opencode-multiagent 0.6.0 → 0.7.0

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.
Files changed (29) hide show
  1. package/CHANGELOG.md +19 -0
  2. package/agents/architect.md +104 -0
  3. package/agents/auditor.md +17 -9
  4. package/agents/brainstormer.md +9 -2
  5. package/agents/coder.md +19 -8
  6. package/agents/docmaster.md +31 -7
  7. package/agents/executor.md +46 -10
  8. package/agents/ops-coder.md +122 -0
  9. package/agents/planner.md +35 -9
  10. package/agents/reviewer.md +16 -8
  11. package/agents/scout.md +15 -6
  12. package/agents/sec-coder.md +20 -9
  13. package/agents/tester.md +107 -0
  14. package/agents/ui-coder.md +20 -9
  15. package/defaults/opencode-multiagent.json +51 -21
  16. package/dist/index.js +42 -5
  17. package/dist/opencode-multiagent/compiler.d.ts.map +1 -1
  18. package/dist/opencode-multiagent/task-manager.d.ts +1 -1
  19. package/dist/opencode-multiagent/task-manager.d.ts.map +1 -1
  20. package/dist/opencode-multiagent/tools.d.ts +7 -1
  21. package/dist/opencode-multiagent/tools.d.ts.map +1 -1
  22. package/docs/AGENTS.md +49 -2
  23. package/docs/agents.md +81 -29
  24. package/docs/agents.tr.md +81 -29
  25. package/docs/configuration.md +17 -14
  26. package/docs/configuration.tr.md +17 -14
  27. package/docs/usage-guide.md +61 -10
  28. package/docs/usage-guide.tr.md +61 -10
  29. package/package.json +1 -1
package/docs/agents.md CHANGED
@@ -4,12 +4,12 @@
4
4
 
5
5
  # Agent Reference
6
6
 
7
- `opencode-multiagent` ships 10 bundled agents:
7
+ `opencode-multiagent` ships 13 bundled agents:
8
8
 
9
9
  - 3 primary agents
10
- - 7 subagents
10
+ - 10 subagents
11
11
 
12
- Most bundled agents default to `temperature: 0`. The exception is `brainstormer` which uses `temperature: 0.3` for more creative exploration.
12
+ Each agent uses a model and temperature tuned for its role. The system is multi-provider: Anthropic (Opus/Sonnet), OpenAI (GPT-5.4), and OpenCode Go (MiniMax M2.5, Kimi K2.5, GLM-5).
13
13
 
14
14
  ## System Shape
15
15
 
@@ -27,17 +27,23 @@ brainstormer (pre-planning exploration + idea refinement)
27
27
  └── auditor (stress-test ideas)
28
28
 
