@qwen-code/arena-agent 0.0.10-dev.13 → 0.0.10-dev.14
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/CHANGELOG.md +9 -1
- package/dist/model-capabilities.js +18 -8
- package/dist/model-capabilities.js.map +1 -1
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -12,6 +12,13 @@
|
|
|
12
12
|
|
|
13
13
|
## [Unreleased]
|
|
14
14
|
|
|
15
|
+
## [0.0.10-dev.14] - 2026-10-26
|
|
16
|
+
|
|
17
|
+
### 通用
|
|
18
|
+
|
|
19
|
+
- 修复 Qwen4 及后续代际的图片输入识别,默认仅支持文字和图片。
|
|
20
|
+
- 修复 Claude Fable/Mythos、Vertex 日期型号及 DeepSeek Flash 视觉型号的输入模态识别。
|
|
21
|
+
|
|
15
22
|
## [0.0.10-dev.13] - 2026-10-23
|
|
16
23
|
|
|
17
24
|
### Agent Arena
|
|
@@ -260,7 +267,8 @@
|
|
|
260
267
|
- `bash` 用于一次性命令,区分超时与信号终止,并清理遗留后台进程;需要持续运行
|
|
261
268
|
的任务改用进程工具。预览端口支持 E2B Host 探测及按需返回告警。
|
|
262
269
|
|
|
263
|
-
[Unreleased]: https://code.alibaba-inc.com/know-how/arena-harness/compare?from=v0.0.10-dev.
|
|
270
|
+
[Unreleased]: https://code.alibaba-inc.com/know-how/arena-harness/compare?from=v0.0.10-dev.14&to=master&mergeBase=true
|
|
271
|
+
[0.0.10-dev.14]: https://code.alibaba-inc.com/know-how/arena-harness/compare?from=v0.0.10-dev.13&to=v0.0.10-dev.14&mergeBase=true
|
|
264
272
|
[0.0.10-dev.13]: https://code.alibaba-inc.com/know-how/arena-harness/compare?from=v0.0.10-dev.12&to=v0.0.10-dev.13&mergeBase=true
|
|
265
273
|
[0.0.10-dev.12]: https://code.alibaba-inc.com/know-how/arena-harness/compare?from=v0.0.10-dev.11&to=v0.0.10-dev.12&mergeBase=true
|
|
266
274
|
[0.0.10-dev.11]: https://code.alibaba-inc.com/know-how/arena-harness/compare?from=v0.0.10-dev.10&to=v0.0.10-dev.11&mergeBase=true
|
|
@@ -3,8 +3,8 @@ const allModalities = [
|
|
|
3
3
|
"video",
|
|
4
4
|
"audio",
|
|
5
5
|
];
|
|
6
|
-
function rule(id, pattern, modalities) {
|
|
7
|
-
return { id, pattern, modalities };
|
|
6
|
+
function rule(id, pattern, modalities, qualifiedPattern) {
|
|
7
|
+
return { id, pattern, modalities, qualifiedPattern };
|
|
8
8
|
}
|
|
9
9
|
// Ordered from the most specific model/version to broader families. These are
|
|
10
10
|
// input capabilities for the chat model, not image/audio/video generation
|
|
@@ -19,6 +19,9 @@ const capabilityRules = [
|
|
|
19
19
|
rule("qwen3.8-max", /^qwen3\.8-max(?:[-.:]|$)/, ["image"]),
|
|
20
20
|
rule("qwen3.7-max-text", /^qwen3\.7-max(?:[-.:]|$)/, []),
|
|
21
21
|
rule("qwen3.6-27b-text", /^qwen3\.6-27b(?:[-.:]|$)/, []),
|
|
22
|
+
// Qwen4 and later default to text + image input, including Max/Coder
|
|
23
|
+
// variants. Do not inherit audio/video capabilities from older families.
