@bendyline/gilde 0.1.29 → 0.1.30

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (11) hide show
  1. package/authoring/chat-models/qwen3.8-27b-q4.json +11 -7
  2. package/authoring/chat-models/qwen3.8-27b-q8.json +185 -0
  3. package/data/chat-models/index.json +1 -1
  4. package/data/chat-models/qw/qwen3.8-27b-q4/manifest.json +83 -10
  5. package/data/chat-models/qw/qwen3.8-27b-q4/versions/1.0.1/manifest.json +93 -0
  6. package/data/chat-models/qw/qwen3.8-27b-q8/manifest.json +279 -0
  7. package/data/chat-models/qw/qwen3.8-27b-q8/versions/1.0.0/manifest.json +113 -0
  8. package/data/craftbook-templates/index.json +1 -1
  9. package/data/craftbook-templates/pu/pull-request-review/versions/1.5.0/craftbook.json +188 -0
  10. package/data/craftbook-templates/pu/pull-request-review/versions/1.5.0/test.json +123 -0
  11. package/package.json +1 -1
package/authoring/chat-models/qwen3.8-27b-q4.json CHANGED
@@ -18,14 +18,14 @@
18
18
  "name": "Alibaba",
19
19
  "url": "https://huggingface.co/Qwen/Qwen3.8-27B"
20
20
  },
21
- "version": "1.0.0",
22
- "updatedAt": "2026-08-14T00:00:00Z",
21
+ "version": "1.0.1",
22
+ "updatedAt": "2026-08-15T00:00:00Z",
23
23
  "license": "Apache-2.0",
24
24
  "licenseClass": "open",
25
25
  "licenseShortName": "Apache 2.0",
26
26
  "recoScore": 20,
27
27
  "parameterSize": "27B",
28
- "approxSizeBytes": 17106773984,
28
+ "approxSizeBytes": 17106775008,
29
29
  "supportsTools": true,
30
30
  "contextWindow": 262144,
31
31
  "upstream": "https://huggingface.co/Qwen/Qwen3.8-27B",
@@ -64,7 +64,7 @@
64
64
  },
65
65
  "reasoning": {
66
66
  "enableThinking": true,
67
- "thinkingBudget": 8192,
67
+ "thinkingBudget": 4096,
68
68
  "templateKwargs": {
69
69
  "reasoning_effort": "xhigh"
70
70
  }
@@ -86,7 +86,7 @@
86
86
  },
87
87
  "reasoning": {
88
88
  "enableThinking": true,
89
- "thinkingBudget": 8192,
89
+ "thinkingBudget": 4096,
90
90
  "templateKwargs": {
91
91
  "reasoning_effort": "xhigh"
92
92
  }
@@ -104,7 +104,7 @@
104
104
  },
105
105
  "reasoning": {
106
106
  "enableThinking": true,
107
- "thinkingBudget": 8192,
107
+ "thinkingBudget": 4096,
108
108
  "templateKwargs": {
109
109
  "reasoning_effort": "xhigh"
110
110
  }
@@ -163,9 +163,13 @@
163
163
  "huggingfaceRepo": "unsloth/Qwen3.8-27B-GGUF",
164
164
  "filename": "Qwen3.8-27B-Q4_K_M.gguf",
165
165
  "quantization": "Q4_K_M",
166
- "residentBytes": 20528128781,
166
+ "residentBytes": 20528130010,
167
167
  "mmprojFilename": "mmproj-F16.gguf"
168
168
  },
169
+ "mlx": {
170
+ "huggingfaceRepo": "mlx-community/Qwen3.8-27B-4bit",
171
+ "quantization": "4bit"
172
+ },
169
173
  "releasePatch": {
170
174
  "tuning": {
171
175
  "engine": {
package/authoring/chat-models/qwen3.8-27b-q8.json ADDED
@@ -0,0 +1,185 @@
1
+ {
2
+ "id": "qwen3.8-27b-q8",
3
+ "name": "Qwen 3.8 (27B, Q8)",
4
+ "description": "Alibaba's Qwen 3.8 at 27B parameters: a dense, multimodal agentic model with native text, image, and video understanding, 262K context, flexible reasoning effort, MTP, and native tool calling. Q8_0 and 8-bit MLX preserve more fidelity than the Q4 build at roughly double the memory footprint; intended for machines with 40GB+ usable memory.",
5
+ "tags": [
6
+ "alibaba",
7
+ "agentic",
8
+ "coding",
9
+ "multimodal",
10
+ "vision",
11
+ "video",
12
+ "tools",
13
+ "reasoning",
14
+ "long-context"
15
+ ],
16
+ "category": "general",
17
+ "maintainer": {
18
+ "name": "Alibaba",
19
+ "url": "https://huggingface.co/Qwen/Qwen3.8-27B"
20
+ },
21
+ "version": "1.0.0",
22
+ "updatedAt": "2026-08-15T00:00:00Z",
23
+ "license": "Apache-2.0",
24
+ "licenseClass": "open",
25