29
29
  planner (triage + planning + inspection)
30
+ ├── architect (technical direction + module boundaries)
30
31
  ├── executor (execution coordination)
32
+ │ ├── architect (architecture decisions)
31
33
  │ ├── coder (standard coding) -> reviewer, scout
32
34
  │ ├── ui-coder (UI/UX coding) -> reviewer, scout
33
35
  │ ├── sec-coder (security coding) -> reviewer, scout
36
+ │ ├── tester (test writing) -> reviewer, scout
37
+ │ ├── ops-coder (CI/CD + infra) -> reviewer, scout
34
38
  │ ├── reviewer (quality gate)
35
39
  │ ├── scout (research)
36
40
  │ └── docmaster (.magent + docs + GitHub)
37
41
  ├── auditor (plan gap finder)
38
42
  ├── scout (research)
39
43
  ├── reviewer (evidence)
40
- └── docmaster (plan persistence)
44
+ ├── docmaster (plan persistence)
45
+ ├── tester (test planning)
46
+ └── ops-coder (infra planning)
41
47
  ```
42
48
 
43
49
  ## Primary Agents
@@ -45,8 +51,8 @@ planner (triage + planning + inspection)
45
51
  ### `brainstormer`
46
52
 
47
53
  - Mode: `primary`
48
- - Default model: `anthropic/claude-opus-4-6`
49
- - Temperature: `0.3`
54
+ - Default model: `anthropic/claude-sonnet-4-6`
55
+ - Temperature: `0.45`
50
56
  - Steps: `200`
51
57
  - Purpose: pre-planning exploration and idea refinement
52
58
  - Best for: exploring ideas with the user before planning begins, surfacing gaps and alternatives, challenging assumptions, and producing a planner-ready brief when the discussion concludes
@@ -54,15 +60,18 @@ planner (triage + planning + inspection)
54
60
  ### `planner`
55
61
 
56
62
  - Mode: `primary`
57
- - Default model: `anthropic/claude-opus-4-6`
63
+ - Default model: `openai/gpt-5.4`
64
+ - Temperature: `0.15`
58
65
  - Steps: `200`
66
+ - Options: `reasoning_effort: high`
59
67
  - Purpose: triage, planning, challenge, and inspection
60
68
  - Best for: deciding the lightest safe route, writing durable execution plans, challenging assumptions, inspecting repository state without implementing changes, and integrating results for the user
61
69
 
62
70
  ### `executor`
63
71
 
64
72
  - Mode: `primary`
65
- - Default model: `anthropic/claude-sonnet-4-6`
73
+ - Default model: `anthropic/claude-opus-4-6`
74
+ - Temperature: `0.10`
66
75
  - Steps: `200`
67
76
  - Purpose: execution orchestration
68
77
  - Best for: running an approved plan by dispatching bounded coding tasks, coordinating reviewer and scout passes, and enforcing quality gates
@@ -72,7 +81,8 @@ planner (triage + planning + inspection)
72
81
  ### `coder`
73
82
 
74
83
  - Mode: `subagent`
75
- - Default model: `anthropic/claude-sonnet-4-6`
84
+ - Default model: `opencodego/minimax-m2.5`
85
+ - Temperature: `0.20`
76
86
  - Steps: `40`
77
87
  - Purpose: standard coding work
78
88
  - Use when: the task is a bounded, non-UI, non-security-sensitive implementation slice
@@ -80,7 +90,8 @@ planner (triage + planning + inspection)
80
90
  ### `ui-coder`
81
91
 
82
92
  - Mode: `subagent`
83
- - Default model: `anthropic/claude-sonnet-4-6`
93
+ - Default model: `opencodego/kimi-k2.5`
94
+ - Temperature: `0.30`
84
95
  - Steps: `40`
85
96
  - Purpose: UI and UX work
86
97
  - Use when: the work is mainly components, view-local state, layout, accessibility, or responsive behavior
@@ -89,16 +100,48 @@ planner (triage + planning + inspection)
89
100
 
90
101
  - Mode: `subagent`
91
102
  - Default model: `anthropic/claude-opus-4-6`
103
+ - Temperature: `0.08`
92
104
  - Steps: `60`
93
105
  - Purpose: security-sensitive work
94
106
  - Use when: the change touches authentication, authorization, cryptography, secrets handling, or other high-risk runtime behavior
95
107
 
108
+ ### `tester`
109
+
110
+ - Mode: `subagent`
111
+ - Default model: `openai/gpt-5.4`
112
+ - Temperature: `0.18`
113
+ - Steps: `40`
114
+ - Options: `reasoning_effort: medium`
115
+ - Purpose: independent test writing
116
+ - Use when: the task requires test suites, regression tests, integration tests, test fixtures, or test infrastructure work independent from implementation
117
+
118
+ ### `ops-coder`
119
+
120
+ - Mode: `subagent`
121
+ - Default model: `anthropic/claude-sonnet-4-6`
122
+ - Temperature: `0.14`
123
+ - Steps: `40`
124
+ - Purpose: CI/CD and infrastructure work
125
+ - Use when: the work involves CI pipelines, Docker, deployment config, build systems, GitHub Actions, or infrastructure-as-code
126
+
127
+ ## Architecture Agent
128
+
129
+ ### `architect`
130
+
131
+ - Mode: `subagent`
132
+ - Default model: `anthropic/claude-opus-4-6`
133
+ - Temperature: `0.12`
134
+ - Steps: `40`
135
+ - Purpose: read-only architecture advisor
136
+ - Best for: technical direction choices, module boundary analysis, migration and refactor strategy, compatibility assessment, and producing implementation briefs for `executor`
137
+
96
138
  ## Quality Agents
97
139
 
98
140
  ### `reviewer`
99
141
 
100
142
  - Mode: `subagent`
101
- - Default model: `anthropic/claude-sonnet-4-6`
143
+ - Default model: `anthropic/claude-opus-4-6`
144
+ - Temperature: `0.05`
102
145
  - Steps: `30`
103
146
  - Purpose: code review, QA, and validation
104
147
  - Best for: reviewing code changes, running build and test evidence, and issuing an acceptance verdict before work is reported as done
@@ -106,7 +149,8 @@ planner (triage + planning + inspection)
106
149
  ### `auditor`
107
150
 
108
151
  - Mode: `subagent`
109
- - Default model: `anthropic/claude-opus-4-6`
152
+ - Default model: `anthropic/claude-sonnet-4-6`
153
+ - Temperature: `0.22`
110
154
  - Steps: `40`
111
155
  - Purpose: stubborn plan gap finder
112
156
  - Best for: finding missing steps, weak acceptance criteria, sequencing failures, and verification gaps before execution starts
@@ -116,7 +160,8 @@ planner (triage + planning + inspection)
116
160
  ### `scout`
117
161
 
118
162
  - Mode: `subagent`
119
- - Default model: `anthropic/claude-sonnet-4-6`
163
+ - Default model: `opencodego/glm-5`
164
+ - Temperature: `0.12`
120
165
  - Steps: `30`
121
166
  - Purpose: file discovery and external research
122
167
  - Best for: repository mapping, symbol tracing, codebase structure questions, git-aware local exploration, and live documentation or API lookup
@@ -126,8 +171,10 @@ planner (triage + planning + inspection)