|
|
24
|
+
rule("qwen4-plus-vision", /^qwen(?:[4-9]|[1-9]\d+)(?:\.\d+)?(?:[-.:]|$)/, ["image"]),
|
|
22
25
|
rule("qwen-omni", /^(?:qwen(?:\d+(?:\.\d+)?)?-omni|qwen3\.5-omni)(?:[-.:]|$)/, ["image", "video", "audio"]),
|
|
23
26
|
rule("qwen-vl-qvq", /^(?:qwen(?:\d+(?:\.\d+)?)?-vl|qwen-vl|qvq)(?:[-.:]|$)/, ["image", "video"]),
|
|
24
27
|
rule("qwen-plus-flash-multimodal", /^(?:qwen-plus|qwen3\.(?:5|6|7)-plus|qwen3\.(?:5|6)-flash)(?:[-.:]|$)/, ["image", "video"]),
|
|
@@ -27,8 +30,8 @@ const capabilityRules = [
|
|
|
27
30
|
rule("qwen-known-text", /^(?:qwen(?:\d+(?:\.\d+)?)?-(?:max|turbo|coder|long)|qwq)(?:[-.:]|$)/, []),
|
|
28
31
|
// Claude 3 and later accept image blocks. Claude Messages does not accept
|
|
29
32
|
// native audio/video blocks; those require preprocessing or another model.
|
|
30
|
-
rule("claude-3-plus-vision", /(?:^|[._-])claude-(?:(?:3|[4-9]|[1-9]\d)(?:[._
|
|
31
|
-
rule("claude-legacy-text", /(?:^|[._-])claude-(?:instant|v?[12])(?:[._
|
|
33
|
+
rule("claude-3-plus-vision", /(?:^|[._-])claude-(?:(?:3|[4-9]|[1-9]\d)(?:[._:@-]|$)|(?:opus|sonnet|haiku|fable|mythos)-(?:3|[4-9]|[1-9]\d)(?:[._:@-]|$)|mythos-preview(?:[._:@-]|$))/, ["image"]),
|
|
34
|
+
rule("claude-legacy-text", /(?:^|[._-])claude-(?:instant|v?[12])(?:[._:@-]|$)/, []),
|
|
32
35
|
// Audio chat models are intentionally checked before the broader GPT-4o
|
|
33
36
|
// vision rule: their documented Chat Completions inputs are text + audio.
|
|
34
37
|
rule("gpt-realtime-multimodal", /^(?:gpt-realtime|gpt-4o(?:-mini)?-realtime)(?:[-.:]|$)/, ["image", "audio"]),
|
|
@@ -39,9 +42,14 @@ const capabilityRules = [
|
|
|
39
42
|
rule("openai-reasoning-vision", /^(?:o1(?:-pro)?|o3(?:-pro)?|o4-mini)(?:-\d{4}-\d{2}-\d{2})?$/, ["image"]),
|
|
40
43
|
rule("openai-computer-use-vision", /^computer-use-preview(?:[-.:]|$)/, ["image"]),
|
|
41
44
|
rule("openai-known-text", /^(?:gpt-3\.5|gpt-4|gpt-oss|o1-(?:mini|preview)|o3-mini)(?:[-.:]|$)/, []),
|
|
42
|
-
// The
|
|
43
|
-
//
|
|
44
|
-
|
|
45
|
+
// The original V4 Flash weights remain text-only, while the hosted alias
|
|
46
|
+
// now routes to V4.1 Flash. Preserve the explicit original weight identity.
|
|
47
|
+
rule("deepseek-v4-flash-original-text", /^deepseek-v4-flash$/, [], /(?:^|\/)deepseek-ai\/deepseek-v4-flash$/),
|
|
48
|
+
// Only confirmed visual Flash names and hosted aliases bypass the legacy
|
|
49
|
+
// text family below; this does not assume future DeepSeek versions can see.
|
|
50
|
+
// https://api-docs.deepseek.com/quick_start/pricing/
|
|
51
|
+
rule("deepseek-flash-vision", /^deepseek-(?:flash|v4-flash(?:-vision-exp)?|v4\.1-flash(?:-1017)?)$/, ["image"]),
|
|
52
|
+
// Separately deployed VL/OCR families also accept images.
