agentic-dataset-builder 0.1.1 → 0.2.0
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.
- package/README.md +29 -91
- package/dist/cli.d.ts +2 -0
- package/dist/cli.js +149 -0
- package/dist/labeling.d.ts +12 -0
- package/dist/labeling.js +22 -0
- package/dist/parquet.d.ts +4 -0
- package/dist/parquet.js +115 -0
- package/dist/platform/paths.d.ts +4 -0
- package/dist/platform/paths.js +62 -0
- package/dist/schemas/qwen35.d.ts +338 -0
- package/dist/schemas/qwen35.js +139 -0
- package/dist/sources/claude.d.ts +2 -0
- package/dist/sources/claude.js +64 -0
- package/dist/sources/codex.d.ts +2 -0
- package/dist/sources/codex.js +261 -0
- package/dist/sources/pi.d.ts +2 -0
- package/dist/sources/pi.js +276 -0
- package/dist/utils/common.d.ts +7 -0
- package/dist/utils/common.js +46 -0
- package/dist/utils/jsonl.d.ts +2 -0
- package/dist/utils/jsonl.js +17 -0
- package/package.json +24 -12
- package/agentic_dataset/__init__.py +0 -1
- package/agentic_dataset/build_agentic_dataset.py +0 -368
- package/agentic_dataset/export_codex_session_to_qwen35.py +0 -466
- package/agentic_dataset/export_pi_session.py +0 -701
- package/agentic_dataset/export_pi_session_to_qwen35.py +0 -742
- package/agentic_dataset/export_qwen35_training.py +0 -1559
- package/agentic_dataset/label_qwen35_agentic.py +0 -156
- package/agentic_dataset/platform_paths.py +0 -85
- package/agentic_dataset/qwen35_training_record.py +0 -179
- package/bin/agentic-dataset-builder.js +0 -77
- package/requirements.txt +0 -2
- package/run.py +0 -8
|
@@ -0,0 +1,338 @@
|
|
|
1
|
+
import { z } from 'zod';
|
|
2
|
+
export declare const Qwen35TextBlockSchema: z.ZodObject<{
|
|
3
|
+
type: z.ZodLiteral<"text">;
|
|
4
|
+
text: z.ZodString;
|
|
5
|
+
}, z.core.$strip>;
|
|
6
|
+
export declare const Qwen35ImageBlockSchema: z.ZodObject<{
|
|
7
|
+
type: z.ZodLiteral<"image">;
|
|
8
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
9
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
10
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
11
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
12
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
13
|
+
}, z.core.$strip>;
|
|
14
|
+
export declare const Qwen35VideoBlockSchema: z.ZodObject<{
|
|
15
|
+
type: z.ZodLiteral<"video">;
|
|
16
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
17
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
18
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
19
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
20
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
21
|
+
}, z.core.$strip>;
|
|
22
|
+
export declare const Qwen35ContentSchema: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
23
|
+
type: z.ZodLiteral<"text">;
|
|
24
|
+
text: z.ZodString;
|
|
25
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
26
|
+
type: z.ZodLiteral<"image">;
|
|
27
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
28
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
29
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
30
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
31
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
32
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
33
|
+
type: z.ZodLiteral<"video">;
|
|
34
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
35
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
36
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
37
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
38
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
39
|
+
}, z.core.$strip>]>>]>;
|
|
40
|
+
export declare const Qwen35ToolCallSchema: z.ZodObject<{
|
|
41
|
+
type: z.ZodDefault<z.ZodLiteral<"function">>;
|
|
42
|
+
id: z.ZodOptional<z.ZodString>;
|
|
43
|
+
function: z.ZodObject<{
|
|
44
|
+
name: z.ZodString;