126
171
  ### `docmaster`
127
172
 
128
173
  - Mode: `subagent`
129
- - Default model: `anthropic/claude-sonnet-4-6`
174
+ - Default model: `openai/gpt-5.4`
175
+ - Temperature: `0.10`
130
176
  - Steps: `30`
177
+ - Options: `reasoning_effort: medium`
131
178
  - Purpose: documentation writer and `.magent` artifact manager
132
179
  - Best for: writing or updating `README.md`, `AGENTS.md`, `CHANGELOG.md`, `CONTRIBUTING.md`, `docs/**`, `.magent/**` artifacts, and GitHub workflow files
133
180
 
@@ -136,25 +183,30 @@ planner (triage + planning + inspection)
136
183
  | Category | Agents |
137
184
  | --------------------- | ------------------------------------------ |
138
185
  | Primary orchestration | `brainstormer`, `planner`, `executor` |
139
- | Coding | `coder`, `ui-coder`, `sec-coder` |
186
+ | Coding | `coder`, `ui-coder`, `sec-coder`, `tester` |
187
+ | Infrastructure | `ops-coder` |
188
+ | Architecture | `architect` |
140
189
  | Quality | `reviewer`, `auditor` |
141
190
  | Research | `scout` |
142
191
  | Documentation | `docmaster` |
143
192
 
144
193
  ## Summary Table
145
194
 
146
- | Agent | Category | Mode | Model | Steps | Primary use |
147
- | -------------- | ------------- | ---------- | ----------------------------- | ----- | ---------------------------------------- |
148
- | `brainstormer` | Primary | `primary` | `anthropic/claude-opus-4-6` | `200` | Pre-planning exploration and ideation |
149
- | `planner` | Primary | `primary` | `anthropic/claude-opus-4-6` | `200` | Triage, planning, challenge, inspection |
150
- | `executor` | Primary | `primary` | `anthropic/claude-sonnet-4-6` | `200` | Execution orchestration |
151
- | `coder` | Coding | `subagent` | `anthropic/claude-sonnet-4-6` | `40` | Standard bounded coding work |
152
- | `ui-coder` | Coding | `subagent` | `anthropic/claude-sonnet-4-6` | `40` | UI and UX work |
153
- | `sec-coder` | Coding | `subagent` | `anthropic/claude-opus-4-6` | `60` | Security-sensitive implementation |
154
- | `reviewer` | Quality | `subagent` | `anthropic/claude-sonnet-4-6` | `30` | Code review, QA, and validation |
155
- | `auditor` | Quality | `subagent` | `anthropic/claude-opus-4-6` | `40` | Plan gap finding before execution |
156
- | `scout` | Research | `subagent` | `anthropic/claude-sonnet-4-6` | `30` | File discovery and external research |
157
- | `docmaster` | Documentation | `subagent` | `anthropic/claude-sonnet-4-6` | `30` | Docs and `.magent` writing |
195
+ | Agent | Category | Mode | Model | Temp | Steps | Primary use |
196
+ | -------------- | ------------- | ---------- | ----------------------------- | ------ | ----- | ---------------------------------------- |
197
+ | `brainstormer` | Primary | `primary` | `anthropic/claude-sonnet-4-6` | `0.45` | `200` | Pre-planning exploration and ideation |
198
+ | `planner` | Primary | `primary` | `openai/gpt-5.4` | `0.15` | `200` | Triage, planning, challenge, inspection |
199
+ | `executor` | Primary | `primary` | `anthropic/claude-opus-4-6` | `0.10` | `200` | Execution orchestration |
200
+ | `coder` | Coding | `subagent` | `opencodego/minimax-m2.5` | `0.20` | `40` | Standard bounded coding work |
201
+ | `ui-coder` | Coding | `subagent` | `opencodego/kimi-k2.5` | `0.30` | `40` | UI and UX work |
202
+ | `sec-coder` | Coding | `subagent` | `anthropic/claude-opus-4-6` | `0.08` | `60` | Security-sensitive implementation |
203
+ | `tester` | Testing | `subagent` | `openai/gpt-5.4` | `0.18` | `40` | Independent test writing |
204
+ | `ops-coder` | Infrastructure| `subagent` | `anthropic/claude-sonnet-4-6` | `0.14` | `40` | CI/CD, Docker, and infrastructure |
205
+ | `architect` | Architecture | `subagent` | `anthropic/claude-opus-4-6` | `0.12` | `40` | Technical direction and architecture |
206
+ | `reviewer` | Quality | `subagent` | `anthropic/claude-opus-4-6` | `0.05` | `30` | Code review, QA, and validation |
207
+ | `auditor` | Quality | `subagent` | `anthropic/claude-sonnet-4-6` | `0.22` | `40` | Plan gap finding before execution |
208
+ | `scout` | Research | `subagent` | `opencodego/glm-5` | `0.12` | `30` | File discovery and external research |
209
+ | `docmaster` | Documentation | `subagent` | `openai/gpt-5.4` | `0.10` | `30` | Docs and `.magent` writing |
158
210
 
159
211
  ## Model Parameters
160
212
 
@@ -182,9 +234,9 @@ Example: enabling thinking mode for the planner agent:
182
234
  {
183
235
  "agentSettings": {
184
236
  "planner": {
185
- "model": "anthropic/claude-opus-4-6",
237
+ "model": "openai/gpt-5.4",
186
238
  "options": {
187
- "thinking": { "type": "enabled", "budget_tokens": 10000 }
239
+ "reasoning_effort": "high"
188
240
  }
189
241
  }
190
242
  }
package/docs/agents.tr.md CHANGED
@@ -4,12 +4,12 @@
4
4
 
5
5
  # Ajan Referansı
6
6
 
7
- `opencode-multiagent`, 10 paketlenmiş ajan ile gelir:
7
+ `opencode-multiagent`, 13 paketlenmiş ajan ile gelir:
8
8
 
9
9
  - 3 ana ajan
10
- - 7 alt ajan
10
+ - 10 alt ajan
11
11
 
12
- Paketlenmiş ajanların çoğu varsayılan olarak `temperature: 0` kullanır. Tek istisna, daha yaratıcı keşif için `temperature: 0.3` kullanan `brainstormer` ajanıdır.
12
+ Her ajan, rolüne uygun model ve temperature ile yapılandırılmıştır. Sistem çoklu provider destekler: Anthropic (Opus/Sonnet), OpenAI (GPT-5.4) ve OpenCode Go (MiniMax M2.5, Kimi K2.5, GLM-5).
13
13
 
14
14
  ## Sistem Yapısı
15
15
 
@@ -28,10 +28,14 @@ kullanici
28
28
  -> auditor (fikirleri stres-test)
29
29
 
30
30
  -> planner (triage + planlama + inspeksiyon)
31
+ -> architect (teknik yön + modül sınırları)
31
32
  -> executor (yürütme koordinasyonu)
33
+ -> architect (mimari kararlar)
32
34
  -> coder (standart kodlama) -> reviewer, scout
33
35
  -> ui-coder (UI/UX kodlama) -> reviewer, scout
34
36
  -> sec-coder (güvenlik kodlama) -> reviewer, scout
37
+ -> tester (test yazımı) -> reviewer, scout
38
+ -> ops-coder (CI/CD + altyapı) -> reviewer, scout
35
39
  -> reviewer (kalite kapısı)
36
40
  -> scout (araştırma)
37
41
  -> docmaster (.magent + doküman + GitHub)
@@ -39,6 +43,8 @@ kullanici
39
43
  -> scout (araştırma)
40
44
  -> reviewer (kanıt)
41
45
  -> docmaster (plan kalıcılığı)
46
+ -> tester (test planlama)
47
+ -> ops-coder (altyapı planlama)
42
48
  ```
43
49
 