|
|
45
53
|
rule("deepseek-vl-ocr", /^deepseek-(?:vl|ocr)(?:[-.:\d]|$)/, ["image"]),
|
|
46
54
|
rule("deepseek-hosted-text", /^deepseek-(?:chat|reasoner|coder|r1|v[234])(?:[-.:]|$)/, []),
|
|
47
55
|
// Kimi's current K2/K3 families and managed coding model advertise both
|
|
@@ -68,7 +76,9 @@ export function normalizeModelId(model) {
|
|
|
68
76
|
}
|
|
69
77
|
export function modelInputCapabilities(model) {
|
|
70
78
|
const normalizedModel = normalizeModelId(model);
|
|
71
|
-
const
|
|
79
|
+
const qualifiedModel = model.trim().toLowerCase();
|
|
80
|
+
const matched = capabilityRules.find((candidate) => candidate.pattern.test(normalizedModel) &&
|
|
81
|
+
(!candidate.qualifiedPattern || candidate.qualifiedPattern.test(qualifiedModel)));
|
|
72
82
|
if (!matched) {
|
|
73
83
|
return {
|
|
74
84
|
model,
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"model-capabilities.js","sourceRoot":"","sources":["../src/model-capabilities.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"model-capabilities.js","sourceRoot":"","sources":["../src/model-capabilities.ts"],"names":[],"mappings":"AA0BA,MAAM,aAAa,GAAkC;IACnD,OAAO;IACP,OAAO;IACP,OAAO;CACR,CAAC;AAEF,SAAS,IAAI,CACX,EAAU,EACV,OAAe,EACf,UAAyC,EACzC,gBAAyB;IAEzB,OAAO,EAAE,EAAE,EAAE,OAAO,EAAE,UAAU,EAAE,gBAAgB,EAAE,CAAC;AACvD,CAAC;AAED,8EAA8E;AAC9E,0EAA0E;AAC1E,4EAA4E;AAC5E,+EAA+E;AAC/E,MAAM,eAAe,GAA8B;IACjD,0EAA0E;IAC1E,yEAAyE;IACzE,yEAAyE;IACzE,IAAI,CAAC,aAAa,EAAE,yBAAyB,EAAE,CAAC,OAAO,CAAC,CAAC;IACzD,IAAI,CACF,wBAAwB,EACxB,qCAAqC,EACrC,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CAAC,aAAa,EAAE,0BAA0B,EAAE,CAAC,OAAO,CAAC,CAAC;IAC1D,IAAI,CAAC,kBAAkB,EAAE,0BAA0B,EAAE,EAAE,CAAC;IACxD,IAAI,CAAC,kBAAkB,EAAE,0BAA0B,EAAE,EAAE,CAAC;IACxD,qEAAqE;IACrE,yEAAyE;IACzE,IAAI,CACF,mBAAmB,EACnB,8CAA8C,EAC9C,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,WAAW,EACX,2DAA2D,EAC3D,CAAC,OAAO,EAAE,OAAO,EAAE,OAAO,CAAC,CAC5B;IACD,IAAI,CACF,aAAa,EACb,uDAAuD,EACvD,CAAC,OAAO,EAAE,OAAO,CAAC,CACnB;IACD,IAAI,CACF,4BAA4B,EAC5B,sEAAsE,EACtE,CAAC,OAAO,EAAE,OAAO,CAAC,CACnB;IACD,IAAI,CACF,6BAA6B,EAC7B,4CAA4C,EAC5C,CAAC,OAAO,EAAE,OAAO,CAAC,CACnB;IACD,IAAI,CACF,YAAY,EACZ,yDAAyD,EACzD,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,iBAAiB,EACjB,qEAAqE,EACrE,EAAE,CACH;IAED,0EAA0E;IAC1E,2EAA2E;IAC3E,IAAI,CACF,sBAAsB,EACtB,wJAAwJ,EACxJ,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,oBAAoB,EACpB,mDAAmD,EACnD,EAAE,CACH;IAED,wEAAwE;IACxE,0EAA0E;IAC1E,IAAI,CACF,yBAAyB,EACzB,wDAAwD,EACxD,CAAC,OAAO,EAAE,OAAO,CAAC,CACnB;IACD,IAAI,CACF,WAAW,EACX,wCAAwC,EACxC,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,cAAc,EACd,mBAAmB,EACnB,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,uBAAuB,EACvB,qDAAqD,EACrD,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,oBAAoB,EACpB,oCAAoC,EACpC,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,yBAAyB,EACzB,8DAA8D,EAC9D