|
|
45
|
+
arguments: z.ZodDefault<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
46
|
+
}, z.core.$strip>;
|
|
47
|
+
}, z.core.$strip>;
|
|
48
|
+
export declare const Qwen35ToolSpecSchema: z.ZodObject<{
|
|
49
|
+
name: z.ZodString;
|
|
50
|
+
description: z.ZodOptional<z.ZodString>;
|
|
51
|
+
parameters: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
52
|
+
}, z.core.$strip>;
|
|
53
|
+
export declare const Qwen35MessageSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
|
|
54
|
+
role: z.ZodLiteral<"system">;
|
|
55
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
56
|
+
type: z.ZodLiteral<"text">;
|
|
57
|
+
text: z.ZodString;
|
|
58
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
59
|
+
type: z.ZodLiteral<"image">;
|
|
60
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
61
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
62
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
63
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
64
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
65
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
66
|
+
type: z.ZodLiteral<"video">;
|
|
67
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
68
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
69
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
70
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
71
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
72
|
+
}, z.core.$strip>]>>]>;
|
|
73
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
74
|
+
role: z.ZodLiteral<"user">;
|
|
75
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
76
|
+
type: z.ZodLiteral<"text">;
|
|
77
|
+
text: z.ZodString;
|
|
78
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
79
|
+
type: z.ZodLiteral<"image">;
|
|
80
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
81
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
82
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
83
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
84
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
85
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
86
|
+
type: z.ZodLiteral<"video">;
|
|
87
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
88
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
89
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
90
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
91
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
92
|
+
}, z.core.$strip>]>>]>;
|
|
93
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
94
|
+
role: z.ZodLiteral<"assistant">;
|
|
95
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
96
|
+
type: z.ZodLiteral<"text">;
|
|
97
|
+
text: z.ZodString;
|
|
98
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
99
|
+
type: z.ZodLiteral<"image">;
|
|
100
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
101
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
102
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
103
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
104
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
105
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
106
|
+
type: z.ZodLiteral<"video">;
|
|
107
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
108
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
109
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
110
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
111
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
112
|
+
}, z.core.$strip>]>>]>;
|
|
113
|
+
reasoning_content: z.ZodOptional<z.ZodString>;
|
|
114
|
+
tool_calls: z.ZodOptional<z.ZodArray<z.ZodObject<{
|
|
115
|
+
type: z.ZodDefault<z.ZodLiteral<"function">>;
|
|
116
|
+
id: z.ZodOptional<z.ZodString>;
|
|
117
|
+
function: z.ZodObject<{
|
