@polderlabs/bizar 3.12.4 → 3.14.1

package/README.md

@@ -66,17 +66,17 @@
 
 | Agent | Rune | Model | Cost | Role |
 |---|---|---|---|---|
-| **Odin** | ᛟ | OpenRouter minimax-m3 | $0.30/M · $1.20/M out | Primary router — never executes, only delegates |
+| **Odin** | ᛟ | MiniMax M3 | $0.30/M · $1.20/M out | Primary router — never executes, only delegates |
 | **Frigg** | ᚠ | DeepSeek V4 Flash | **Free** | Read-only Q&A — ask questions, get answers, never modifies |
 | **Vör** | ᛢ | DeepSeek V4 Flash | **Free** | Clarifies ambiguous requests — asks questions until task is well-defined |
 | **Mimir** | ᛗ | DeepSeek V4 Flash | **Free** | Research & codebase exploration (Semble-first) |
 | **Heimdall** | ᚹ | DeepSeek V4 Flash | **Free** | Simple tasks, quick edits, file operations |
-| **Hermod** | ᚱ | OpenRouter minimax-m2.7 | $0.30/M · $1.20/M out | Git & GitHub operations (commit, PR, merge, rebase) |
-| **Thor** | ᚦ | OpenRouter minimax-m2.7 | $0.30/M · $1.20/M out | Moderate implementation, debugging, refactoring |
-| **Baldr** | ᛒ | OpenRouter minimax-m2.7 | $0.30/M · $1.20/M out | UI/UX design system (DESIGN.md plans, visual audits) |
-| **Tyr** | ᛏ | OpenRouter minimax-m3 | $0.30/M · $1.20/M out | Complex implementation, architecture, deep debugging |
+| **Hermod** | ᚱ | MiniMax M2.7 | $0.30/M · $1.20/M out | Git & GitHub operations (commit, PR, merge, rebase) |
+| **Thor** | ᚦ | MiniMax M2.7 | $0.30/M · $1.20/M out | Moderate implementation, debugging, refactoring |
+| **Baldr** | ᛒ | MiniMax M2.7 | $0.30/M · $1.20/M out | UI/UX design system (DESIGN.md plans, visual audits) |
+| **Tyr** | ᛏ | MiniMax M3 | $0.30/M · $1.20/M out | Complex implementation, architecture, deep debugging |
 | **Vidarr** | ᛉ | GPT-5.5 | Subscription | Last resort — GPT-5.5 |
-| **Forseti** | ᚨ | OpenRouter minimax-m3 | $0.30/M · $1.20/M out | Adversarial plan reviewer (edit: deny, audit-only) |
+| **Forseti** | ᚨ | MiniMax M3 | $0.30/M · $1.20/M out | Adversarial plan reviewer (edit: deny, audit-only) |
 
 ---
 
@@ -303,7 +303,7 @@ After installation, run `/connect` in opencode to add API keys:
 | Provider | Models | Auth |
 |---|---|---|
 | **OpenCode Zen** | `opencode/deepseek-v4-flash-free` | Free API key from [opencode.ai](https://opencode.ai) — create account, get key, no charges |
-| **OpenRouter** | `openrouter/minimax/minimax-m2.7`, `openrouter/minimax/minimax-m3` | API key from [openrouter.ai](https://openrouter.ai) |
+| **MiniMax (direct)** | `minimax/MiniMax-M2.7`, `minimax/MiniMax-M3` | API key from [MiniMax](https://platform.minimaxi.com) |
 | **OpenAI** | `openai/gpt-5.5` | ChatGPT subscription (OAuth) |
 
 Then run `/models` to verify connectivity.

