description": "DeepReinforce's Ornith 1.0 at 9B parameters (dense, built on Qwen 3.5 + Gemma 4), an agentic-coding model trained with a self-improving reinforcement-learning framework. Native <think> reasoning, OpenAI-style tool calling, 256K context. The small-tier pick for tool-heavy work on a 12GB+ inference budget. Q4_K_M quantization. MIT licensed.",
+  "tags": [
+    "deepreinforce",
+    "coding",
+    "agentic",
+    "tools",
+    "reasoning",
+    "long-context"
+  ],
+  "category": "coding",
+  "maintainer": {
+    "name": "DeepReinforce",
+    "url": "https://huggingface.co/deepreinforce-ai"
+  },
+  "version": "1.0.0",
+  "updatedAt": "2026-07-01T00:00:00Z",
+  "license": "MIT",
+  "licenseClass": "open",
+  "licenseShortName": "MIT",
+  "parameterSize": "9B",
+  "approxSizeBytes": 5600000000,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "kvBytesPerTokenF16": 32768,
+  "kvFixedBytesF16": 51511296,
+  "upstream": "https://huggingface.co/deepreinforce-ai/Ornith-1.0-9B",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "mcp.relax-required-fields",
+    "mcp.default-missing-fields",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "provider.merge-system-messages"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 0.6,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "maxTokens": 16384
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.6,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 2048
+    },
+    "promptTags": {
+      "enableThinkingTag": "/think",
+      "disableThinkingTag": "/no_think"
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 2048
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.4,
+          "topP": 0.9,
+          "topK": 20,
+          "minP": 0,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 2048
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.8,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1.1,
+          "topP": 0.95,
+          "topK": 40,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "deepreinforce-ai/Ornith-1.0-9B-GGUF",
+    "filename": "ornith-1.0-9b-Q4_K_M.gguf",
+    "quantization": "Q4_K_M",
+    "residentBytes": 6754930445
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Ornith-1.0-9B-4bit",
+    "quantization": "4bit",
+    "residentBytes": 7770193739
+  }
+}

