my-opencode 2.0.0 → 3.0.0

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 CHANGED
@@ -62,7 +62,6 @@ npm link # 將 bin/moc 註冊到 PATH,讓 moc 指令全域可用
62
62
  ```bash
63
63
  moc # 預設 opencode profile
64
64
  moc openai # 切到 OpenAI profile
65
- moc google # 切到 Google profile
66
65
  moc copilot # 切到 Copilot profile
67
66
  moc --help # 透明轉發到 opencode --help
68
67
  ```
@@ -80,17 +79,27 @@ npx my-opencode help # 顯示說明
80
79
 
81
80
  內建 `provider-profile` plugin / resolver,主要透過 `moc` 切換整套模型配置。
82
81
 
83
- 支援值:`opencode`(預設)、`openai`、`google`、`copilot`
82
+ 支援值:`opencode`(預設)、`openai`、`copilot`、`anthropic`
84
83
 
85
84
  | Profile | Primary / heavy | Small / verifier |
86
85
  |---------|------------------|------------------|
87
- | `opencode` | `opencode-go/deepseek-v4-pro` | `opencode-go/deepseek-v4-flash` |
88
- | `openai` | `openai/gpt-5.6` | `openai/gpt-5.6-luna` |
89
- | `google` | `google/gemini-3.1-pro-preview-customtools` | `google/gemini-3.5-flash` |
90
- | `copilot` | `github-copilot/gpt-5.6-sol` | `github-copilot/gpt-5.6-luna` |
91
-
92
- > 💡 `openai` 與 `copilot` profile 的 `gpt-5.6-luna`(verifier / 小型任務)支援 reasoning effort 切換:
93
- > **high / medium / low** 三檔(預設 `low`)。在 opencode 內用 `variant_cycle` 快捷鍵循環切換。
86
+ | `opencode` | model:`opencode-go/gpt-6-luna#max`;heavy:`opencode-go/claude-haiku-5-5#max` | `opencode-go/deepseek-v4.1-flash#max` |
87
+ | `openai` | heavy:`openai/gpt-6.2-sol#medium`;model:`openai/gpt-6.2-sol#low` | `openai/gpt-6.1-terra#medium` |
88
+ | `copilot` | heavy:`github-copilot/gpt-6.1-sol#medium`;model:`github-copilot/claude-haiku-5.5#xhigh` | `github-copilot/gpt-6-luna#max` |
89
+ | `anthropic` | heavy:`anthropic/claude-sonnet-5-5#high`;default:`anthropic/claude-haiku-5-5#max` | `anthropic/claude-haiku-5-5#high` |
90
+
91
+ > 💡 `anthropic` profile 以 `#variant` 指定 effort(opencode 的 `provider/model#variant` 語法):
92
+ > heavy agents(SA / SD / reviewer / team-lead / architect)用 `sonnet-5-5#high`(最高);
93
+ > 其餘 implementation agents 預設 `haiku-5-5#max`(其次);verifier 與 title 用 `haiku-5-5#high`(最低)。
94
+ >
95
+ > 💡 `opencode` profile 全部使用 `max` effort:預設 implementation agents 用 `gpt-6-luna#max`;heavy agents(SA / SD / reviewer / team-lead / architect)用 `claude-haiku-5-5#max`;verifier 與 title 用 `deepseek-v4.1-flash#max`(最輕量)。
96
+ > `deepseek-v4.1-flash` 的 effort 可在 opencode 內用 `variant_cycle` 快捷鍵循環切換(low / high / max)。
97
+ >
98
+ > 💡 `openai` profile 以 `#variant` 指定 effort:heavy agents(SA / SD / reviewer / team-lead / architect)用 `gpt-6.2-sol#medium`(最高);
99
+ > 其餘 implementation agents 預設 `gpt-6.2-sol#low`(其次);verifier 與 title 用 `gpt-6.1-terra#medium`(最低)。
100
+ >
101
+ > 💡 `copilot` profile 以 `#variant` 指定 effort:heavy agents(SA / SD / reviewer / team-lead / architect)用 `gpt-6.1-sol#medium`(最高);
102
+ > 其餘 implementation agents 預設 `claude-haiku-5.5#xhigh`(其次);verifier 與 title 用 `gpt-6-luna#max`(最低)。
94
103
 