|
118
|
+
name: z.ZodString;
|
|
119
|
+
arguments: z.ZodDefault<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
120
|
+
}, z.core.$strip>;
|
|
121
|
+
}, z.core.$strip>>>;
|
|
122
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
123
|
+
role: z.ZodLiteral<"tool">;
|
|
124
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
125
|
+
type: z.ZodLiteral<"text">;
|
|
126
|
+
text: z.ZodString;
|
|
127
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
128
|
+
type: z.ZodLiteral<"image">;
|
|
129
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
130
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
131
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
132
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
133
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
134
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
135
|
+
type: z.ZodLiteral<"video">;
|
|
136
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
137
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
138
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
139
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
140
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
141
|
+
}, z.core.$strip>]>>]>;
|
|
142
|
+
tool_call_id: z.ZodOptional<z.ZodString>;
|
|
143
|
+
name: z.ZodOptional<z.ZodString>;
|
|
144
|
+
}, z.core.$strip>], "role">;
|
|
145
|
+
export declare const Qwen35MetaSchema: z.ZodObject<{
|
|
146
|
+
endpoint: z.ZodString;
|
|
147
|
+
status: z.ZodNumber;
|
|
148
|
+
ts: z.ZodString;
|
|
149
|
+
key: z.ZodOptional<z.ZodString>;
|
|
150
|
+
source: z.ZodOptional<z.ZodString>;
|
|
151
|
+
requested_model: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
152
|
+
actual_model: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
153
|
+
stream: z.ZodOptional<z.ZodBoolean>;
|
|
154
|
+
thinking_level: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
155
|
+
reasoning_summary_mode: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnknown>, z.ZodRecord<z.ZodString, z.ZodUnknown>]>>;
|
|
156
|
+
thinking_type: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
157
|
+
thinking_budget_tokens: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
|
|
158
|
+
max_output_tokens: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
|
|
159
|
+
tool_spec_count: z.ZodOptional<z.ZodNumber>;
|
|
160
|
+
tool_choice: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnknown>, z.ZodRecord<z.ZodString, z.ZodUnknown>]>>;
|
|
161
|
+
request_contains_non_text_content: z.ZodDefault<z.ZodBoolean>;
|
|
162
|
+
request_image_block_count: z.ZodDefault<z.ZodNumber>;
|
|
163
|
+
request_video_block_count: z.ZodDefault<z.ZodNumber>;
|
|
164
|
+
request_tool_call_block_count: z.ZodDefault<z.ZodNumber>;
|
|
165
|
+
request_tool_result_block_count: z.ZodDefault<z.ZodNumber>;
|
|
166
|
+
request_thinking_block_count: z.ZodDefault<z.ZodNumber>;
|
|
167
|
+
response_contains_non_text_content: z.ZodDefault<z.ZodBoolean>;
|
|
168
|
+
response_image_block_count: z.ZodDefault<z.ZodNumber>;
|
|
169
|
+
response_video_block_count: z.ZodDefault<z.ZodNumber>;
|
|
170
|
+
response_tool_call_block_count: z.ZodDefault<z.ZodNumber>;
|
|
171
|
+
response_tool_result_block_count: z.ZodDefault<z.ZodNumber>;
|
|
172
|
+
response_thinking_block_count: z.ZodDefault<z.ZodNumber>;
|
|
173
|
+
request_truncated: z.ZodDefault<z.ZodBoolean>;
|
|
174
|
+
response_truncated: z.ZodDefault<z.ZodBoolean>;
|
|
175
|
+
lossy_source: z.ZodDefault<z.ZodBoolean>;
|
|
176
|
+
lossy_reasons: z.ZodDefault<z.ZodArray<z.ZodString>>;
|
|
177
|
+
dataset_label: z.ZodOptional<z.ZodString>;
|
|
178
|
+
dataset_source_system: z.ZodOptional<z.ZodString>;
|
|
179
|
+
dataset_source_bucket: z.ZodOptional<z.ZodString>;
|
|
180
|
+
dataset_source_file: z.ZodOptional<z.ZodString>;