44
50
  ## Ana Ajanlar
@@ -46,8 +52,8 @@ kullanici
46
52
  ### `brainstormer`
47
53
 
48
54
  - Mod: `primary`
49
- - Varsayılan model: `anthropic/claude-opus-4-6`
50
- - Temperature: `0.3`
55
+ - Varsayılan model: `anthropic/claude-sonnet-4-6`
56
+ - Temperature: `0.45`
51
57
  - Steps: `200`
52
58
  - Amaç: planlama öncesi keşif ve fikir olgunlaştırma
53
59
  - En uygun kullanım: planlama başlamadan önce kullanıcıyla birlikte fikirleri keşfetmek, boşlukları ve alternatifleri ortaya çıkarmak, varsayımları sorgulamak ve tartışma tamamlandığında planner'a hazır bir brief üretmek
@@ -55,15 +61,18 @@ kullanici
55
61
  ### `planner`
56
62
 
57
63
  - Mod: `primary`
58
- - Varsayılan model: `anthropic/claude-opus-4-6`
64
+ - Varsayılan model: `openai/gpt-5.4`
65
+ - Temperature: `0.15`
59
66
  - Steps: `200`
67
+ - Options: `reasoning_effort: high`
60
68
  - Amaç: triage, planlama, challenge ve inspeksiyon
61
69
  - En uygun kullanım: route seçimi, iş delegasyonu, büyük işlerde kalıcı ve yürütmeye hazır plan üretmek, repo durumunu incelemek
62
70
 
63
71
  ### `executor`
64
72
 
65
73
  - Mod: `primary`
66
- - Varsayılan model: `anthropic/claude-sonnet-4-6`
74
+ - Varsayılan model: `anthropic/claude-opus-4-6`
75
+ - Temperature: `0.10`
67
76
  - Steps: `200`
68
77
  - Amaç: yürütme orkestrasyonu
69
78
  - En uygun kullanım: onaylanmış planı bounded uzman görevlerine bölüp doğrulama katmanları ile yürütmek, task board güncellemeleri
@@ -73,7 +82,8 @@ kullanici
73
82
  ### `coder`
74
83
 