95
104
  ```bash
96
105
  # 預設 profile(opencode)
@@ -99,15 +108,15 @@ moc
99
108
  # 切到 OpenAI
100
109
  moc openai
101
110
 
102
- # 切到 Google
103
- moc google
111
+ # 切到 Anthropic
112
+ moc anthropic
104
113
 
105
114
  # 轉發額外參數
106
115
  moc openai --list-agents
107
116
  ```
108
117
 
109
118
  `moc` 會在目前工作目錄尋找 `.opencode/plugins/provider-profile.mjs` 與 `.opencode/opencode.json`。
110
- 對 `opencode` / `openai` / `google` / `copilot` 這四個 profile,會把產生出的 `OPENCODE_CONFIG_CONTENT` 快取到 `.opencode/cache/provider-profile/`;只要 profile、plugin mtime、config mtime 都沒變,就直接重用快取,不再重新跑 Node。若 plugin 缺失,則會警告並回退為 plain `opencode`。
119
+ 對 `opencode` / `openai` / `copilot` / `anthropic` 這四個 profile,會把產生出的 `OPENCODE_CONFIG_CONTENT` 快取到 `.opencode/cache/provider-profile/`;只要 profile、plugin mtime、config mtime 都沒變,就直接重用快取,不再重新跑 Node。若 plugin 缺失,則會警告並回退為 plain `opencode`。
111
120
 
112
121
  若你直接用 repo 內的 `bin/moc`,仍會走既有 POSIX shell launcher;透過 npm 安裝、`npm link`、`npx` 取得的 `moc` 則會走跨平台的 `bin/moc.js`。
113
122
 
@@ -124,7 +133,7 @@ opencode
124
133
  ```json
125
134
  {
126
135
  "plugin": [
127
- ["plugins/provider-profile.mjs", { "profile": "google" }]
136
+ ["plugins/provider-profile.mjs", { "profile": "openai" }]
128
137
  ]
129
138
  }
130
139
  ```
package/bin/cli.js CHANGED
@@ -127,7 +127,7 @@ function cmdHelp() {
127
127
  \x1b[1mLAUNCHER\x1b[0m
128
128
  moc 預設使用 opencode profile 啟動 opencode
129
129
  moc openai 使用 openai profile 啟動 opencode
130
- moc google 使用 google profile 啟動 opencode
130
+ moc anthropic 使用 anthropic profile 啟動 opencode
131
131
  moc --help 透明轉發到 opencode --help
132
132
 
133
133
  \x1b[1mEXAMPLES\x1b[0m
package/bin/moc CHANGED
@@ -1,7 +1,7 @@
1
1
  #!/bin/sh
2
2
 
3
3
  profile=
4
- known_profiles="opencode openai google copilot"
4
+ known_profiles="opencode openai copilot anthropic"
5
5
 
6
6
  print_help() {
7
7
  cat <<'EOF'
@@ -12,14 +12,13 @@ Usage:
12
12
  moc print-config [profile]
13
13
 
14
14
  Profiles:
15
- opencode | openai | google | copilot
15
+ opencode | openai | copilot | anthropic
16
16
 
17
17
  Examples:
18
18
  moc
19
- moc google
20
19
  moc openai --list-agents
21
- moc doctor google
22
- moc print-config google
20
+ moc doctor anthropic
21
+ moc print-config anthropic
23
22
 