|
|
181
|
+
dataset_has_reasoning: z.ZodOptional<z.ZodBoolean>;
|
|
182
|
+
dataset_reasoning_chars: z.ZodOptional<z.ZodNumber>;
|
|
183
|
+
}, z.core.$strip>;
|
|
184
|
+
export declare const Qwen35RecordSchema: z.ZodObject<{
|
|
185
|
+
id: z.ZodString;
|
|
186
|
+
request_id: z.ZodOptional<z.ZodString>;
|
|
187
|
+
messages: z.ZodArray<z.ZodDiscriminatedUnion<[z.ZodObject<{
|
|
188
|
+
role: z.ZodLiteral<"system">;
|
|
189
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
190
|
+
type: z.ZodLiteral<"text">;
|
|
191
|
+
text: z.ZodString;
|
|
192
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
193
|
+
type: z.ZodLiteral<"image">;
|
|
194
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
195
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
196
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
197
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
198
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
199
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
200
|
+
type: z.ZodLiteral<"video">;
|
|
201
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
202
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
203
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
204
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
205
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
206
|
+
}, z.core.$strip>]>>]>;
|
|
207
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
208
|
+
role: z.ZodLiteral<"user">;
|
|
209
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
210
|
+
type: z.ZodLiteral<"text">;
|
|
211
|
+
text: z.ZodString;
|
|
212
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
213
|
+
type: z.ZodLiteral<"image">;
|
|
214
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
215
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
216
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
217
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
218
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
219
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
220
|
+
type: z.ZodLiteral<"video">;
|
|
221
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
222
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
223
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
224
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
225
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
226
|
+
}, z.core.$strip>]>>]>;
|
|
227
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
228
|
+
role: z.ZodLiteral<"assistant">;
|
|
229
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
230
|
+
type: z.ZodLiteral<"text">;
|
|
231
|
+
text: z.ZodString;
|
|
232
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
233
|
+
type: z.ZodLiteral<"image">;
|
|
234
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
235
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
236
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
237
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
238
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
239
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
240
|
+
type: z.ZodLiteral<"video">;
|
|
241
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
242
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
243
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
244
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
245
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
246
|
+
}, z.core.$strip>]>>]>;
|
|
247
|
+
reasoning_content: z.ZodOptional<z.ZodString>;
|
|
248
|
+
tool_calls: z.ZodOptional<z.ZodArray<z.ZodObject<{
|
|
249
|
+
type: z.ZodDefault<z.ZodLiteral<"function">>;
|
|
250
|
+
id: z.ZodOptional<z.ZodString>;
|
|
251
|
+
function: z.ZodObject<{
|
|
252
|
+
name: z.ZodString;
|
|
253
|
+