package/cli/audit.mjs

@@ -78,8 +78,8 @@ export async function runAudit() {
       const model = modelMatch[1].trim();
       const validModels = [
         'opencode/deepseek-v4-flash-free',
-        'openrouter/minimax/minimax-m2.7',
-        'openrouter/minimax/minimax-m3',
+        'minimax/MiniMax-M2.7',
+        'minimax/MiniMax-M3',
         'openai/gpt-5.5',
       ];
       if (!validModels.includes(model)) {

package/cli/bin.mjs

@@ -284,7 +284,7 @@ function showDoctorHelp() {
       • core agent files are installed (odin, quick, thor, tyr)
       • rtk / semble / skills on PATH (lenient — at least one)
       • dashboard reachable (skipped if no port file)
-      • provider.openrouter block + MiniMax model flags are sane
+      • provider.minimax block + MiniMax model flags are sane
 
     Prints ✓/✗ for each check and a final summary. Exits non-zero
     if any check fails. Use \`bizar doctor\` after a manual config

package/cli/doctor.mjs

@@ -211,11 +211,11 @@ async function checkProviderConfigSanity() {
     throw new Error('opencode.json missing');
   }
   const cfg = JSON.parse(readFileSync(cfgPath, 'utf8'));
-  const openrouter = cfg.provider && cfg.provider.openrouter;
-  if (!openrouter) {
-    throw new Error('provider.openrouter block missing');
+  const minimax = cfg.provider && cfg.provider.minimax;
+  if (!minimax) {
+    throw new Error('provider.minimax block missing');
   }
-  const models = openrouter.models || {};
+  const models = minimax.models || {};
   const saneNames = Object.entries(models).filter(([, m]) => {
     return (
       m &&
@@ -229,7 +229,7 @@ async function checkProviderConfigSanity() {
       'no MiniMax-style model with interleaved + reasoning flags',
     );
   }
-  return `provider.openrouter + ${saneNames.length} model(s) sane`;
+  return `provider.minimax + ${saneNames.length} model(s) sane`;
 }
 
 const CHECKS = [

package/cli/doctor.test.mjs

@@ -296,18 +296,18 @@ describe('runDoctor() with fixture HOME', () => {
     assert.equal(r.ok, true, r.message);
   });
 
-  test('provider-config-sanity fails without openrouter block', async () => {
+  test('provider-config-sanity fails without minimax block', async () => {
     writeOpencodeConfig({ provider: {} });
     const result = await runDoctor({ silent: true });
     const r = findCheck(result, 'provider-config-sanity');
     assert.equal(r.ok, false);
-    assert.match(r.message, /openrouter/);
+    assert.match(r.message, /minimax/);
   });
 
   test('provider-config-sanity fails when models lack interleaved+reasoning', async () => {
     writeOpencodeConfig({
       provider: {
-        openrouter: {
+        minimax: {
           models: { 'some-model': { id: 'foo' } },
         },
       },
@@ -321,7 +321,7 @@ describe('runDoctor() with fixture HOME', () => {
   test('provider-config-sanity passes with interleaved+reasoning', async () => {
     writeOpencodeConfig({
       provider: {
-        openrouter: {
+        minimax: {
           models: {
             'MiniMax/MiniMax-M3': { interleaved: true, reasoning: true },
           },

package/config/AGENTS.md

@@ -126,7 +126,7 @@ Follow `rules/uncertainty.md` strictly. When uncertain or stuck, the next move i
 
 This system uses a 5-tier model architecture with a verification gate:
 
-### Odin (default agent, OpenRouter minimax-m3)
+### Odin (default agent, MiniMax M3 direct)
 
 Odin (`@odin`) is the All-Father and primary/default agent. He analyzes each request and **decomposes it into independent work streams** — **he never executes work himself**:
 - **Identifies parallelizable work** and launches multiple subagent `task` calls in a **single message** (always 2+)
@@ -171,25 +171,25 @@ Odin (`@odin`) is the All-Father and primary/default agent. He analyzes each req
 
 ### Hermod
 
-- **Model**: `openrouter/minimax/minimax-m2.7` (via OpenRouter)
+- **Model**: `minimax/MiniMax-M2.7` (via MiniMax (direct))
 - **Use for**: Git and GitHub operations — commit, push, merge, PRs, branches, conflict resolution. The swift messenger.
 - **Cost**: $0.30/M input, $1.20/M output
 