24
23
  Notes:
25
24
  - `moc` without extra args launches interactive opencode.
@@ -54,7 +53,7 @@ case "${1-}" in
54
53
  print-config)
55
54
  profile=${2-opencode}
56
55
  ;;
57
- opencode|openai|google|copilot)
56
+ opencode|openai|copilot|anthropic)
58
57
  profile=$1
59
58
  shift
60
59
  ;;
@@ -65,7 +64,7 @@ if [ -z "$profile" ]; then
65
64
  fi
66
65
 
67
66
  case "$profile" in
68
- opencode|openai|google|copilot)
67
+ opencode|openai|copilot|anthropic)
69
68
  ;;
70
69
  *)
71
70
  printf '%s\n' "moc: unknown profile '$profile'. Must be one of: $known_profiles" >&2
@@ -74,10 +73,19 @@ case "$profile" in
74
73
  esac
75
74
 
76
75
  exec_opencode_with_profile() {
76
+ # Profile launches always use a private server (--standalone). A background
77
+ # service started without this profile env would otherwise ignore it.
78
+ standalone_flag=--standalone
79
+ for arg in "$@"; do
80
+ if [ "$arg" = "--standalone" ]; then
81
+ standalone_flag=
82
+ fi
83
+ done
84
+
77
85
  exec env \
78
86
  OPENCODE_PROVIDER_PROFILE="$profile" \
79
87
  OPENCODE_CONFIG_CONTENT="$config_content" \
80
- opencode "$@"
88
+ opencode ${standalone_flag:+"$standalone_flag"} "$@"
81
89
  }
82
90
 
83
91
  plugin_path=.opencode/plugins/provider-profile.mjs
package/bin/moc.js CHANGED
@@ -5,8 +5,8 @@ import fs from 'node:fs';
5
5
  import path from 'node:path';
6
6
  import { spawn, spawnSync } from 'node:child_process';
7
7
 
8
- const knownProfiles = new Set(['opencode', 'openai', 'google', 'copilot']);
9
- const knownProfilesText = 'opencode openai google copilot';
8
+ const knownProfiles = new Set(['opencode', 'openai', 'copilot', 'anthropic']);
9
+ const knownProfilesText = 'opencode openai copilot anthropic';
10
10
  const signalExitCodes = {
11
11
  SIGHUP: 1,
12
12
  SIGINT: 2,
@@ -341,7 +341,13 @@ function main() {
341
341
  process.exit(0);
342
342
  }
343
343
 