arguments: z.ZodDefault<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
254
|
+
}, z.core.$strip>;
|
|
255
|
+
}, z.core.$strip>>>;
|
|
256
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
257
|
+
role: z.ZodLiteral<"tool">;
|
|
258
|
+
content: z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnion<readonly [z.ZodObject<{
|
|
259
|
+
type: z.ZodLiteral<"text">;
|
|
260
|
+
text: z.ZodString;
|
|
261
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
262
|
+
type: z.ZodLiteral<"image">;
|
|
263
|
+
image_url: z.ZodOptional<z.ZodString>;
|
|
264
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
265
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
266
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
267
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
268
|
+
}, z.core.$strip>, z.ZodObject<{
|
|
269
|
+
type: z.ZodLiteral<"video">;
|
|
270
|
+
video_url: z.ZodOptional<z.ZodString>;
|
|
271
|
+
placeholder: z.ZodOptional<z.ZodBoolean>;
|
|
272
|
+
placeholder_token: z.ZodOptional<z.ZodString>;
|
|
273
|
+
source_kind: z.ZodOptional<z.ZodString>;
|
|
274
|
+
metadata: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
275
|
+
}, z.core.$strip>]>>]>;
|
|
276
|
+
tool_call_id: z.ZodOptional<z.ZodString>;
|
|
277
|
+
name: z.ZodOptional<z.ZodString>;
|
|
278
|
+
}, z.core.$strip>], "role">>;
|
|
279
|
+
tools: z.ZodDefault<z.ZodArray<z.ZodObject<{
|
|
280
|
+
name: z.ZodString;
|
|
281
|
+
description: z.ZodOptional<z.ZodString>;
|
|
282
|
+
parameters: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
|
|
283
|
+
}, z.core.$strip>>>;
|
|
284
|
+
meta: z.ZodObject<{
|
|
285
|
+
endpoint: z.ZodString;
|
|
286
|
+
status: z.ZodNumber;
|
|
287
|
+
ts: z.ZodString;
|
|
288
|
+
key: z.ZodOptional<z.ZodString>;
|
|
289
|
+
source: z.ZodOptional<z.ZodString>;
|
|
290
|
+
requested_model: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
291
|
+
actual_model: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
292
|
+
stream: z.ZodOptional<z.ZodBoolean>;
|
|
293
|
+
thinking_level: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
294
|
+
reasoning_summary_mode: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnknown>, z.ZodRecord<z.ZodString, z.ZodUnknown>]>>;
|
|
295
|
+
thinking_type: z.ZodOptional<z.ZodNullable<z.ZodString>>;
|
|
296
|
+
thinking_budget_tokens: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
|
|
297
|
+
max_output_tokens: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
|
|
298
|
+
tool_spec_count: z.ZodOptional<z.ZodNumber>;
|
|
299
|
+
tool_choice: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodArray<z.ZodUnknown>, z.ZodRecord<z.ZodString, z.ZodUnknown>]>>;
|
|
300
|
+
request_contains_non_text_content: z.ZodDefault<z.ZodBoolean>;
|
|
301
|
+
request_image_block_count: z.ZodDefault<z.ZodNumber>;
|
|
302
|
+
request_video_block_count: z.ZodDefault<z.ZodNumber>;
|
|
303
|
+
request_tool_call_block_count: z.ZodDefault<z.ZodNumber>;
|
|
304
|
+
request_tool_result_block_count: z.ZodDefault<z.ZodNumber>;
|
|
305
|
+
request_thinking_block_count: z.ZodDefault<z.ZodNumber>;
|
|
306
|
+
response_contains_non_text_content: z.ZodDefault<z.ZodBoolean>;
|
|
307
|
+
response_image_block_count: z.ZodDefault<z.ZodNumber>;
|
|
308
|
+
response_video_block_count: z.ZodDefault<z.ZodNumber>;
|
|
309
|
+
response_tool_call_block_count: z.ZodDefault<z.ZodNumber>;
|
|
310
|
+
response_tool_result_block_count: z.ZodDefault<z.ZodNumber>;
|
|
311
|
+
response_thinking_block_count: z.ZodDefault<z.ZodNumber>;
|
|
312
|
+
request_truncated: z.ZodDefault<z.ZodBoolean>;
|
|
313
|
+
response_truncated: z.ZodDefault<z.ZodBoolean>;
|
|
314
|
+
lossy_source: z.ZodDefault<z.ZodBoolean>;
|
|
315
|
+
lossy_reasons: z.ZodDefault<z.ZodArray<z.ZodString>>;
|
|
316
|
+
dataset_label: z.ZodOptional<z.ZodString>;
|
|
317
|