 ### Thor
 
-- **Model**: `openrouter/minimax/minimax-m2.7` (via OpenRouter)
+- **Model**: `minimax/MiniMax-M2.7` (via MiniMax (direct))
 - **Use for**: Moderate complexity features, debugging, code review, refactoring
 - **Cost**: $0.30/M input, $1.20/M output — cheaper than Tyr, more capable than Heimdall
 
 ### Baldr
 
-- **Model**: `openrouter/minimax/minimax-m2.7` (via OpenRouter)
+- **Model**: `minimax/MiniMax-M2.7` (via MiniMax (direct))
 - **Use for**: Design system creation, DESIGN.md, visual audit, usability planning. Creates design plans — does not implement.
 - **Cost**: $0.30/M input, $1.20/M output
 
 ### Tyr
 
-- **Model**: `openrouter/minimax/minimax-m3` (via OpenRouter)
+- **Model**: `minimax/MiniMax-M3` (via MiniMax (direct))
 - **Use for**: Highest complexity implementation, debugging, architecture, multi-step engineering
 - **Cost**: Higher — reserved for the hardest problems
 
@@ -201,7 +201,7 @@ Odin (`@odin`) is the All-Father and primary/default agent. He analyzes each req
 
 ### Forseti
 
-- **Model**: `openrouter/minimax/minimax-m3` (via OpenRouter, audit-only, no edit permissions)
+- **Model**: `minimax/MiniMax-M3` (via MiniMax (direct), audit-only, no edit permissions)
 - **Use for**: Adversarial plan review — audits completeness, correctness, consistency, feasibility, security
 - **Always runs before any Tier 4 or Tier 5 implementation begins**
 

package/config/agents/baldr.md

@@ -1,7 +1,7 @@
 ---
 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.
 mode: subagent
-model: openrouter/minimax/minimax-m2.7
+model: minimax/MiniMax-M2.7
 color: "#ec4899"
 permission:
   read: allow

package/config/agents/forseti.md

@@ -1,7 +1,7 @@
 ---
 description: Forseti — Audits, criticizes, and corrects implementation plans before execution using MiniMax M3. No write permissions — review only.
 mode: subagent
-model: openrouter/minimax/minimax-m3
+model: minimax/MiniMax-M3
 color: "#ef4444"
 permission:
   read: allow

package/config/agents/hermod.md

@@ -1,7 +1,7 @@
 ---
 description: Hermod — Git and GitHub operations specialist using MiniMax M2.7. Branching, commits, PRs, merge/rebase, conflict resolution, CI/CD, releases, gh CLI.
 mode: subagent
-model: openrouter/minimax/minimax-m2.7
+model: minimax/MiniMax-M2.7
 color: "#06b6d4"
 permission:
   read: allow

package/config/agents/odin.md

@@ -1,7 +1,7 @@
 ---
 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).
 mode: primary
-model: openrouter/minimax/minimax-m3
+model: minimax/MiniMax-M3
 color: "#6366f1"
 permission:
   task: allow
@@ -53,8 +53,8 @@ When you get ANY request:
 4. After all return, synthesize the results
 
 For implementation work, you have two parallel implementation agents:
-- **@thor** (OpenRouter minimax-m2.7) — moderate complexity, cheaper
-- **@tyr** (OpenRouter minimax-m3) — complex work, more expensive
+- **@thor** (MiniMax M2.7) — moderate complexity, cheaper
+- **@tyr** (MiniMax M3) — complex work, more expensive
 
 **ALWAYS use both.** Split each implementation task across them. For example:
 - Frontend parts → @thor, Backend parts → @tyr
@@ -100,7 +100,7 @@ For deep codebase research, pattern discovery, documentation analysis:
 ### Simple Tasks & Quick Edits — Route to @heimdall (DeepSeek V4 Flash Free, free)
 For any simple, mechanical, or deterministic work:
 