75
84
  - Mod: `subagent`
76
- - Varsayılan model: `anthropic/claude-sonnet-4-6`
85
+ - Varsayılan model: `opencodego/minimax-m2.5`
86
+ - Temperature: `0.20`
77
87
  - Steps: `40`
78
88
  - Amaç: standart kodlama
79
89
  - En uygun kullanım: UI dışı, orta zorlukta, varsayılan kodlama yolu gerektiğinde
@@ -81,7 +91,8 @@ kullanici
81
91
  ### `ui-coder`
82
92
 
83
93
  - Mod: `subagent`
84
- - Varsayılan model: `anthropic/claude-sonnet-4-6`
94
+ - Varsayılan model: `opencodego/kimi-k2.5`
95
+ - Temperature: `0.30`
85
96
  - Steps: `40`
86
97
  - Amaç: UI/UX
87
98
  - En uygun kullanım: iş daha çok component, view-local state, layout, accessibility veya responsive davranış ile ilgiliyse
@@ -90,16 +101,48 @@ kullanici
90
101
 
91
102
  - Mod: `subagent`
92
103
  - Varsayılan model: `anthropic/claude-opus-4-6`
104
+ - Temperature: `0.08`
93
105
  - Steps: `60`
94
106
  - Amaç: güvenlik hassas kodlama
95
107
  - En uygun kullanım: değişiklik hassas runtime davranışını, kimlik doğrulama veya güvenlik mantığını etkiliyorsa
96
108
 
109
+ ### `tester`
110
+
111
+ - Mod: `subagent`
112
+ - Varsayılan model: `openai/gpt-5.4`
113
+ - Temperature: `0.18`
114
+ - Steps: `40`
115
+ - Options: `reasoning_effort: medium`
116
+ - Amaç: bağımsız test yazımı
117
+ - En uygun kullanım: test suite'leri, regresyon testleri, entegrasyon testleri, test fixture'ları ve implementasyondan bağımsız test altyapısı çalışmaları
118
+
119
+ ### `ops-coder`
120
+
121
+ - Mod: `subagent`
122
+ - Varsayılan model: `anthropic/claude-sonnet-4-6`
123
+ - Temperature: `0.14`
124
+ - Steps: `40`
125
+ - Amaç: CI/CD ve altyapı
126
+ - En uygun kullanım: CI pipeline'ları, Docker, deployment konfigürasyonu, build sistemleri, GitHub Actions veya altyapı-kod (IaC) işleri
127
+
128
+ ## Mimari Ajanı
129
+
130
+ ### `architect`
131
+
132
+ - Mod: `subagent`
133
+ - Varsayılan model: `anthropic/claude-opus-4-6`
134
+ - Temperature: `0.12`
135
+ - Steps: `40`
136
+ - Amaç: salt-okunur mimari danışman
137
+ - En uygun kullanım: teknik yön seçimi, modül sınırı analizi, migration ve refactor stratejisi, uyumluluk değerlendirmesi ve `executor`'a implementation brief üretmek
138
+
97
139
  ## Kalite Ajanları
98
140
 
99
141
  ### `reviewer`
100
142
 
101
143
  - Mod: `subagent`
102
- - Varsayılan model: `anthropic/claude-sonnet-4-6`
144
+ - Varsayılan model: `anthropic/claude-opus-4-6`
145
+ - Temperature: `0.05`
103
146
  - Steps: `30`
104
147
  - Amaç: kod inceleme, QA, doğrulama
105
148
  - En uygun kullanım: implementasyon tamamlandıktan sonra değişiklik yüzeyini incelemek, test/build/lint ile kanıt üretmek ve son kabul kararı vermek
@@ -107,7 +150,8 @@ kullanici
107
150
  ### `auditor`
108
151
 
109
152
  - Mod: `subagent`
110
- - Varsayılan model: `anthropic/claude-opus-4-6`
153
+ - Varsayılan model: `anthropic/claude-sonnet-4-6`
154
+ - Temperature: `0.22`
111
155
  - Steps: `40`
112
156
  - Amaç: plan açığı bulucu
113
157
  - En uygun kullanım: iş yürütülmeden önce eksik adımları, zayıf kabul kriterlerini, sıralama sorunlarını ve verification boşluklarını yakalamak
@@ -117,7 +161,8 @@ kullanici
117
161
  ### `scout`
118
162
 
119
163
  - Mod: `subagent`
120
- - Varsayılan model: `anthropic/claude-sonnet-4-6`
164
+ - Varsayılan model: `opencodego/glm-5`
165
+ - Temperature: `0.12`
121
166
  - Steps: `30`
122
167
  - Amaç: dosya keşfi, araştırma
123
168
  - En uygun kullanım: dosya bulma, sembol takibi, kod tabanı yapısını anlama ve yerel keşif
@@ -127,8 +172,10 @@ kullanici
127
172
  ### `docmaster`
128
173
 
129
174
  - Mod: `subagent`
130
- - Varsayılan model: `anthropic/claude-sonnet-4-6`
175
+ - Varsayılan model: `openai/gpt-5.4`
176
+ - Temperature: `0.10`
131
177
  - Steps: `30`
178
+ - Options: `reasoning_effort: medium`
132
179
  - Amaç: dokümantasyon, `.magent`, GitHub
133
180
  - En uygun kullanım: `README.md`, `AGENTS.md`, `CHANGELOG.md`, `CONTRIBUTING.md`, `docs/**` ve `.magent/**` güncellemeleri, GitHub işlemleri
134
181
 