344
- spawnOpencode(forwardedArgs, {
344
+ // Profile launches always use a private server (--standalone). A background
345
+ // service started without this profile env would otherwise ignore it.
346
+ const launchArgs = forwardedArgs.includes('--standalone')
347
+ ? forwardedArgs
348
+ : ['--standalone', ...forwardedArgs];
349
+
350
+ spawnOpencode(launchArgs, {
345
351
  OPENCODE_PROVIDER_PROFILE: profile,
346
352
  OPENCODE_CONFIG_CONTENT: configContent,
347
353
  });
package/opencode.json CHANGED
@@ -1,7 +1,29 @@
1
1
  {
2
2
  "$schema": "https://opencode.ai/config.json",
3
- "model": "opencode-go/deepseek-v4-pro",
3
+ "model": "opencode-go/deepseek-v4.1-flash",
4
4
  "small_model": "opencode-go/deepseek-v4-flash",
5
+ "provider": {
6
+ "opencode-go": {
7
+ "models": {
8
+ "deepseek-v4.1-flash": {
9
+ "options": {
10
+ "reasoningEffort": "max"
11
+ },
12
+ "variants": {
13
+ "low": {
14
+ "reasoningEffort": "low"
15
+ },
16
+ "high": {
17
+ "reasoningEffort": "high"
18
+ },
19
+ "max": {
20
+ "reasoningEffort": "max"
21
+ }
22
+ }
23
+ }
24
+ }
25
+ }
26
+ },
5
27
  "default_agent": "team-lead",
6
28
  "instructions": [
7
29
  "skills/adr/SKILL.md",
@@ -33,7 +55,7 @@
33
55
  "design-evidence": {
34
56
  "description": "Extract source-backed Design Evidence Packs from Figma, screenshots, wireframes, and prototypes.",
35
57
  "mode": "subagent",
36
- "model": "opencode-go/deepseek-v4-pro",
58
+ "model": "opencode-go/deepseek-v4.1-flash",
37
59
  "prompt": "{file:agents/design-evidence.md}",
38
60
  "tools": {
39
61
  "read": true,
@@ -45,7 +67,7 @@
45
67
  "business-analyst": {
46
68
  "description": "SA role: translate evidence into business requirements, rules, states, data meaning, and acceptance intent.",
47
69
  "mode": "subagent",
48
- "model": "opencode-go/deepseek-v4-pro",
70
+ "model": "opencode-go/deepseek-v4.1-flash",
49
71
  "prompt": "{file:agents/business-analyst.md}",
50
72
  "tools": {
51
73
  "read": true,
@@ -57,7 +79,7 @@
57
79
  "system-designer": {
58
80
  "description": "SD role: translate business requirements into technical, API, data, security, and operational contracts.",
59
81
  "mode": "subagent",
60
- "model": "opencode-go/deepseek-v4-pro",
82
+ "model": "opencode-go/deepseek-v4.1-flash",
61
83
  "prompt": "{file:agents/system-designer.md}",
62
84
  "tools": {
63
85
  "read": true,
@@ -103,7 +125,7 @@
103
125
  "code-reviewer": {
104
126
  "description": "Systematic code review across correctness, security, quality, testing, and style dimensions.",
105
127
  "mode": "subagent",
106
- "model": "opencode-go/deepseek-v4-pro",
128
+ "model": "opencode-go/deepseek-v4.1-flash",
107
129
  "prompt": "{file:agents/code-reviewer.md}",
108
130
  "tools": {
109
131
  "read": true,
@@ -137,7 +159,7 @@
137
159
  "architect": {
138
160
  "description": "Capture architecture decisions as ADR documents. Analyzes tradeoffs and records context.",
139
161
  "mode": "subagent",
140
- "model": "opencode-go/deepseek-v4-pro",
162
+ "model": "opencode-go/deepseek-v4.1-flash",
141
163
  "prompt": "{file:agents/architect.md}",
142
164
  "tools": {
143
165
  "read": true,
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "my-opencode",
3
- "version": "2.0.0",
3
+ "version": "3.0.0",
4
4
  "description": "高品質 opencode 開發流程 skill 與 agents 集合 — Evidence-First 開發流程、ADR、PR Review、測試生成、文件生成",
5
5
  "keywords": [
6
6
  "opencode",
package/plugins/README.md CHANGED
@@ -9,19 +9,26 @@ opencode plugin 是透過 `opencode.json` 的 `plugin` 區塊註冊的 npm 或
9
9
  零依賴 ESM plugin / resolver,用來切換本 repo 的模型 profile。
10
10
 
11
11
  - 預設 profile:`opencode`
12
- - 支援:`opencode`、`openai`、`google`、`copilot`
12
+ - 支援:`opencode`、`openai`、`copilot`、`anthropic`
13
13
  - 讀取:`OPENCODE_PROVIDER_PROFILE`
14
- - 可透過 plugin tuple options 覆寫:`["plugins/provider-profile.mjs", { "profile": "google" }]`
14
+ - 可透過 plugin tuple options 覆寫:`["plugins/provider-profile.mjs", { "profile": "openai" }]`
15
15
 