-### Git Operations — Route to @hermod (MiniMax M2.7 via OpenRouter)
+### Git Operations — Route to @hermod (MiniMax M2.7)
 For any git or GitHub workflow:
 - Committing, pushing, pulling, branching, merging, rebasing
 - Pull request creation, review, and management
@@ -117,7 +117,7 @@ When the user asks for `@hermod /pr-review` or a PR review:
 2. @hermod waits for both, synthesizes the review, and posts as a PR comment
 3. @hermod has write access to post PR comments via `gh pr comment`
 
-### Design System & Visual Planning — Route to @baldr (MiniMax M2.7 via OpenRouter)
+### Design System & Visual Planning — Route to @baldr (MiniMax M2.7)
 For any task that touches visuals, usability, or design systems:
 - Creating DESIGN.md files (Google design.md standard — YAML tokens + prose sections)
 - Auditing visual consistency across a codebase (10-dimension scoring)
@@ -129,7 +129,7 @@ For any task that touches visuals, usability, or design systems:
 
 Baldr creates design plans. Baldr does NOT implement code — that goes to @thor or @tyr after the plan is approved.
 
-### Moderate Complexity — Route to @thor (MiniMax M2.7 via OpenRouter)
+### Moderate Complexity — Route to @thor (MiniMax M2.7)
 For tasks that need more reasoning than DeepSeek but aren't the hardest problems:
 - Implementing new features of moderate complexity
 - Debugging non-trivial issues
@@ -137,7 +137,7 @@ For tasks that need more reasoning than DeepSeek but aren't the hardest problems
 - Writing tests for non-trivial logic
 - Multi-step tasks that are well-scoped and understood
 
-### Complex Work — Route to @tyr (MiniMax M3 via OpenRouter)
+### Complex Work — Route to @tyr (MiniMax M3)
 For the most demanding engineering work:
 - Complex new feature implementation from scratch
 - Deep debugging of subtle or intermittent bugs
@@ -306,7 +306,7 @@ Call `bizar_spawn_background` with:
 
 - `agent`: the agent name (e.g., "mimir", "thor", "tyr")
 - `prompt`: what to do (specific, with context)
-- `model`: optional, `"<providerID>/<modelID>"` format (e.g., `"openrouter/minimax/minimax-m3"`)
+- `model`: optional, `"<providerID>/<modelID>"` format (e.g., `"minimax/MiniMax-M3"`)
 - `timeoutMs`: optional, default 5 min, max 30 min, min 1s
 
 You get an `instanceId` back immediately.

package/config/agents/quick.md

@@ -1,7 +1,7 @@
 ---
 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.
 mode: primary
-model: openrouter/minimax/minimax-m2.7
+model: minimax/MiniMax-M2.7
 color: "#22d3ee"
 permission:
   read: allow

package/config/agents/thor.md

@@ -1,7 +1,7 @@
 ---
-description: Thor — Handles medium-complexity tasks using MiniMax M2.7 from OpenRouter. Strong and reliable, cheaper than Tyr but more capable than Heimdall.
+description: Thor — Handles medium-complexity tasks using MiniMax M2.7 (direct). Strong and reliable, cheaper than Tyr but more capable than Heimdall.
 mode: subagent
-model: openrouter/minimax/minimax-m2.7
+model: minimax/MiniMax-M2.7
 color: "#a855f7"
 permission:
   read: allow

package/config/agents/tyr.md

@@ -1,7 +1,7 @@
 ---
-description: Tyr — Handles the most complex implementation, debugging, and architectural work using MiniMax M3 via OpenRouter. Unmatched wisdom for the hardest problems.
+description: Tyr — Handles the most complex implementation, debugging, and architectural work using MiniMax M3 (direct). Unmatched wisdom for the hardest problems.
 mode: subagent
-model: openrouter/minimax/minimax-m3
+model: minimax/MiniMax-M3
 color: "#f59e0b"
 permission:
   read: allow