@@ -137,25 +184,30 @@ kullanici
137
184
  | Kategori | Ajanlar |
138
185
  | ------------------- | ---------------------------------------------- |
139
186
  | Ana orkestrasyon | `brainstormer`, `planner`, `executor` |
140
- | Kodlama ajanları | `coder`, `ui-coder`, `sec-coder` |
187
+ | Kodlama ajanları | `coder`, `ui-coder`, `sec-coder`, `tester` |
188
+ | Altyapı | `ops-coder` |
189
+ | Mimari | `architect` |
141
190
  | Kalite | `reviewer`, `auditor` |
142
191
  | Araştırma | `scout` |
143
192
  | Dokümantasyon | `docmaster` |
144
193
 
145
194
  ## Özet Tablo
146
195
 
147
- | Agent | Kategori | Mod | Model | Steps | Ana kullanım |
148
- | -------------- | ------------- | ---------- | ----------------------------- | ----- | ------------------------------------------------------- |
149
- | `brainstormer` | Primary | `primary` | `anthropic/claude-opus-4-6` | `200` | Planlama öncesi keşif ve fikir olgunlaştırma |
150
- | `planner` | Primary | `primary` | `anthropic/claude-opus-4-6` | `200` | Triage, planlama, challenge, inspeksiyon |
151
- | `executor` | Primary | `primary` | `anthropic/claude-sonnet-4-6` | `200` | Yürütme orkestrasyonu |
152
- | `coder` | Coding | `subagent` | `anthropic/claude-sonnet-4-6` | `40` | Standart kodlama |
153
- | `ui-coder` | Coding | `subagent` | `anthropic/claude-sonnet-4-6` | `40` | UI/UX kodlama |
154
- | `sec-coder` | Coding | `subagent` | `anthropic/claude-opus-4-6` | `60` | Güvenlik hassas kodlama |
155
- | `reviewer` | Quality | `subagent` | `anthropic/claude-sonnet-4-6` | `30` | Kod inceleme, QA, doğrulama |
156
- | `auditor` | Quality | `subagent` | `anthropic/claude-opus-4-6` | `40` | Plan açığı bulma ve denetim |
157
- | `scout` | Research | `subagent` | `anthropic/claude-sonnet-4-6` | `30` | Dosya keşfi ve araştırma |
158
- | `docmaster` | Documentation | `subagent` | `anthropic/claude-sonnet-4-6` | `30` | Doküman, `.magent` yazımı ve GitHub |
196
+ | Agent | Kategori | Mod | Model | Temp | Steps | Ana kullanım |
197
+ | -------------- | ------------- | ---------- | ----------------------------- | ------ | ----- | ------------------------------------------------------- |
198
+ | `brainstormer` | Primary | `primary` | `anthropic/claude-sonnet-4-6` | `0.45` | `200` | Planlama öncesi keşif ve fikir olgunlaştırma |
199
+ | `planner` | Primary | `primary` | `openai/gpt-5.4` | `0.15` | `200` | Triage, planlama, challenge, inspeksiyon |
200
+ | `executor` | Primary | `primary` | `anthropic/claude-opus-4-6` | `0.10` | `200` | Yürütme orkestrasyonu |
201
+ | `coder` | Coding | `subagent` | `opencodego/minimax-m2.5` | `0.20` | `40` | Standart kodlama |
202
+ | `ui-coder` | Coding | `subagent` | `opencodego/kimi-k2.5` | `0.30` | `40` | UI/UX kodlama |
203
+ | `sec-coder` | Coding | `subagent` | `anthropic/claude-opus-4-6` | `0.08` | `60` | Güvenlik hassas kodlama |
204
+ | `tester` | Testing | `subagent` | `openai/gpt-5.4` | `0.18` | `40` | Bağımsız test yazımı |
205
+ | `ops-coder` | Infrastructure| `subagent` | `anthropic/claude-sonnet-4-6` | `0.14` | `40` | CI/CD, Docker ve altyapı |
206
+ | `architect` | Architecture | `subagent` | `anthropic/claude-opus-4-6` | `0.12` | `40` | Teknik yön ve mimari kararlar |
207
+ | `reviewer` | Quality | `subagent` | `anthropic/claude-opus-4-6` | `0.05` | `30` | Kod inceleme, QA, doğrulama |
208
+ | `auditor` | Quality | `subagent` | `anthropic/claude-sonnet-4-6` | `0.22` | `40` | Plan açığı bulma ve denetim |
209
+ | `scout` | Research | `subagent` | `opencodego/glm-5` | `0.12` | `30` | Dosya keşfi ve araştırma |
210
+ | `docmaster` | Documentation | `subagent` | `openai/gpt-5.4` | `0.10` | `30` | Doküman, `.magent` yazımı ve GitHub |
159
211
 
