@polderlabs/bizar 3.12.0 → 3.12.1

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
@@ -303,7 +303,7 @@ After installation, run `/connect` in opencode to add API keys:
303
303
  | Provider | Models | Auth |
304
304
  |---|---|---|
305
305
  | **OpenCode Zen** | `opencode/deepseek-v4-flash-free` | Free API key from [opencode.ai](https://opencode.ai) — create account, get key, no charges |
306
- | **OpenRouter** | `openrouter/minimax-m2.7`, `openrouter/minimax-m3` | API key from [openrouter.ai](https://openrouter.ai) |
306
+ | **OpenRouter** | `openrouter/minimax/minimax-m2.7`, `openrouter/minimax/minimax-m3` | API key from [openrouter.ai](https://openrouter.ai) |
307
307
  | **OpenAI** | `openai/gpt-5.5` | ChatGPT subscription (OAuth) |
308
308
 
309
309
  Then run `/models` to verify connectivity.
package/cli/audit.mjs CHANGED
@@ -78,8 +78,8 @@ export async function runAudit() {
78
78
  const model = modelMatch[1].trim();
79
79
  const validModels = [
80
80
  'opencode/deepseek-v4-flash-free',
81
- 'openrouter/minimax-m2.7',
82
- 'openrouter/minimax-m3',
81
+ 'openrouter/minimax/minimax-m2.7',
82
+ 'openrouter/minimax/minimax-m3',
83
83
  'openai/gpt-5.5',
84
84
  ];
85
85
  if (!validModels.includes(model)) {
package/config/AGENTS.md CHANGED
@@ -103,6 +103,7 @@ BizarHarness ships always-on coding rules organized by language and concern. All
103
103
  | `rules/git.md` | Git and commit conventions |
104
104
  | `rules/testing.md` | Test methodology and coverage |
105
105
  | `rules/thinking.md` | All agents — concise thinking behavior |
106
+ | `rules/uncertainty.md` | All agents — stop-and-research rule; reach for `websearch` / `webfetch` when uncertain or stuck; self-catch loops before the plugin loop-guard fires |
106
107
 
107
108
  ### How to Use
108
109
 
@@ -115,6 +116,10 @@ BizarHarness ships always-on coding rules organized by language and concern. All
115
116
 
116
117
  For agents with `reasoning: true` + `variant: "high"`, follow `rules/thinking.md` strictly. Cap reasoning at 2–4 sentences. No informal self-talk, no "what if" loops, no mid-thought self-correction. Think once, decide, act.
117
118
 
119
+ ### Research-Loop Rule
120
+
121
+ Follow `rules/uncertainty.md` strictly. When uncertain or stuck, the next move is a research tool call (`websearch` for outside-the-repo facts, `webfetch` for official docs, `semble search` for codebase patterns, `hindsight_recall` for project memory) — not a third variation of the same edit. If you catch yourself about to retry the same failed command with slightly different arguments, stop and search first. The plugin's loop-guard (`loopThresholdWarn: 5`) is the safety net; self-correct at attempt 2.
122
+
118
123
  ---
119
124
 
120
125
  ## Model Routing & Agents
@@ -166,25 +171,25 @@ Odin (`@odin`) is the All-Father and primary/default agent. He analyzes each req
166
171
 
167
172
  ### Hermod
168
173
 
169
- - **Model**: `openrouter/minimax-m2.7` (via OpenRouter)
174
+ - **Model**: `openrouter/minimax/minimax-m2.7` (via OpenRouter)
170
175
  - **Use for**: Git and GitHub operations — commit, push, merge, PRs, branches, conflict resolution. The swift messenger.
171
176
  - **Cost**: $0.30/M input, $1.20/M output
172
177
 
173
178
  ### Thor
174
179
 
175
- - **Model**: `openrouter/minimax-m2.7` (via OpenRouter)
180
+ - **Model**: `openrouter/minimax/minimax-m2.7` (via OpenRouter)
176
181
  - **Use for**: Moderate complexity features, debugging, code review, refactoring
177
182
  - **Cost**: $0.30/M input, $1.20/M output — cheaper than Tyr, more capable than Heimdall
178
183
 