package/config/opencode.json.template

@@ -1,7 +1,7 @@
 {
   "$schema": "https://opencode.ai/config.json",
-  "model": "openrouter/minimax/minimax-m3",
-  "small_model": "openrouter/minimax/minimax-m2.7",
+  "model": "minimax/MiniMax-M3",
+  "small_model": "minimax/MiniMax-M2.7",
   "default_agent": "odin",
   "permission": "allow",
   "snapshot": false,
@@ -53,7 +53,7 @@
     "odin": {
       "description": "Odin — Pure router that delegates all work to subagents.",
       "mode": "primary",
-      "model": "openrouter/minimax/minimax-m3",
+      "model": "minimax/MiniMax-M3",
       "color": "#6366f1",
       "permission": {
         "task": "allow",
@@ -67,7 +67,7 @@
     "vor": {
       "description": "Vör — The Questioning One.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#8b5cf6",
       "permission": {
         "read": "allow",
@@ -80,7 +80,7 @@
     "frigg": {
       "description": "Frigg — All-knowing Q&A agent.",
       "mode": "primary",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#06b6d4",
       "permission": {
         "read": "allow",
@@ -98,7 +98,7 @@
     "quick": {
       "description": "Quick — Fast single-shot tasks.",
       "mode": "primary",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#22d3ee",
       "permission": {
         "read": "allow",
@@ -117,7 +117,7 @@
     "mimir": {
       "description": "Mimir — Research and codebase exploration agent.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#0ea5e9",
       "permission": {
         "read": "allow",
@@ -135,7 +135,7 @@
     "heimdall": {
       "description": "Heimdall — Simple, routine, deterministic tasks.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#10b981",
       "permission": {
         "read": "allow",
@@ -152,7 +152,7 @@
     "hermod": {
       "description": "Hermod — Git and GitHub operations specialist.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#06b6d4",
       "permission": {
         "read": "allow",
@@ -168,7 +168,7 @@
     "thor": {
       "description": "Thor — Medium-complexity tasks, strong and reliable.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#a855f7",
       "permission": {
         "read": "allow",
@@ -185,7 +185,7 @@
     "baldr": {
       "description": "Baldr — UI/UX design system specialist.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m2.7",
+      "model": "minimax/MiniMax-M2.7",
       "color": "#ec4899",
       "permission": {
         "read": "allow",
@@ -202,7 +202,7 @@
     "tyr": {
       "description": "Tyr — Complex implementation, debugging, architecture.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m3",
+      "model": "minimax/MiniMax-M3",
       "color": "#f59e0b",
       "permission": {
         "read": "allow",
@@ -236,7 +236,7 @@
     "forseti": {
       "description": "Forseti — Audits and corrects plans before execution.",
       "mode": "subagent",
-      "model": "openrouter/minimax/minimax-m3",
+      "model": "minimax/MiniMax-M3",
       "color": "#ef4444",
       "permission": {
         "read": "allow",
@@ -314,32 +314,14 @@
           "interleaved": { "field": "reasoning_details" },
           "reasoning": true,
           "options": {
-            "thinking": "adaptive"
+            "thinking": { "type": "enabled" }
           }
         },
         "MiniMax-M2.7": {
           "interleaved": { "field": "reasoning_details" },
           "reasoning": true,
           "options": {
-            "thinking": "adaptive"
-          }
-        }
-      }
-    },
-    "openrouter": {
-      "models": {
-        "minimax/minimax-m3": {
-          "interleaved": { "field": "reasoning_details" },
-          "reasoning": true,
-          "options": {
-            "reasoning": { "enabled": true }
-          }
-        },
-        "minimax/minimax-m2.7": {
-          "interleaved": { "field": "reasoning_details" },
-          "reasoning": true,
-          "options": {
-            "reasoning": { "enabled": true }
+            "thinking": { "type": "enabled" }
           }
         }
       }

package/config/skills/bizar/SKILL.md

@@ -26,14 +26,14 @@ All subagents use Hindsight memory with **per-project banks**. Call `hindsight_l
 