+
dataset_source_system: z.ZodOptional<z.ZodString>;
|
|
318
|
+
dataset_source_bucket: z.ZodOptional<z.ZodString>;
|
|
319
|
+
dataset_source_file: z.ZodOptional<z.ZodString>;
|
|
320
|
+
dataset_has_reasoning: z.ZodOptional<z.ZodBoolean>;
|
|
321
|
+
dataset_reasoning_chars: z.ZodOptional<z.ZodNumber>;
|
|
322
|
+
}, z.core.$strip>;
|
|
323
|
+
label: z.ZodOptional<z.ZodString>;
|
|
324
|
+
source_system: z.ZodOptional<z.ZodString>;
|
|
325
|
+
source_bucket: z.ZodOptional<z.ZodString>;
|
|
326
|
+
source_file: z.ZodOptional<z.ZodString>;
|
|
327
|
+
agentic_label: z.ZodOptional<z.ZodObject<{
|
|
328
|
+
label: z.ZodString;
|
|
329
|
+
tool_call_count: z.ZodOptional<z.ZodNumber>;
|
|
330
|
+
tool_message_count: z.ZodOptional<z.ZodNumber>;
|
|
331
|
+
dialogue_rounds_est: z.ZodOptional<z.ZodNumber>;
|
|
332
|
+
reasoning_chars: z.ZodOptional<z.ZodNumber>;
|
|
333
|
+
has_reasoning: z.ZodOptional<z.ZodBoolean>;
|
|
334
|
+
lossy_source: z.ZodOptional<z.ZodBoolean>;
|
|
335
|
+
lossy_reasons: z.ZodOptional<z.ZodArray<z.ZodString>>;
|
|
336
|
+
}, z.core.$strip>>;
|
|
337
|
+
}, z.core.$strip>;
|
|
338
|
+
export type Qwen35Record = z.infer<typeof Qwen35RecordSchema>;
|
|
@@ -0,0 +1,139 @@
|
|
|
1
|
+
import { z } from 'zod';
|
|
2
|
+
export const Qwen35TextBlockSchema = z.object({
|
|
3
|
+
type: z.literal('text'),
|
|
4
|
+
text: z.string(),
|
|
5
|
+
});
|
|
6
|
+
export const Qwen35ImageBlockSchema = z.object({
|
|
7
|
+
type: z.literal('image'),
|
|
8
|
+
image_url: z.string().optional(),
|
|
9
|
+
placeholder: z.boolean().optional(),
|
|
10
|
+
placeholder_token: z.string().optional(),
|
|
11
|
+
source_kind: z.string().optional(),
|
|
12
|
+
metadata: z.record(z.string(), z.unknown()).optional(),
|
|
13
|
+
});
|
|
14
|
+
export const Qwen35VideoBlockSchema = z.object({
|
|
15
|
+
type: z.literal('video'),
|
|
16
|
+
video_url: z.string().optional(),
|
|
17
|
+
placeholder: z.boolean().optional(),
|
|
18
|
+
placeholder_token: z.string().optional(),
|
|
19
|
+
source_kind: z.string().optional(),
|
|
20
|
+
metadata: z.record(z.string(), z.unknown()).optional(),
|
|
21
|
+
});
|
|
22
|
+
export const Qwen35ContentSchema = z.union([
|
|
23
|
+
z.string(),
|
|
24
|
+
z.array(z.union([Qwen35TextBlockSchema, Qwen35ImageBlockSchema, Qwen35VideoBlockSchema])),
|
|
25
|
+
]);
|
|
26
|
+
export const Qwen35ToolCallSchema = z.object({
|
|
27
|
+
type: z.literal('function').default('function'),
|
|
28
|
+
id: z.string().optional(),
|
|
29
|
+
function: z.object({
|
|
30
|
+
name: z.string(),
|
|
31
|
+
arguments: z.record(z.string(), z.unknown()).default({}),
|
|
32
|
+
}),
|
|
33
|
+
});
|
|
34
|
+
export const Qwen35ToolSpecSchema = z.object({
|
|
35
|
+
name: z.string(),
|
|
36
|
+
description: z.string().optional(),
|
|
37
|
+
parameters: z.record(z.string(), z.unknown()).optional(),
|
|
38
|
+
});
|
|
39
|
+
export const Qwen35MessageSchema = z.discriminatedUnion('role', [
|
|
40
|
+
z.object({ role: z.literal('system'), content: Qwen35ContentSchema }),
|
|
41
|
+
z.object({ role: z.literal('user'), content: Qwen35ContentSchema }),
|
|
42
|
+
z.object({
|
|
43
|
+
role: z.literal('assistant'),
|
|
44
|
+
content: Qwen35ContentSchema,
|
|
45
|
+
reasoning_content: z.string().optional(),
|
|
46
|
+
tool_calls: z.array(Qwen35ToolCallSchema).optional(),
|
|
47
|
+
}),
|
|
48
|
+
z.object({
|
|
49
|
+
role: z.literal('tool'),
|
|
50
|
+
content: Qwen35ContentSchema,
|
|
51
|
+
tool_call_id: z.string().optional(),
|
|
52
|
+
name: z.string().optional(),
|
|
53
|
+
}),
|
|
54
|
+
]);
|
|
55
|
+
export const Qwen35MetaSchema = z.object({
|
|
56
|
+
endpoint: z.string(),
|
|
57
|
+
status: z.number().int().min(100).max(599),
|
|
58
|
+
ts: z.string(),
|
|
59
|
+
key: z.string().optional(),
|
|
60
|
+
source: z.string().optional(),