179
184
  ### Baldr
180
185
 
181
- - **Model**: `openrouter/minimax-m2.7` (via OpenRouter)
186
+ - **Model**: `openrouter/minimax/minimax-m2.7` (via OpenRouter)
182
187
  - **Use for**: Design system creation, DESIGN.md, visual audit, usability planning. Creates design plans — does not implement.
183
188
  - **Cost**: $0.30/M input, $1.20/M output
184
189
 
185
190
  ### Tyr
186
191
 
187
- - **Model**: `openrouter/minimax-m3` (via OpenRouter)
192
+ - **Model**: `openrouter/minimax/minimax-m3` (via OpenRouter)
188
193
  - **Use for**: Highest complexity implementation, debugging, architecture, multi-step engineering
189
194
  - **Cost**: Higher — reserved for the hardest problems
190
195
 
@@ -196,7 +201,7 @@ Odin (`@odin`) is the All-Father and primary/default agent. He analyzes each req
196
201
 
197
202
  ### Forseti
198
203
 
199
- - **Model**: `openrouter/minimax-m3` (via OpenRouter, audit-only, no edit permissions)
204
+ - **Model**: `openrouter/minimax/minimax-m3` (via OpenRouter, audit-only, no edit permissions)
200
205
  - **Use for**: Adversarial plan review — audits completeness, correctness, consistency, feasibility, security
201
206
  - **Always runs before any Tier 4 or Tier 5 implementation begins**
202
207
 
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Baldr — UI/UX design system specialist. Creates DESIGN.md files using Google's design.md standard (alpha). Focuses on visual consistency, usability, accessibility, and design tokens.
3
3
  mode: subagent
4
- model: openrouter/minimax-m2.7
4
+ model: openrouter/minimax/minimax-m2.7
5
5
  color: "#ec4899"
6
6
  permission:
7
7
  read: allow
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Forseti — Audits, criticizes, and corrects implementation plans before execution using MiniMax M3. No write permissions — review only.
3
3
  mode: subagent
4
- model: openrouter/minimax-m3
4
+ model: openrouter/minimax/minimax-m3
5
5
  color: "#ef4444"
6
6
  permission:
7
7
  read: allow
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Hermod — Git and GitHub operations specialist using MiniMax M2.7. Branching, commits, PRs, merge/rebase, conflict resolution, CI/CD, releases, gh CLI.
3
3
  mode: subagent
4
- model: openrouter/minimax-m2.7
4
+ model: openrouter/minimax/minimax-m2.7
5
5
  color: "#06b6d4"
6
6
  permission:
7
7
  read: allow
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Odin — Pure router that delegates all work to subagents. Routes across Frigg (DeepSeek/Q&A), Vör (DeepSeek/clarify), Mimir (DeepSeek/research), Heimdall (DeepSeek/simple), Hermod (M2.7/git), Thor (M2.7/mid), Baldr (M2.7/design), Tyr (M3/top), Vidarr (GPT-5.5/ultra), Forseti (verifier/M3).
3
3
  mode: primary
4
- model: openrouter/minimax-m3
4
+ model: openrouter/minimax/minimax-m3
5
5
  color: "#6366f1"
6
6
  permission:
7
7
  task: allow
@@ -306,7 +306,7 @@ Call `bizar_spawn_background` with:
306
306
 
307
307
  - `agent`: the agent name (e.g., "mimir", "thor", "tyr")
308
308
  - `prompt`: what to do (specific, with context)
309
- - `model`: optional, `"<providerID>/<modelID>"` format (e.g., `"openrouter/minimax-m3"`)
309
+ - `model`: optional, `"<providerID>/<modelID>"` format (e.g., `"openrouter/minimax/minimax-m3"`)
310
310
  - `timeoutMs`: optional, default 5 min, max 30 min, min 1s
311
311
 