package/authoring/chat-models/qwen3.5-122b-a10b-q4.json

@@ -0,0 +1,172 @@
+{
+  "id": "qwen3.5-122b-a10b-q4",
+  "name": "Qwen 3.5 (122B-A10B, Q4)",
+  "description": "Alibaba's Qwen 3.5 flagship open-weight model: 122B total parameters with 10B active per token, sparse MoE/Gated DeltaNet architecture, multimodal text + image + video support, native tool calling, and 262K native context extensible to roughly 1M tokens. Workstation-tier on-device option via Q4 GGUF or MLX 4-bit; expect around 90GB of usable memory before large KV-cache workloads.",
+  "tags": [
+    "alibaba",
+    "multimodal",
+    "vision",
+    "video",
+    "tools",
+    "long-context",
+    "moe",
+    "reasoning",
+    "large"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Alibaba",
+    "url": "https://huggingface.co/Qwen/Qwen3.5-122B-A10B"
+  },
+  "version": "1.0.1",
+  "updatedAt": "2026-07-28T00:00:00Z",
+  "license": "Apache-2.0",
+  "licenseClass": "open",
+  "licenseShortName": "Apache 2.0",
+  "parameterSize": "122B",
+  "approxSizeBytes": 78260403200,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://huggingface.co/Qwen/Qwen3.5-122B-A10B",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "prompt.meester-build-prelude",
+    "fabrication.detect-past-tense-no-tools",
+    "fabrication.detect-claim-without-tool",
+    "mcp.relax-required-fields",
+    "mcp.default-missing-fields",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding",
+    "turn.ramble-detection",
+    "tools.mlx-grammar",
+    "provider.merge-system-messages",
+    "validate.inline-js-parses",
+    "prompt.prefer-writefile-edits",
+    "prompt.derive-by-execution"
+  ],
+  "tuning": {
+    "sampling": {
+      "temperature": 1,
+      "topP": 0.95,
+      "topK": 20,
+      "minP": 0,
+      "presencePenalty": 1.5,
+      "repetitionPenalty": 1,
+      "maxTokens": 16384
+    },
+    "samplingWhenThinking": {
+      "temperature": 0.6,
+      "topP": 0.95
+    },
+    "reasoning": {
+      "enableThinking": true,
+      "thinkingBudget": 4096
+    },
+    "profiles": {
+      "thinking-general": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 12288
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "thinking-coding": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "maxTokens": 16384
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 6144
+        }
+      },
+      "thinking-precise": {
+        "sampling": {
+          "temperature": 0.6,
+          "topP": 0.95,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 0,
+          "repetitionPenalty": 1,
+          "seed": 0,
+          "maxTokens": 6144
+        },
+        "reasoning": {
+          "enableThinking": true,
+          "thinkingBudget": 4096
+        }
+      },
+      "instruct": {
+        "sampling": {
+          "temperature": 0.7,
+          "topP": 0.8,
+          "topK": 20,
+          "minP": 0,
+          "presencePenalty": 1.5,
+          "repetitionPenalty": 1,
+          "maxTokens": 4096
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      },
+      "creative": {
+        "sampling": {
+          "temperature": 1,
+          "topP": 1,
+          "topK": 40,
+          "minP": 0,
+          "presencePenalty": 2,
+          "repetitionPenalty": 1,
+          "maxTokens": 8192
+        },
+        "reasoning": {
+          "enableThinking": false
+        }
+      }
+    }
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Qwen3.5-122B-A10B-MTP-GGUF",
+    "shardsDir": "UD-Q4_K_M",
+    "quantization": "UD-Q4_K_M",
+    "residentBytes": 93912483840,
+    "mmprojFilename": "mmproj-BF16.gguf"
+  },
+  "releasePatch": {
+    "tuning": {
+      "engine": {
+        "llamaCpp": {
+          "spec": {
+            "mtp": true,
+            "nMax": 4
+          }
+        }
+      }
+    }
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Qwen3.5-122B-A10B-4bit",
+    "quantization": "4bit",
+    "residentBytes": 90506802715
+  }
+}

package/authoring/chat-models/qwen3.5-2b-q4.json

@@ -0,0 +1,70 @@
+{
+  "id": "qwen3.5-2b-q4",
+  "name": "Qwen 3.5 (2B)",
+  "description": "Alibaba's Qwen 3.5 at 2B parameters. Multimodal (text + images) with native tool calling and a 256K context window. A very compact choice for constrained devices.",
+  "tags": [
+    "alibaba",
+    "multimodal",
+    "vision",
+    "tools",
+    "small"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Alibaba",
+    "url": "https://ollama.com/library/qwen3.5"
+  },
+  "version": "1.1.2",
+  "updatedAt": "2026-07-28T00:00:00Z",
+  "license": "Apache-2.0",
+  "parameterSize": "2B",
+  "approxSizeBytes": 1329851808,
+  "supportsTools": true,
+  "contextWindow": 256000,
+  "upstream": "https://ollama.com/library/qwen3.5",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-full",
+    "fabrication.detect-past-tense-no-tools",
+    "mcp.relax-required-fields",
+    "mcp.default-missing-fields",
+    {
+      "id": "turn.continuation-budget",
+      "config": {
+        "count": 4
+      }
+    },
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding"
+  ],
+  "ollama": {
+    "tag": "qwen3.5:2b"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Qwen3.5-2B-MTP-GGUF",
+    "filename": "Qwen3.5-2B-Q4_K_M.gguf",
+    "quantization": "Q4_K_M",
+    "residentBytes": 1595822170
+  },
+  "releasePatch": {
+    "tuning": {
+      "engine": {
+        "llamaCpp": {
+          "spec": {
+            "mtp": true,
+            "nMax": 4
+          }
+        }
+      }
+    }
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Qwen3.5-2B-4bit",
+    "quantization": "4bit"
+  }
+}