 | Agent | Model | Tier | Cost | When to Route |
 |---|---|---|---|---|
-| **Odin** ᛟ | OpenRouter minimax-m3 | Router | $0.30/M · $1.20/M out | Primary entry point. Decomposes and dispatches. |
+| **Odin** ᛟ | MiniMax-M3 | Router | $0.30/M · $1.20/M out | Primary entry point. Decomposes and dispatches. |
 | **Mimir** ᛗ | DeepSeek V4 Flash | Free | **$0** | Deep codebase research, Semble-first exploration, docs analysis |
 | **Heimdall** ᚹ | DeepSeek V4 Flash | Free | **$0** | Simple edits, file ops, mechanical CRUD, quick answers |
-| **Hermod** ᚱ | OpenRouter minimax-m2.7 | Mid | $0.30/M · $1.20/M out | Git ops: commit, push, PR, merge, rebase, branches, `gh` CLI |
-| **Thor** ᚦ | OpenRouter minimax-m2.7 | Mid | $0.30/M · $1.20/M out | Moderate implementation, tests, debugging, refactoring |
-| **Tyr** ᛏ | OpenRouter minimax-m3 | High | $0.30/M · $1.20/M out | Complex features, architecture, deep debugging, cross-cutting refactor |
+| **Hermod** ᚱ | MiniMax-M2.7 | Mid | $0.30/M · $1.20/M out | Git ops: commit, push, PR, merge, rebase, branches, `gh` CLI |
+| **Thor** ᚦ | MiniMax-M2.7 | Mid | $0.30/M · $1.20/M out | Moderate implementation, tests, debugging, refactoring |
+| **Tyr** ᛏ | MiniMax-M3 | High | $0.30/M · $1.20/M out | Complex features, architecture, deep debugging, cross-cutting refactor |
 | **Vidarr** ᛉ | GPT-5.5 | Ultra | ChatGPT sub | Last resort when Tyr fails or debugging is stuck |
-| **Forseti** ᚨ | OpenRouter minimax-m3 | Gate | $0.30/M · $1.20/M out | Plan auditor — reviews Tyr/Vidarr plans before execution. `edit: deny`. |
+| **Forseti** ᚨ | MiniMax-M3 | Gate | $0.30/M · $1.20/M out | Plan auditor — reviews Tyr/Vidarr plans before execution. `edit: deny`. |
 | **Semble** | — | — | **$0** | MCP search tool, not an agent. Semble-first code search. |
 
 ## Odin Routing Rules
@@ -98,21 +98,42 @@ permission:
 
 **Fix:** Check `~/.config/opencode/agents/<name>.md` for the `model:` field. Valid models:
 - `opencode/deepseek-v4-flash-free` — free
-- `openrouter/minimax/minimax-m2.7` — M2.7
-- `openrouter/minimax/minimax-m3` — M3
+- `minimax/MiniMax-M2.7` — M2.7
+- `minimax/MiniMax-M3` — M3
 - `openai/gpt-5.5` — GPT-5.5
 
-### OpenRouter 404 Errors
+### MiniMax direct provider 404 errors
 
-**Symptoms:** 404 errors when calling the OpenRouter-hosted MiniMax models.
+When using the `minimax/MiniMax-M3` or `minimax/MiniMax-M2.7` model ids, do NOT set a custom `baseURL` on the `minimax` provider — opencode ships a built-in MiniMax provider that resolves the correct API endpoint. Adding an explicit baseURL is a common cause of 404s.
 