312
312
  You get an `instanceId` back immediately.
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Quick (quick) — fast single-shot tasks. No delegation, no parallel streams. Use for small edits, mechanical changes, one-shot questions. Routes to no one.
3
3
  mode: primary
4
- model: openrouter/minimax-m2.7
4
+ model: openrouter/minimax/minimax-m2.7
5
5
  color: "#22d3ee"
6
6
  permission:
7
7
  read: allow
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Thor — Handles medium-complexity tasks using MiniMax M2.7 from OpenRouter. Strong and reliable, cheaper than Tyr but more capable than Heimdall.
3
3
  mode: subagent
4
- model: openrouter/minimax-m2.7
4
+ model: openrouter/minimax/minimax-m2.7
5
5
  color: "#a855f7"
6
6
  permission:
7
7
  read: allow
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  description: Tyr — Handles the most complex implementation, debugging, and architectural work using MiniMax M3 via OpenRouter. Unmatched wisdom for the hardest problems.
3
3
  mode: subagent
4
- model: openrouter/minimax-m3
4
+ model: openrouter/minimax/minimax-m3
5
5
  color: "#f59e0b"
6
6
  permission:
7
7
  read: allow
@@ -1,23 +1,10 @@
1
1
  {
2
2
  "$schema": "https://opencode.ai/config.json",
3
- "model": "openrouter/minimax-m3",
4
- "small_model": "openrouter/minimax-m2.7",
3
+ "model": "openrouter/minimax/minimax-m3",
4
+ "small_model": "openrouter/minimax/minimax-m2.7",
5
5
  "default_agent": "odin",
6
6
  "permission": "allow",
7
7
  "snapshot": false,
8
- "provider": {
9
- "openrouter": {
10
- "npm": "@openrouter/ai-sdk-provider",
11
- "name": "OpenRouter",
12
- "options": {
13
- "apiKey": "{env:OPENROUTER_API_KEY}"
14
- },
15
- "models": {
16
- "minimax-m2.7": {},
17
- "minimax-m3": {}
18
- }
19
- }
20
- },
21
8
  "mcp": {
22
9
  "supabase": {
23
10
  "type": "remote",
@@ -66,7 +53,7 @@
66
53
  "odin": {
67
54
  "description": "Odin — Pure router that delegates all work to subagents.",
68
55
  "mode": "primary",
69
- "model": "openrouter/minimax-m3",
56
+ "model": "openrouter/minimax/minimax-m3",
70
57
  "color": "#6366f1",
71
58
  "permission": {
72
59
  "task": "allow",
@@ -80,7 +67,7 @@
80
67
  "vor": {
81
68
  "description": "Vör — The Questioning One.",
82
69
  "mode": "subagent",
83
- "model": "openrouter/minimax-m2.7",
70
+ "model": "openrouter/minimax/minimax-m2.7",
84
71
  "color": "#8b5cf6",
85
72
  "permission": {
86
73
  "read": "allow",
@@ -93,7 +80,7 @@
93
80
  "frigg": {
94
81
  "description": "Frigg — All-knowing Q&A agent.",
95
82
  "mode": "primary",
96
- "model": "openrouter/minimax-m2.7",
83
+ "model": "openrouter/minimax/minimax-m2.7",
97
84
  "color": "#06b6d4",
98
85
  "permission": {
99
86
  "read": "allow",
@@ -111,7 +98,7 @@
111
98
  "quick": {
112
99
  "description": "Quick — Fast single-shot tasks.",
113
100
  "mode": "primary",
114
- "model": "openrouter/minimax-m2.7",
101
+ "model": "openrouter/minimax/minimax-m2.7",
115
102
  "color": "#22d3ee",
116
103
  "permission": {
117
104
  "read": "allow",
@@ -130,7 +117,7 @@
130
117
  "mimir": {
131
118
  "description": "Mimir — Research and codebase exploration agent.",
132
119
  "mode": "subagent",
133
- "model": "openrouter/minimax-m2.7",
120
+ "model": "openrouter/minimax/minimax-m2.7",
134
121
  "color": "#0ea5e9",
135
122
  "permission": {
136
123
  "read": "allow",
@@ -148,7 +135,7 @@
148
135
  "heimdall": {
149
136
  "description": "Heimdall — Simple, routine, deterministic tasks.",
150
137
  "mode": "subagent",
151
- "model": "openrouter/minimax-m2.7",
138
+ "model": "openrouter/minimax/minimax-m2.7",
152
139
  "color": "#10b981",
153
140
  "permission": {
154
141
  "read": "allow",
@@ -165,7 +152,7 @@
165
152
  "hermod": {
166
153
  "description": "Hermod — Git and GitHub operations specialist.",