package/authoring/chat-models/qwen3.5-4b-q4.json

@@ -0,0 +1,61 @@
+{
+  "id": "qwen3.5-4b-q4",
+  "name": "Qwen 3.5 (4B)",
+  "description": "Alibaba's Qwen 3.5 at 4B parameters. Multimodal (text + images) with native tool calling and a 256K context window. A balanced compact default for modern laptops.",
+  "tags": [
+    "alibaba",
+    "multimodal",
+    "vision",
+    "tools"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Alibaba",
+    "url": "https://ollama.com/library/qwen3.5"
+  },
+  "version": "1.1.2",
+  "updatedAt": "2026-07-28T00:00:00Z",
+  "license": "Apache-2.0",
+  "parameterSize": "4B",
+  "approxSizeBytes": 2834975040,
+  "supportsTools": true,
+  "contextWindow": 256000,
+  "upstream": "https://ollama.com/library/qwen3.5",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding"
+  ],
+  "ollama": {
+    "tag": "qwen3.5:4b"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Qwen3.5-4B-MTP-GGUF",
+    "filename": "Qwen3.5-4B-Q4_K_M.gguf",
+    "quantization": "Q4_K_M",
+    "residentBytes": 3401970048
+  },
+  "releasePatch": {
+    "tuning": {
+      "engine": {
+        "llamaCpp": {
+          "spec": {
+            "mtp": true,
+            "nMax": 4
+          }
+        }
+      }
+    }
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Qwen3.5-4B-4bit",
+    "quantization": "4bit"
+  }
+}

package/authoring/chat-models/qwen3.5-9b-q4.json

@@ -0,0 +1,61 @@
+{
+  "id": "qwen3.5-9b-q4",
+  "name": "Qwen 3.5 (9B)",
+  "description": "Alibaba's Qwen 3.5 at 9B parameters. Multimodal (text + images), native tool calling, 256K context. Strong reasoning + tool use for devices with a 12GB+ inference budget — the small-tier sweet spot for tool-heavy work.",
+  "tags": [
+    "alibaba",
+    "multimodal",
+    "vision",
+    "tools"
+  ],
+  "category": "general",
+  "maintainer": {
+    "name": "Alibaba",
+    "url": "https://ollama.com/library/qwen3.5"
+  },
+  "version": "1.1.2",
+  "updatedAt": "2026-07-28T00:00:00Z",
+  "license": "Apache-2.0",
+  "parameterSize": "9B",
+  "approxSizeBytes": 5868826976,
+  "supportsTools": true,
+  "contextWindow": 262144,
+  "upstream": "https://ollama.com/library/qwen3.5",
+  "style": {
+    "family": "qwen",
+    "reasoningFormat": "think",
+    "toolCallFormat": "function-call"
+  },
+  "behaviors": [
+    "reasoning.strip-think-tags",
+    "prompt.tool-cookbook-condensed",
+    "fabrication.detect-past-tense-no-tools",
+    "turn.ollama-num-predict-bumped",
+    "turn.preamble-folding"
+  ],
+  "ollama": {
+    "tag": "qwen3.5:9b"
+  },
+  "llamaCpp": {
+    "huggingfaceRepo": "unsloth/Qwen3.5-9B-MTP-GGUF",
+    "filename": "Qwen3.5-9B-Q4_K_M.gguf",
+    "quantization": "Q4_K_M",
+    "residentBytes": 7042592371
+  },
+  "releasePatch": {
+    "tuning": {
+      "engine": {
+        "llamaCpp": {
+          "spec": {
+            "mtp": true,
+            "nMax": 4
+          }
+        }
+      }
+    }
+  },
+  "mlx": {
+    "huggingfaceRepo": "mlx-community/Qwen3.5-9B-4bit",
+    "quantization": "4bit"
+  }
+}