-**Cause:** A `baseURL` was set on the `openrouter` provider in `opencode.json`, which conflicts with the `@openrouter/ai-sdk-provider`'s built-in URL construction.
+### MiniMax rate-limit / quota failures
 
-**Fix:** Remove any `baseURL` from the openrouter provider section. OpenRouter's URL is built into the SDK. Example fix:
-```diff
-- "baseURL": "https://openrouter.ai/api/v1"
+**Symptoms:** A session dies with HTTP 429, 402, or 5xx from the MiniMax API. The user has multiple MiniMax accounts and wants them to share the load.
+
+**Cause:** opencode has no built-in multi-key rotation; it reads one key from `auth.json` and uses it for the entire session. Hitting that key's rate limit or quota is fatal.
+
+**Fix:** Set additional keys via env vars. The `bizar` plugin rotates through them on 429/402/5xx automatically.
+
+```bash
+# Option A — single comma-separated env var
+export MINIMAX_API_KEYS="key1,key2,key3"
+
+# Option B — numbered env vars (read in order, gaps skipped)
+export MINIMAX_API_KEY="primary-key"
+export MINIMAX_API_KEY_2="second-key"
+export MINIMAX_API_KEY_3="third-key"
 ```
-The provider uses the correct default URL internally.
+
+Rotation policy:
+- Triggers on 429 (rate limit), 402 (quota exhausted), 500/502/503/504 (server error)
+- Does NOT trigger on 401 (unauthorized — your key is wrong; fix the key, don't rotate)
+- Does NOT trigger on other 4xx (client error — the request itself is bad)
+- Network errors (ECONNRESET etc.) DO trigger rotation
+- Caps at N attempts where N = number of configured keys (try each once)
+- If all keys fail, the last error response is returned so opencode surfaces it normally
+- Round-robin on success: the next request starts on the next key, spreading load across accounts
+
+Single-key mode is unchanged — if only `MINIMAX_API_KEY` is set (or no key rotation env vars at all), the plugin works exactly as before.
 
 ### Forseti Rejects Every Plan
 
@@ -132,7 +153,7 @@ The provider uses the correct default URL internally.
 
 | File | Purpose |
 |---|---|
-| `~/.config/opencode/opencode.json` | Main config (no openrouter baseURL) |
+| `~/.config/opencode/opencode.json` | Main config (no external baseURL needed) |
 | `~/.config/opencode/AGENTS.md` | Routing table and conventions |
 | `~/.config/opencode/agents/odin.md` | Primary router agent |
 | `~/.config/opencode/agents/mimir.md` | Research agent |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@polderlabs/bizar",
-  "version": "3.12.4",
+  "version": "3.14.1",
   "description": "Norse-pantheon multi-agent system for opencode — 13 agents across 4 cost tiers with cost-aware routing, plans, and a configurable agent harness.",
   "type": "module",
   "bin": {
@@ -19,7 +19,7 @@
     "build:sdk": "npm run build -w @polderlabs/bizar-sdk",
     "test:sdk": "npm run test -w @polderlabs/bizar-sdk",
     "test:sdk:watch": "npm run test:watch -w @polderlabs/bizar-sdk",
-    "test": "npm run typecheck && npm run test:sdk && (cd plugins/bizar && bun test tests/loop.test.ts tests/block.test.ts tests/tools/bg-get-comments.test.ts tests/settings.test.ts tests/tools/plan-action.test.ts tests/tools/wait-for-feedback.test.ts) && node bizar-dash/tests/smoke-v2.mjs && node --test bizar-dash/tests/path-safe.test.mjs bizar-dash/tests/tmux-wrap.test.mjs bizar-dash/tests/opencode-runner.test.mjs",
+    "test": "npm run typecheck && npm run test:sdk && (cd plugins/bizar && bun test tests/loop.test.ts tests/block.test.ts tests/stall-think.test.ts tests/tools/bg-get-comments.test.ts tests/tools/bg-spawn-delegation.test.ts tests/tools/opencode-runner.test.ts tests/settings.test.ts tests/commands.test.ts tests/commands-impl.test.ts tests/tools/plan-action.test.ts tests/tools/wait-for-feedback.test.ts tests/reasoning-clean.test.ts tests/key-rotation.test.ts) && node bizar-dash/tests/smoke-v2.mjs && node --test bizar-dash/tests/path-safe.test.mjs bizar-dash/tests/tmux-wrap.test.mjs bizar-dash/tests/opencode-runner.test.mjs",
     "build": "npm run build:sdk"
   },
   "keywords": [