16
16
  | Profile | Primary / heavy | Small / verifier |
17
17
  |---------|------------------|------------------|
18
- | `opencode` | `opencode-go/deepseek-v4-pro` | `opencode-go/deepseek-v4-flash` |
19
- | `openai` | `openai/gpt-5.6` | `openai/gpt-5.6-luna` |
20
- | `google` | `google/gemini-3.1-pro-preview-customtools` | `google/gemini-3.5-flash` |
21
- | `copilot` | `github-copilot/gpt-5.6-sol` | `github-copilot/gpt-5.6-luna` |
18
+ | `opencode` | model:`opencode-go/gpt-6-luna#max`;heavy:`opencode-go/claude-haiku-5-5#max` | `opencode-go/deepseek-v4.1-flash#max` |
19
+ | `openai` | heavy:`openai/gpt-6.2-sol#medium`;model:`openai/gpt-6.2-sol#low` | `openai/gpt-6.1-terra#medium` |
20
+ | `copilot` | heavy:`github-copilot/gpt-6.1-sol#medium`;model:`github-copilot/claude-haiku-5.5#xhigh` | `github-copilot/gpt-6-luna#max` |
21
+ | `anthropic` | heavy:`anthropic/claude-sonnet-5-5#high`;default:`anthropic/claude-haiku-5-5#max` | `anthropic/claude-haiku-5-5#high` |
22
22
 
23
- > 💡 `openai` 與 `copilot` profile 的 `gpt-5.6-luna` 支援 reasoning effort 切換(high / medium / low,預設 `low`),
24
- > 在 opencode 內用 `variant_cycle` 快捷鍵循環切換。
23
+ > 💡 `anthropic` profile 的 `#<variant>` 後綴為 opencode 的 effort 指定語法(`provider/model#variant`),
24
+ > 同一顆 haiku 可依 agent 使用不同 effort。variant 由 opencode 內建 model catalog 提供,因此此 profile 不產生 `provider` 覆寫區塊。
25
+
26
+ > 💡 `opencode` profile 全部使用 `max` effort(variant 皆由 opencode 內建 catalog 提供):model 用 `gpt-6-luna#max`,heavy 用 `claude-haiku-5-5#max`,small 用 `deepseek-v4.1-flash#max`。
27
+ > `deepseek-v4.1-flash` 另有 low / high / max 可在 opencode 內用 `variant_cycle` 快捷鍵循環切換。
28
+ >
29
+ > 💡 `openai` profile 以 `#variant` 指定 effort(heavy `gpt-6.2-sol#medium`、model `gpt-6.2-sol#low`、small `gpt-6.1-terra#medium`;variant 由 opencode 內建 catalog 提供)。
30
+
31
+ > 💡 `copilot` profile 以 `#variant` 指定 effort(heavy `gpt-6.1-sol#medium`、model `claude-haiku-5.5#xhigh`、small `gpt-6-luna#max`;variant 由 opencode 內建 catalog 提供)。
25
32
 
26
33
  ### 用法
27
34
 
@@ -30,7 +37,7 @@ opencode plugin 是透過 `opencode.json` 的 `plugin` 區塊註冊的 npm 或
30
37
  ```bash
31
38
  moc
32
39
  moc openai
33
- moc google --list-agents
40
+ moc anthropic --list-agents
34
41
  ```
35
42
 
36
43
  `moc` 會對四個已知 profile 快取 provider-profile 產生出的 JSON override,位置在 `.opencode/cache/provider-profile/`。快取是否有效由以下條件決定:
@@ -7,37 +7,59 @@ const __dirname = path.dirname(__filename);
7
7
  const CONFIG_PATH = path.resolve(__dirname, "..", "opencode.json");
8
8
 