|
|
61
|
+
requested_model: z.string().nullable().optional(),
|
|
62
|
+
actual_model: z.string().nullable().optional(),
|
|
63
|
+
stream: z.boolean().optional(),
|
|
64
|
+
thinking_level: z.string().nullable().optional(),
|
|
65
|
+
reasoning_summary_mode: z.union([z.string(), z.array(z.unknown()), z.record(z.string(), z.unknown())]).optional(),
|
|
66
|
+
thinking_type: z.string().nullable().optional(),
|
|
67
|
+
thinking_budget_tokens: z.number().int().nonnegative().nullable().optional(),
|
|
68
|
+
max_output_tokens: z.number().int().nonnegative().nullable().optional(),
|
|
69
|
+
tool_spec_count: z.number().int().nonnegative().optional(),
|
|
70
|
+
tool_choice: z.union([z.string(), z.array(z.unknown()), z.record(z.string(), z.unknown())]).optional(),
|
|
71
|
+
request_contains_non_text_content: z.boolean().default(false),
|
|
72
|
+
request_image_block_count: z.number().int().nonnegative().default(0),
|
|
73
|
+
request_video_block_count: z.number().int().nonnegative().default(0),
|
|
74
|
+
request_tool_call_block_count: z.number().int().nonnegative().default(0),
|
|
75
|
+
request_tool_result_block_count: z.number().int().nonnegative().default(0),
|
|
76
|
+
request_thinking_block_count: z.number().int().nonnegative().default(0),
|
|
77
|
+
response_contains_non_text_content: z.boolean().default(false),
|
|
78
|
+
response_image_block_count: z.number().int().nonnegative().default(0),
|
|
79
|
+
response_video_block_count: z.number().int().nonnegative().default(0),
|
|
80
|
+
response_tool_call_block_count: z.number().int().nonnegative().default(0),
|
|
81
|
+
response_tool_result_block_count: z.number().int().nonnegative().default(0),
|
|
82
|
+
response_thinking_block_count: z.number().int().nonnegative().default(0),
|
|
83
|
+
request_truncated: z.boolean().default(false),
|
|
84
|
+
response_truncated: z.boolean().default(false),
|
|
85
|
+
lossy_source: z.boolean().default(false),
|
|
86
|
+
lossy_reasons: z.array(z.string()).default([]),
|
|
87
|
+
dataset_label: z.string().optional(),
|
|
88
|
+
dataset_source_system: z.string().optional(),
|
|
89
|
+
dataset_source_bucket: z.string().optional(),
|
|
90
|
+
dataset_source_file: z.string().optional(),
|
|
91
|
+
dataset_has_reasoning: z.boolean().optional(),
|
|
92
|
+
dataset_reasoning_chars: z.number().int().nonnegative().optional(),
|
|
93
|
+
});
|
|
94
|
+
export const Qwen35RecordSchema = z.object({
|
|
95
|
+
id: z.string(),
|
|
96
|
+
request_id: z.string().optional(),
|
|
97
|
+
messages: z.array(Qwen35MessageSchema).min(1),
|
|
98
|
+
tools: z.array(Qwen35ToolSpecSchema).default([]),
|
|
99
|
+
meta: Qwen35MetaSchema,
|
|
100
|
+
label: z.string().optional(),
|
|
101
|
+
source_system: z.string().optional(),
|
|
102
|
+
source_bucket: z.string().optional(),
|
|
103
|
+
source_file: z.string().optional(),
|
|
104
|
+
agentic_label: z
|
|
105
|
+
.object({
|
|
106
|
+
label: z.string(),
|
|
107
|
+
tool_call_count: z.number().int().nonnegative().optional(),
|
|
108
|
+
tool_message_count: z.number().int().nonnegative().optional(),
|
|
109
|
+
dialogue_rounds_est: z.number().int().nonnegative().optional(),
|
|
110
|
+
reasoning_chars: z.number().int().nonnegative().optional(),
|
|
111
|
+
has_reasoning: z.boolean().optional(),
|
|
112
|
+
lossy_source: z.boolean().optional(),
|
|
113
|
+
lossy_reasons: z.array(z.string()).optional(),
|
|
114
|
+
})
|
|
115
|
+
.optional(),
|
|
116
|
+
}).superRefine((record, ctx) => {
|
|
117
|
+
const seenUser = record.messages.some((message) => message.role === 'user');
|
|
118
|
+
if (!seenUser) {
|
|
119
|
+
ctx.addIssue({ code: 'custom', message: 'at least one user message is required' });
|
|
120
|
+
}
|
|
121
|
+
let seenNonSystem = false;
|
|
122
|
+
for (const message of record.messages) {
|
|
123
|
+