160
212
  ## Model Parametreleri
161
213
 
@@ -177,15 +229,15 @@ Bu parametreler iki katmanda uygulanır:
177
229
  1. **Config zamanı** — `compileAgents` bunları agent config'e yazar
178
230
  2. **Runtime** — `chat.params` hook'u `top_p` ve `options`'ı doğrudan LLM çağrı parametrelerine enjekte eder
179
231
 
180
- Örnek: planner ajanında thinking modunu açma:
232
+ Örnek: planner ajanında reasoning effort ayarlama:
181
233
 
182
234
  ```json
183
235
  {
184
236
  "agentSettings": {
185
237
  "planner": {
186
- "model": "anthropic/claude-opus-4-6",
238
+ "model": "openai/gpt-5.4",
187
239
  "options": {
188
- "thinking": { "type": "enabled", "budget_tokens": 10000 }
240
+ "reasoning_effort": "high"
189
241
  }
190
242
  }
191
243
  }
package/docs/configuration.md CHANGED
@@ -137,20 +137,23 @@ This section centralizes per-agent defaults for:
137
137
  - `retry` — maximum retry attempts on failure. `0` = no retry.
138
138
  - `escalation_model` — model to escalate to when retries are exhausted. Empty string = disabled.
139
139
 
140
- Most bundled agents default to `temperature: 0`. The exception is `brainstormer` which uses `temperature: 0.3` for more creative exploration.
141
-
142
- | Agent | Category | Default model | Steps | top_p | options | timeout_ms | retry | escalation_model |
143
- | -------------- | ------------- | ----------------------------- | ----- | ----- | ------- | ---------- | ----- | ---------------- |
144
- | `brainstormer` | Primary | `anthropic/claude-opus-4-6` | `200` | `1` | `{}` | `0` | `0` | |
145
- | `planner` | Primary | `anthropic/claude-opus-4-6` | `200` | `1` | `{}` | `0` | `0` | |
146
- | `executor` | Primary | `anthropic/claude-sonnet-4-6` | `200` | `1` | `{}` | `0` | `0` | |
147
- | `coder` | Coding | `anthropic/claude-sonnet-4-6` | `40` | `1` | `{}` | `0` | `0` | |
148
- | `ui-coder` | Coding | `anthropic/claude-sonnet-4-6` | `40` | `1` | `{}` | `0` | `0` | |
149
- | `sec-coder` | Coding | `anthropic/claude-opus-4-6` | `60` | `1` | `{}` | `0` | `0` | |
150
- | `reviewer` | Quality | `anthropic/claude-sonnet-4-6` | `30` | `1` | `{}` | `0` | `0` | |
151
- | `auditor` | Quality | `anthropic/claude-opus-4-6` | `40` | `1` | `{}` | `0` | `0` | |
152
- | `scout` | Research | `anthropic/claude-sonnet-4-6` | `30` | `1` | `{}` | `0` | `0` | |
153
- | `docmaster` | Documentation | `anthropic/claude-sonnet-4-6` | `30` | `1` | `{}` | `0` | `0` | |
140
+ Each agent uses a model and temperature tuned for its role. The system is multi-provider.
141
+
142
+ | Agent | Category | Default model | Temp | Steps | options | timeout_ms | retry | escalation_model |
143
+ | -------------- | ------------- | ----------------------------- | ------ | ----- | ------------------------------ | ---------- | ----- | ---------------- |
144
+ | `brainstormer` | Primary | `anthropic/claude-sonnet-4-6` | `0.45` | `200` | `{}` | `0` | `0` | |
145
+ | `planner` | Primary | `openai/gpt-5.4` | `0.15` | `200` | `{ reasoning_effort: "high" }` | `0` | `0` | |
146
+ | `executor` | Primary | `anthropic/claude-opus-4-6` | `0.10` | `200` | `{}` | `0` | `0` | |
147
+ | `coder` | Coding | `opencodego/minimax-m2.5` | `0.20` | `40` | `{}` | `0` | `0` | |
148
+ | `ui-coder` | Coding | `opencodego/kimi-k2.5` | `0.30` | `40` | `{}` | `0` | `0` | |
149
+ | `sec-coder` | Coding | `anthropic/claude-opus-4-6` | `0.08` | `60` | `{}` | `0` | `0` | |
150
+ | `tester` | Testing | `openai/gpt-5.4` | `0.18` | `40` | `{ reasoning_effort: "medium" }`| `0` | `0` | |
151
+ | `ops-coder` | Infrastructure| `anthropic/claude-sonnet-4-6` | `0.14` | `40` | `{}` | `0` | `0` | |
152
+ | `architect` | Architecture | `anthropic/claude-opus-4-6` | `0.12` | `40` | `{}` | `0` | `0` | |
153
+ | `reviewer` | Quality | `anthropic/claude-opus-4-6` | `0.05` | `30` | `{}` | `0` | `0` | |
154