+ "licenseShortName": "Apache 2.0",
26
+ "recoScore": 20,
27
+ "parameterSize": "27B",
28
+ "approxSizeBytes": 29047086048,
29
+ "supportsTools": true,
30
+ "contextWindow": 262144,
31
+ "upstream": "https://huggingface.co/Qwen/Qwen3.8-27B",
32
+ "style": {
33
+ "family": "qwen",
34
+ "reasoningFormat": "think",
35
+ "toolCallFormat": "function-call"
36
+ },
37
+ "behaviors": [
38
+ "reasoning.strip-think-tags",
39
+ "provider.merge-system-messages",
40
+ "mcp.compact-tool-schemas",
41
+ "provider.compact-write-transcript",
42
+ "fabrication.detect-past-tense-no-tools",
43
+ "turn.ollama-num-predict-bumped",
44
+ "turn.preamble-folding",
45
+ "turn.ramble-detection",
46
+ "tools.mlx-grammar",
47
+ "prompt.retrieval-first",
48
+ "prompt.workspace-gestalt",
49
+ "prompt.derive-by-execution"
50
+ ],
51
+ "tuning": {
52
+ "sampling": {
53
+ "temperature": 1,
54
+ "topP": 0.95,
55
+ "topK": 20,
56
+ "minP": 0,
57
+ "presencePenalty": 0,
58
+ "repetitionPenalty": 1,
59
+ "maxTokens": 12288
60
+ },
61
+ "samplingWhenThinking": {
62
+ "temperature": 1,
63
+ "topP": 0.95
64
+ },
65
+ "reasoning": {
66
+ "enableThinking": true,
67
+ "thinkingBudget": 4096,
68
+ "templateKwargs": {
69
+ "reasoning_effort": "xhigh"
70
+ }
71
+ },
72
+ "promptTags": {
73
+ "enableThinkingTag": "/think",
74
+ "disableThinkingTag": "/no_think"
75
+ },
76
+ "profiles": {
77
+ "thinking-general": {
78
+ "sampling": {
79
+ "temperature": 1,
80
+ "topP": 0.95,
81
+ "topK": 20,
82
+ "minP": 0,
83
+ "presencePenalty": 0,
84
+ "repetitionPenalty": 1,
85
+ "maxTokens": 12288
86
+ },
87
+ "reasoning": {
88
+ "enableThinking": true,
89
+ "thinkingBudget": 4096,
90
+ "templateKwargs": {
91
+ "reasoning_effort": "xhigh"
92
+ }
93
+ }
94
+ },
95
+ "thinking-coding": {
96
+ "sampling": {
97
+ "temperature": 1,
98
+ "topP": 0.95,
99
+ "topK": 20,
100
+ "minP": 0,
101
+ "presencePenalty": 0,
102
+ "repetitionPenalty": 1,
103
+ "maxTokens": 16384
104
+ },
105
+ "reasoning": {
106
+ "enableThinking": true,
107
+ "thinkingBudget": 4096,
108
+ "templateKwargs": {
109
+ "reasoning_effort": "xhigh"
110
+ }
111
+ }
112
+ },
113
+ "thinking-precise": {
114
+ "sampling": {
115
+ "temperature": 0.6,
116
+ "topP": 0.95,
117
+ "topK": 20,
118
+ "minP": 0,
119
+ "presencePenalty": 0,
120
+ "repetitionPenalty": 1,
121
+ "seed": 0,
122
+ "maxTokens": 6144
123
+ },
124
+ "reasoning": {
125
+ "enableThinking": true,
126
+ "thinkingBudget": 4096,
127
+ "templateKwargs": {
128
+ "reasoning_effort": "medium"
129
+ }
130
+ }
131
+ },
132
+ "instruct": {
133
+ "sampling": {
134
+ "temperature": 0.7,
135
+ "topP": 0.8,
136
+ "topK": 20,
137
+ "minP": 0,
138
+ "presencePenalty": 1.5,
139
+ "repetitionPenalty": 1,
140
+ "maxTokens": 4096
141
+ },
142
+ "reasoning": {
143
+ "enableThinking": false
144
+ }
145
+ },
146
+ "creative": {
147
+ "sampling": {
148
+ "temperature": 1,
149
+ "topP": 0.95,
150
+ "topK": 40,
151
+ "minP": 0,
152
+ "presencePenalty": 1.5,
153
+ "repetitionPenalty": 1,
154
+ "maxTokens": 8192
155
+ },
156
+ "reasoning": {
157
+ "enableThinking": false
158
+ }
159
+ }
160
+ }
161
+ },
162
+ "llamaCpp": {
163
+ "huggingfaceRepo": "unsloth/Qwen3.8-27B-GGUF",
164
+ "filename": "Qwen3.8-27B-Q8_0.gguf",
165
+ "quantization": "Q8_0",
166
+ "residentBytes": 34856503258,
167
+ "mmprojFilename": "mmproj-F16.gguf"
168
+ },
169
+ "mlx": {
170
+ "huggingfaceRepo": "mlx-community/Qwen3.8-27B-8bit",
171
+ "quantization": "8bit"
172
+ },
173
+ "releasePatch": {
174
+ "tuning": {
175
+ "engine": {
176
+ "llamaCpp": {
177
+ "spec": {
178
+ "mtp": true,
179
+ "nMax": 4
180
+ }
181
+ }
182
+ }
183
+ }
184
+ }
185
+ }