if (message.role !== 'system') {
|
|
124
|
+
seenNonSystem = true;
|
|
125
|
+
}
|
|
126
|
+
else if (seenNonSystem) {
|
|
127
|
+
ctx.addIssue({ code: 'custom', message: 'system messages must appear only at the beginning' });
|
|
128
|
+
break;
|
|
129
|
+
}
|
|
130
|
+
if (message.role === 'assistant' && typeof message.reasoning_content === 'string') {
|
|
131
|
+
if (message.reasoning_content.includes('<think>') || message.reasoning_content.includes('</think>')) {
|
|
132
|
+
ctx.addIssue({ code: 'custom', message: 'reasoning_content must not include <think> wrappers' });
|
|
133
|
+
}
|
|
134
|
+
}
|
|
135
|
+
}
|
|
136
|
+
if (record.meta.lossy_source && record.meta.lossy_reasons.length === 0) {
|
|
137
|
+
ctx.addIssue({ code: 'custom', message: 'lossy_source requires lossy_reasons' });
|
|
138
|
+
}
|
|
139
|
+
});
|
|
@@ -0,0 +1,64 @@
|
|
|
1
|
+
import fg from 'fast-glob';
|
|
2
|
+
import { z } from 'zod';
|
|
3
|
+
import { Qwen35RecordSchema } from '../schemas/qwen35.js';
|
|
4
|
+
import { readJsonl } from '../utils/jsonl.js';
|
|
5
|
+
const EntrySchema = z.record(z.string(), z.unknown());
|
|
6
|
+
export async function collectClaudePromptOnlyRecords(root) {
|
|
7
|
+
const files = await fg('**/*.jsonl', { cwd: root, absolute: true, onlyFiles: true });
|
|
8
|
+
const records = [];
|
|
9
|
+
for (const file of files.sort()) {
|
|
10
|
+
const entries = (await readJsonl(file)).map((row) => EntrySchema.parse(row));
|
|
11
|
+
for (const entry of entries) {
|
|
12
|
+
if (entry.type !== 'user')
|
|
13
|
+
continue;
|
|
14
|
+
const message = isRecord(entry.message) ? entry.message : {};
|
|
15
|
+
const content = asString(message.content);
|
|
16
|
+
if (!content)
|
|
17
|
+
continue;
|
|
18
|
+
records.push(Qwen35RecordSchema.parse({
|
|
19
|
+
id: `${asString(entry.sessionId) ?? file}:${asString(entry.promptId) ?? asString(entry.uuid) ?? 'prompt'}`,
|
|
20
|
+
request_id: asString(entry.promptId) ?? asString(entry.uuid) ?? undefined,
|
|
21
|
+
messages: [{ role: 'user', content }],
|
|
22
|
+
tools: [],
|
|
23
|
+
meta: {
|
|
24
|
+
endpoint: 'claude/prompt_history',
|
|
25
|
+
status: 200,
|
|
26
|
+
ts: asString(entry.timestamp) ?? '',
|
|
27
|
+
key: asString(entry.sessionId) ?? undefined,
|
|
28
|
+
source: `claude:session=${asString(entry.sessionId)}:cwd=${asString(entry.cwd)}:entrypoint=${asString(entry.entrypoint)}`,
|
|
29
|
+
requested_model: undefined,
|
|
30
|
+
actual_model: undefined,
|
|
31
|
+
stream: false,
|
|
32
|
+
thinking_level: undefined,
|
|
33
|
+
reasoning_summary_mode: 'claude_prompt_only',
|
|
34
|
+
thinking_type: 'prompt_history_only',
|
|
35
|
+
tool_spec_count: 0,
|
|
36
|
+
tool_choice: { mode: 'prompt_only' },
|
|
37
|
+
request_contains_non_text_content: false,
|
|
38
|
+
request_image_block_count: 0,
|
|
39
|
+
request_video_block_count: 0,
|
|
40
|
+
request_tool_call_block_count: 0,
|
|
41
|
+
request_tool_result_block_count: 0,
|
|
42
|
+
request_thinking_block_count: 0,
|
|
43
|
+
response_contains_non_text_content: false,
|
|
44
|
+
response_image_block_count: 0,
|
|
45
|
+
response_video_block_count: 0,
|
|
46
|
+
response_tool_call_block_count: 0,
|
|
47
|
+
response_tool_result_block_count: 0,
|
|
48
|
+
response_thinking_block_count: 0,
|
|
49
|
+
request_truncated: false,
|
|
50
|
+
response_truncated: false,
|
|
51
|
+
lossy_source: true,
|
|
52
|
+
lossy_reasons: ['prompt_history_only', 'assistant_trace_unavailable'],
|
|
53
|
+
},
|
|
54
|
+
}));
|
|
55
|
+
}
|
|
56
|
+
}
|
|
57
|
+
return records;
|
|
58
|
+
}
|
|
59
|
+
function asString(value) {
|
|
60
|
+
return typeof value === 'string' ? value : undefined;
|
|
61
|
+
}
|
|
62
|
+
function isRecord(value) {
|
|
63
|
+
return Boolean(value) && typeof value === 'object' && !Array.isArray(value);
|
|
64
|
+
}
|