+ | `auditor` | Quality | `anthropic/claude-sonnet-4-6` | `0.22` | `40` | `{}` | `0` | `0` | |
155
+ | `scout` | Research | `opencodego/glm-5` | `0.12` | `30` | `{}` | `0` | `0` | |
156
+ | `docmaster` | Documentation | `openai/gpt-5.4` | `0.10` | `30` | `{ reasoning_effort: "medium" }`| `0` | `0` | |
154
157
 
155
158
  ## `mcpDefaults`
156
159
 
package/docs/configuration.tr.md CHANGED
@@ -137,20 +137,23 @@ Bu bölüm, agent bazında şu varsayılanları merkezi olarak yönetir:
137
137
  - `retry` — başarısızlık durumunda maksimum yeniden deneme sayısı. `0` = yeniden deneme yok.
138
138
  - `escalation_model` — yeniden denemeler tükendiğinde yükseltilecek model. Boş string = devre dışı.
139
139
 
140
- Paketlenmiş ajanların çoğu varsayılan olarak `temperature: 0` kullanır. Tek istisna, daha yaratıcı keşif için `temperature: 0.3` kullanan `brainstormer` ajanıdır.
141
-
142
- | Agent | Kategori | Varsayılan model | Steps | top_p | options | timeout_ms | retry | escalation_model |
143
- | -------------- | ------------- | ----------------------------- | ----- | ----- | ------- | ---------- | ----- | ---------------- |
144
- | `brainstormer` | Primary | `anthropic/claude-opus-4-6` | `200` | `1` | `{}` | `0` | `0` | |
145
- | `planner` | Primary | `anthropic/claude-opus-4-6` | `200` | `1` | `{}` | `0` | `0` | |
146
- | `executor` | Primary | `anthropic/claude-sonnet-4-6` | `200` | `1` | `{}` | `0` | `0` | |
147
- | `coder` | Coding | `anthropic/claude-sonnet-4-6` | `40` | `1` | `{}` | `0` | `0` | |
148
- | `ui-coder` | Coding | `anthropic/claude-sonnet-4-6` | `40` | `1` | `{}` | `0` | `0` | |
149
- | `sec-coder` | Coding | `anthropic/claude-opus-4-6` | `60` | `1` | `{}` | `0` | `0` | |
150
- | `reviewer` | Quality | `anthropic/claude-sonnet-4-6` | `30` | `1` | `{}` | `0` | `0` | |
151
- | `auditor` | Quality | `anthropic/claude-opus-4-6` | `40` | `1` | `{}` | `0` | `0` | |
152
- | `scout` | Research | `anthropic/claude-sonnet-4-6` | `30` | `1` | `{}` | `0` | `0` | |
153
- | `docmaster` | Documentation | `anthropic/claude-sonnet-4-6` | `30` | `1` | `{}` | `0` | `0` | |
140
+ Her ajan, rolüne uygun model ve temperature ile yapılandırılmıştır. Sistem çoklu provider destekler.
141
+
142
+ | Agent | Kategori | Varsayılan model | Temp | Steps | options | timeout_ms | retry | escalation_model |
143
+ | -------------- | ------------- | ----------------------------- | ------ | ----- | ------------------------------ | ---------- | ----- | ---------------- |
144
+ | `brainstormer` | Primary | `anthropic/claude-sonnet-4-6` | `0.45` | `200` | `{}` | `0` | `0` | |
145
+ | `planner` | Primary | `openai/gpt-5.4` | `0.15` | `200` | `{ reasoning_effort: "high" }` | `0` | `0` | |
146
+ | `executor` | Primary | `anthropic/claude-opus-4-6` | `0.10` | `200` | `{}` | `0` | `0` | |
147
+ | `coder` | Coding | `opencodego/minimax-m2.5` | `0.20` | `40` | `{}` | `0` | `0` | |
148
+ | `ui-coder` | Coding | `opencodego/kimi-k2.5` | `0.30` | `40` | `{}` | `0` | `0` | |
149
+ | `sec-coder` | Coding | `anthropic/claude-opus-4-6` | `0.08` | `60` | `{}` | `0` | `0` | |
150
+ | `tester` | Testing | `openai/gpt-5.4` | `0.18` | `40` | `{ reasoning_effort: "medium" }`| `0` | `0` | |
151
+ | `ops-coder` | Infrastructure| `anthropic/claude-sonnet-4-6` | `0.14` | `40` | `{}` | `0` | `0` | |
152
+ | `architect` | Architecture | `anthropic/claude-opus-4-6` | `0.12` | `40` | `{}` | `0` | `0` | |
153
+ | `reviewer` | Quality | `anthropic/claude-opus-4-6` | `0.05` | `30` | `{}` | `0` | `0` | |
154
+ | `auditor` | Quality | `anthropic/claude-sonnet-4-6` | `0.22` | `40` | `{}` | `0` | `0` | |
155
+ | `scout` | Research | `opencodego/glm-5` | `0.12` | `30` | `{}` | `0` | `0` | |
156
+ | `docmaster` | Documentation | `openai/gpt-5.4` | `0.10` | `30` | `{ reasoning_effort: "medium" }`| `0` | `0` | |
154
157
 
155
158
  ## `mcpDefaults`
156
159