9
9
  const PROFILE_MAP = Object.freeze({
10
+ // opencode-go(全部 max effort):
11
+ // model (預設 implementation agents) → gpt-6-luna max
12
+ // heavy (SA/SD/reviewer/team-lead/architect) → claude-haiku-5-5 max
13
+ // small (verifier / title) → deepseek-v4.1-flash max
10
14
  opencode: Object.freeze({
11
- model: "opencode-go/deepseek-v4-pro",
12
- small_model: "opencode-go/deepseek-v4-flash",
13
- heavy_model: "opencode-go/deepseek-v4-pro",
14
- }),
15
- openai: Object.freeze({
16
- model: "openai/gpt-5.6",
17
- small_model: "openai/gpt-5.6-luna",
18
- heavy_model: "openai/gpt-5.6",
15
+ model: "opencode-go/gpt-6-luna",
16
+ model_variant: "max",
17
+ small_model: "opencode-go/deepseek-v4.1-flash",
18
+ small_variant: "max",
19
+ heavy_model: "opencode-go/claude-haiku-5-5",
20
+ heavy_variant: "max",
19
21
  reasoning: Object.freeze({
20
- "openai/gpt-5.6-luna": Object.freeze({
21
- default_effort: "low",
22
- efforts: ["high", "medium", "low"],
22
+ "opencode-go/deepseek-v4.1-flash": Object.freeze({
23
+ default_effort: "max",
24
+ efforts: ["low", "high", "max"],
23
25
  }),
24
26
  }),
25
27
  }),
26
- google: Object.freeze({
27
- model: "google/gemini-3.1-pro-preview-customtools",
28
- small_model: "google/gemini-3.5-flash",
29
- heavy_model: "google/gemini-3.1-pro-preview-customtools",
28
+ // OpenAI 系列:effort 透過 `provider/model#variant` 指定(opencode 內建 variant,不需 provider 覆寫)。
29
+ // heavy (SA/SD/reviewer/team-lead/architect) → gpt-6.2-sol medium (最高)
30
+ // model (預設 implementation agents) → gpt-6.2-sol low (其次)
31
+ // small (verifier / title) → gpt-6.1-terra medium(最低)
32
+ openai: Object.freeze({
33
+ model: "openai/gpt-6.2-sol",
34
+ model_variant: "low",
35
+ small_model: "openai/gpt-6.1-terra",
36
+ small_variant: "medium",
37
+ heavy_model: "openai/gpt-6.2-sol",
38
+ heavy_variant: "medium",
30
39
  }),
40
+ // GitHub Copilot:effort 透過 `provider/model#variant` 指定(opencode 內建 variant,不需 provider 覆寫)。
41
+ // heavy (SA/SD/reviewer/team-lead/architect) → gpt-6.1-sol medium (最高)
42
+ // model (預設 implementation agents) → claude-haiku-5.5 xhigh(其次)
43
+ // small (verifier / title) → gpt-6-luna max (最低)
31
44
  copilot: Object.freeze({
32
- model: "github-copilot/gpt-5.6-sol",
33
- small_model: "github-copilot/gpt-5.6-luna",
34
- heavy_model: "github-copilot/gpt-5.6-sol",
35
- reasoning: Object.freeze({
36
- "github-copilot/gpt-5.6-luna": Object.freeze({
37
- default_effort: "low",
38
- efforts: ["high", "medium", "low"],
39
- }),
40
- }),
45
+ model: "github-copilot/claude-haiku-5.5",
46
+ model_variant: "xhigh",
47
+ small_model: "github-copilot/gpt-6-luna",
48
+ small_variant: "max",
49
+ heavy_model: "github-copilot/gpt-6.1-sol",
50
+ heavy_variant: "medium",
51
+ }),
52
+ // Anthropic 系列:effort 透過 `provider/model#variant` 指定(opencode 內建 variant,不需 provider 覆寫)。
53
+ // heavy (SA/SD/reviewer/team-lead/architect) → sonnet-5-5 high (最高)
54
+ // model (預設 implementation agents) → haiku-5-5 max (其次)
55
+ // small (verifier / title) → haiku-5-5 high (最低)
56
+ anthropic: Object.freeze({
57
+ model: "anthropic/claude-haiku-5-5",
58
+ model_variant: "max",
59
+ small_model: "anthropic/claude-haiku-5-5",
60
+ small_variant: "high",
61
+ heavy_model: "anthropic/claude-sonnet-5-5",
62
+ heavy_variant: "high",
41
63
  }),
42
64
  });
43
65
 