167
154
  "mode": "subagent",
168
- "model": "openrouter/minimax-m2.7",
155
+ "model": "openrouter/minimax/minimax-m2.7",
169
156
  "color": "#06b6d4",
170
157
  "permission": {
171
158
  "read": "allow",
@@ -181,7 +168,7 @@
181
168
  "thor": {
182
169
  "description": "Thor — Medium-complexity tasks, strong and reliable.",
183
170
  "mode": "subagent",
184
- "model": "openrouter/minimax-m2.7",
171
+ "model": "openrouter/minimax/minimax-m2.7",
185
172
  "color": "#a855f7",
186
173
  "permission": {
187
174
  "read": "allow",
@@ -198,7 +185,7 @@
198
185
  "baldr": {
199
186
  "description": "Baldr — UI/UX design system specialist.",
200
187
  "mode": "subagent",
201
- "model": "openrouter/minimax-m2.7",
188
+ "model": "openrouter/minimax/minimax-m2.7",
202
189
  "color": "#ec4899",
203
190
  "permission": {
204
191
  "read": "allow",
@@ -215,7 +202,7 @@
215
202
  "tyr": {
216
203
  "description": "Tyr — Complex implementation, debugging, architecture.",
217
204
  "mode": "subagent",
218
- "model": "openrouter/minimax-m3",
205
+ "model": "openrouter/minimax/minimax-m3",
219
206
  "color": "#f59e0b",
220
207
  "permission": {
221
208
  "read": "allow",
@@ -249,7 +236,7 @@
249
236
  "forseti": {
250
237
  "description": "Forseti — Audits and corrects plans before execution.",
251
238
  "mode": "subagent",
252
- "model": "openrouter/minimax-m3",
239
+ "model": "openrouter/minimax/minimax-m3",
253
240
  "color": "#ef4444",
254
241
  "permission": {
255
242
  "read": "allow",
@@ -325,11 +312,17 @@
325
312
  "models": {
326
313
  "MiniMax-M3": {
327
314
  "interleaved": { "field": "reasoning_details" },
328
- "reasoning": true
315
+ "reasoning": true,
316
+ "options": {
317
+ "thinking": "adaptive"
318
+ }
329
319
  },
330
320
  "MiniMax-M2.7": {
331
321
  "interleaved": { "field": "reasoning_details" },
332
- "reasoning": true
322
+ "reasoning": true,
323
+ "options": {
324
+ "thinking": "adaptive"
325
+ }
333
326
  }
334
327
  }
335
328
  },
@@ -337,11 +330,17 @@
337
330
  "models": {
338
331
  "minimax/minimax-m3": {
339
332
  "interleaved": { "field": "reasoning_details" },
340
- "reasoning": true
333
+ "reasoning": true,
334
+ "options": {
335
+ "reasoning": { "enabled": true }
336
+ }
341
337
  },
342
338
  "minimax/minimax-m2.7": {
343
339
  "interleaved": { "field": "reasoning_details" },
344
- "reasoning": true
340
+ "reasoning": true,
341
+ "options": {
342
+ "reasoning": { "enabled": true }
343
+ }
345
344
  }
346
345
  }
347
346
  }
@@ -5,4 +5,5 @@
5
5
  - All code must be reviewed by another agent before merging
6
6
  - Keep functions small and focused — a function should do one thing
7
7
  - Use meaningful names for variables, functions, and classes
8
- - Prefer readability over cleverness
8
+ - Prefer readability over cleverness
9
+ - When uncertain or stuck, reach for `websearch` / `webfetch` (or `semble search` / `hindsight_recall` for repo and project memory) before trying a third variation of the same edit — see `uncertainty.md`
@@ -98,8 +98,8 @@ permission:
98
98
 
99
99
  **Fix:** Check `~/.config/opencode/agents/<name>.md` for the `model:` field. Valid models:
100
100
  - `opencode/deepseek-v4-flash-free` — free
101
- - `openrouter/minimax-m2.7` — M2.7
102
- - `openrouter/minimax-m3` — M3
101
+ - `openrouter/minimax/minimax-m2.7` — M2.7
102
+ - `openrouter/minimax/minimax-m3` — M3
103
103
  - `openai/gpt-5.5` — GPT-5.5
104
104
 
105
105
  ### OpenRouter 404 Errors
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@polderlabs/bizar",
3
- "version": "3.12.0",
3
+ "version": "3.12.1",
4
4
  "description": "Norse-pantheon multi-agent system for opencode — 13 agents across 4 cost tiers with cost-aware routing, plans, and a configurable agent harness.",
5
5
  "type": "module",
6
6
  "bin": {