,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CAAC,4BAA4B,EAAE,kCAAkC,EAAE,CAAC,OAAO,CAAC,CAAC;IACjF,IAAI,CACF,mBAAmB,EACnB,oEAAoE,EACpE,EAAE,CACH;IAED,yEAAyE;IACzE,4EAA4E;IAC5E,IAAI,CACF,iCAAiC,EACjC,qBAAqB,EACrB,EAAE,EACF,yCAAyC,CAC1C;IACD,yEAAyE;IACzE,4EAA4E;IAC5E,qDAAqD;IACrD,IAAI,CACF,uBAAuB,EACvB,qEAAqE,EACrE,CAAC,OAAO,CAAC,CACV;IACD,0DAA0D;IAC1D,IAAI,CACF,iBAAiB,EACjB,mCAAmC,EACnC,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CACF,sBAAsB,EACtB,wDAAwD,EACxD,EAAE,CACH;IAED,wEAAwE;IACxE,wEAAwE;IACxE,UAAU;IACV,IAAI,CAAC,oBAAoB,EAAE,qBAAqB,EAAE,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;IACrE,IAAI,CACF,oBAAoB,EACpB,qDAAqD,EACrD,CAAC,OAAO,EAAE,OAAO,CAAC,CACnB;IACD,IAAI,CACF,oBAAoB,EACpB,kCAAkC,EAClC,CAAC,OAAO,CAAC,CACV;IACD,IAAI,CAAC,kBAAkB,EAAE,yBAAyB,EAAE,EAAE,CAAC;IAEvD,sEAAsE;IACtE,0EAA0E;IAC1E,0EAA0E;IAC1E,oDAAoD;IACpD,IAAI,CAAC,cAAc,EAAE,sBAAsB,EAAE,EAAE,CAAC;IAEhD,4EAA4E;IAC5E,2EAA2E;IAC3E,2EAA2E;IAC3E,IAAI,CACF,2BAA2B,EAC3B,uFAAuF,EACvF,CAAC,OAAO,EAAE,OAAO,EAAE,OAAO,CAAC,CAC5B;CACF,CAAC;AAEF,MAAM,UAAU,gBAAgB,CAAC,KAAa;IAC5C,MAAM,UAAU,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IAC9C,MAAM,SAAS,GAAG,UAAU,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC;IAC9C,OAAO,SAAS,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAC,UAAU,CAAC,KAAK,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC;AACzE,CAAC;AAED,MAAM,UAAU,sBAAsB,CAAC,KAAa;IAClD,MAAM,eAAe,GAAG,gBAAgB,CAAC,KAAK,CAAC,CAAC;IAChD,MAAM,cAAc,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IAClD,MAAM,OAAO,GAAG,eAAe,CAAC,IAAI,CAAC,CAAC,SAAS,EAAE,EAAE,CACjD,SAAS,CAAC,OAAO,CAAC,IAAI,CAAC,eAAe,CAAC;QACvC,CAAC,CAAC,SAAS,CAAC,gBAAgB,IAAI,SAAS,CAAC,gBAAgB,CAAC,IAAI,CAAC,cAAc,CAAC,CAAC,CACjF,CAAC;IACF,IAAI,CAAC,OAAO,EAAE,CAAC;QACb,OAAO;YACL,KAAK;YACL,eAAe;YACf,KAAK,EAAE,KAAK;SACb,CAAC;IACJ,CAAC;IACD,MAAM,SAAS,GAAG,IAAI,GAAG,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC;IAC9C,OAAO;QACL,KAAK;QACL,eAAe;QACf,KAAK,EAAE,IAAI;QACX,IAAI,EAAE,OAAO,CAAC,EAAE;QAChB,GAAG,MAAM,CAAC,WAAW,CACnB,aAAa,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC,QAAQ,EAAE,SAAS,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC,CAAC,CACrE;KACF,CAAC;AACJ,CAAC;AAED,SAAS,gBAAgB,CACvB,MAAqC,EACrC,QAA4B;IAE5B,IAAI,QAAQ,KAAK,OAAO;QAAE,OAAO,MAAM,CAAC,uBAAuB,CAAC;IAChE,IAAI,QAAQ,KAAK,OAAO;QAAE,OAAO,MAAM,CAAC,uBAAuB,CAAC;IAChE,OAAO,MAAM,CAAC,uBAAuB,CAAC;AACxC,CAAC;AAED,MAAM,UAAU,kBAAkB,CAChC,MAAqC,EACrC,QAA4B;IAE5B,MAAM,QAAQ,GAAG,gBAAgB,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IACpD,IAAI,QAAQ,KAAK,SAAS;QAAE,OAAO,QAAQ,CAAC;IAC5C,MAAM,QAAQ,GAAG,sBAAsB,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,QAAQ,CAAC,CAAC;IAChE,uEAAuE;IACvE,qEAAqE;IACrE,uDAAuD;IACvD,OAAO,QAAQ,IAAI,IAAI,CAAC;AAC1B,CAAC;AAED,MAAM,UAAU,kBAAkB,CAChC,MAAqC;IAErC,OAAO,kBAAkB,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;AAC7C,CAAC"}
|