@@ -74,21 +96,33 @@ function pickProfile(options = {}) {
74
96
  return activeProfile;
75
97
  }
76
98
 
99
+ /**
100
+ * Append an optional reasoning variant to a model ref using opencode's
101
+ * `provider/model#variant` syntax (e.g. `anthropic/claude-haiku-5-5#max`).
102
+ */
103
+ function withVariant(modelRef, variant) {
104
+ return variant ? `${modelRef}#${variant}` : modelRef;
105
+ }
106
+
77
107
  export default function resolveProfile(options = {}) {
78
108
  const activeProfile = pickProfile(options);
79
109
  const profile = PROFILE_MAP[activeProfile];
80
110
 
111
+ const model = withVariant(profile.model, profile.model_variant);
112
+ const smallModel = withVariant(profile.small_model, profile.small_variant);
113
+ const heavyModel = withVariant(profile.heavy_model, profile.heavy_variant);
114
+
81
115
  return {
82
- model: profile.model,
83
- small_model: profile.small_model,
116
+ model,
117
+ small_model: smallModel,
84
118
  agent_overrides: {
85
- "design-evidence": profile.heavy_model,
86
- "business-analyst": profile.heavy_model,
87
- "system-designer": profile.heavy_model,
88
- verifier: profile.small_model,
89
- "code-reviewer": profile.heavy_model,
90
- "team-lead": profile.heavy_model,
91
- architect: profile.heavy_model,
119
+ "design-evidence": heavyModel,
120
+ "business-analyst": heavyModel,
121
+ "system-designer": heavyModel,
122
+ verifier: smallModel,
123
+ "code-reviewer": heavyModel,
124
+ "team-lead": heavyModel,
125
+ architect: heavyModel,
92
126
  },
93
127
  };
94
128
  }
@@ -111,7 +145,8 @@ export function buildProfileConfig(options = {}) {
111
145
 
112
146
  /**
113
147
  * Build the `provider` config fragment for models that need reasoning-effort
114
- * variants (e.g. gpt-5.6-luna with high/medium/low effort switching).
148
+ * variants (e.g. gpt-5.6-luna with high/medium/low effort switching, or
149
+ * deepseek-v4.1-flash with low/high/max effort switching).
115
150
  *
116
151
  * Output shape (merged into opencode config via OPENCODE_CONFIG_CONTENT):
117
152
  * provider.openai.models["gpt-5.6-luna"] = {
@@ -74,16 +74,16 @@ for (const term of ["Business Requirements", "Technical Design", "frontend-contr
74
74
 
75
75
  const plugin = spawnSync(process.execPath, [path.join(root, "plugins/provider-profile.mjs")], {
76
76
  cwd: root,
77
- env: { ...process.env, OPENCODE_PROVIDER_PROFILE: "google" },
77
+ env: { ...process.env, OPENCODE_PROVIDER_PROFILE: "openai" },
78
78
  encoding: "utf8",
79
79
  });
80
80
  expect(plugin.status === 0, "provider profile plugin failed to execute");
81
81
  if (plugin.status === 0) {
82
82
  const output = JSON.parse(plugin.stdout);
83
83
  for (const agent of ["design-evidence", "business-analyst", "system-designer", "code-reviewer", "team-lead", "architect"]) {
84
- expect(output.agent?.[agent]?.model === "google/gemini-3.1-pro-preview-customtools", `google heavy model missing for ${agent}`);
84
+ expect(output.agent?.[agent]?.model === "openai/gpt-6.2-sol#medium", `openai heavy model missing for ${agent}`);
85
85
  }
86
- expect(output.agent?.verifier?.model === "google/gemini-3.5-flash", "google verifier model mismatch");
86
+ expect(output.agent?.verifier?.model === "openai/gpt-6.1-terra#medium", "openai verifier model mismatch");
87
87
  }
88
88
 
89
89
  if (process.exitCode) process.exit(process.exitCode);