pi-opencode-go-provider 1.0.6 → 1.0.8

package/README.md

@@ -68,22 +68,24 @@ pi
 
 | Model | API | Type | Context | Max Tokens | Input Cost | Output Cost |
 |-------|-----|------|---------|------------|------------|-------------|
-| DeepSeek V4 Flash | Completions | Text | 1.0M | 384K | $0.14 | $0.28 |
+| DeepSeek V4 Flash | Completions | Text | 1.0M | 384K | $0.07 | $0.14 |
 | DeepSeek V4 Pro | Completions | Text | 1.0M | 384K | $0.43 | $0.87 |
 | GLM-5.1 | Completions | Text | 203K | 33K | $1.40 | $4.40 |
 | GLM-5.2 | Completions | Text | 1.0M | 131K | $1.40 | $4.40 |
-| Grok 4.5 | Completions | Text + Image | 500K | 500K | $2.00 | $6.00 |
+| GPT-5.6 Luna (2x usage) | Responses | Text + Image | 1.1M | 128K | $0.10 | $0.60 |
+| Grok 4.5 | Responses | Text + Image | 500K | 500K | $2.00 | $6.00 |
 | Hy3 | Completions | Text | 256K | 64K | $0.14 | $0.58 |
 | Kimi K2.6 | Completions | Text + Image | 262K | 66K | $0.95 | $4.00 |
 | Kimi K2.7 Code | Completions | Text + Image | 262K | 262K | $0.95 | $4.00 |
 | Kimi K3 (2x usage) | Completions | Text + Image | 1.0M | 131K | $3.00 | $15.00 |
 | MiMo V2.5 | Completions | Text + Image | 1.0M | 128K | $0.14 | $0.28 |
 | MiMo V2.5 Pro | Completions | Text | 1.0M | 128K | $0.43 | $0.87 |
-| MiniMax-M2.7 | Anthropic | Text | 205K | 131K | $0.30 | $1.20 |
+| MiniMax-M2.7 | Completions | Text | 205K | 131K | $0.30 | $1.20 |
 | MiniMax-M3 | Anthropic | Text + Image | 1.0M | 131K | $0.30 | $1.20 |
-| Qwen3.6 Plus | Anthropic | Text + Image | 1.0M | 66K | $0.50 | $3.00 |
+| Qwen3.6 Plus | Completions | Text + Image | 1.0M | 66K | $0.50 | $3.00 |
 | Qwen3.7 Max | Anthropic | Text | 1.0M | 66K | $2.50 | $7.50 |
 | Qwen3.7 Plus | Anthropic | Text + Image | 1.0M | 66K | $0.40 | $1.60 |
+| Qwen3.8 Max | Anthropic | Text + Image | 1.0M | 131K | $2.00 | $6.00 |
 *Costs are per million tokens. Prices subject to change - check [opencode.ai](https://opencode.ai) for current pricing.*
 
 ## Usage

package/index.ts

@@ -52,7 +52,9 @@ interface JsonModel {
     cacheRead: number;
     cacheWrite: number;
   };
-  contextWindow: number;
+  // null between transformApiModel and mergeWithEmbedded: the live /v1/models
+  // API omits context data; null = "API silent, prefer curated/default".
+  contextWindow: number | null;
   maxTokens: number;
   compat?: Record<string, unknown>;
 }
@@ -185,7 +187,12 @@ function transformApiModel(apiModel: any): JsonModel {
       cacheRead: apiModel.cost?.cache_read || 0,
       cacheWrite: apiModel.cost?.cache_write || 0,
     },
-    contextWindow: apiModel.limit?.context || apiModel.context_length || 131072,
+    // The live /v1/models endpoint exposes only {id, object, created, owned_by}
+    // — no `limit`, no `context_length`. A literal `|| 131072` here used to make
+    // every "API silent" model truthy 131072, which then clobbered the curated
+    // 1M/262K values in mergeWithEmbedded via the `||` truthy check. Use `null`
+    // so the embedded curated value is preferred when the API is silent.
+    contextWindow: apiModel.limit?.context || apiModel.context_length || null,
     maxTokens: apiModel.limit?.output || 0,
   };
 }
@@ -246,10 +253,20 @@ function mergeWithEmbedded(liveModels: JsonModel[], embeddedModels: JsonModel[])
           cacheRead: liveModel.cost.cacheRead || embedded.cost.cacheRead,
           cacheWrite: liveModel.cost.cacheWrite || embedded.cost.cacheWrite,
         },
-        contextWindow: liveModel.contextWindow || embedded.contextWindow,
+        // Nullish coalesce: if the live API was silent (liveModel.contextWindow
+        // is null), fall through to the curated embedded value. The final
+        // `?? 131072` is the safe default for the rare case of a Custom-only
+        // model with neither live nor curated data. Using `||` here was the
+        // bug — it treated the API's 131072-fallback as authoritative and
+        // shadowed curated 1M/262K context windows.
+        contextWindow: liveModel.contextWindow ?? embedded.contextWindow ?? 131072,
       });
     } else {
-      result.push(liveModel);
+      // API-only model (no curated embedded counterpart). The Live API is silent
+      // on contextWindow for all current opencode-go models, so liveModel.contextWindow
+      // is null. Apply the safe default here so downstream consumers (e.g. pi core's
+      // formatTokenCount → count.toString()) don't crash on null.
+      result.push({ ...liveModel, contextWindow: liveModel.contextWindow ?? 131072 });
     }
   }
   // Append any embedded models that the live API didn't return
@@ -346,7 +363,7 @@ export default function (pi: ExtensionAPI) {
       thinkingLevelMap: m.thinkingLevelMap,
       input: m.input,
       cost: m.cost,
-      contextWindow: m.contextWindow,
+      contextWindow: m.contextWindow ?? 131072,
       maxTokens: m.maxTokens,
       compat: m.compat,
     })),
@@ -373,7 +390,7 @@ export default function (pi: ExtensionAPI) {
               thinkingLevelMap: m.thinkingLevelMap,
               input: m.input,
               cost: m.cost,
-              contextWindow: m.contextWindow,
+              contextWindow: m.contextWindow ?? 131072,
               maxTokens: m.maxTokens,
               compat: m.compat,
             })),

package/models.json

@@ -38,7 +38,7 @@
   },
   {
     "id": "deepseek-v4-flash",
-    "name": "DeepSeek V4 Flash",
+    "name": "DeepSeek V4 Flash (2x usage)",
     "api": "openai-completions",
     "baseUrl": "https://opencode.ai/zen/go/v1",
     "reasoning": true,
@@ -46,9 +46,9 @@
       "text"
     ],
     "cost": {
-      "input": 0.14,
-      "output": 0.28,
-      "cacheRead": 0.0028,
+      "input": 0.07,
+      "output": 0.14,
+      "cacheRead": 0.0014,
       "cacheWrite": 0
     },
     "contextWindow": 1000000,
@@ -182,6 +182,25 @@
     "contextWindow": 1000000,
     "maxTokens": 384000
   },
+  {
+    "id": "qwen3.8-max",
+    "name": "Qwen3.8 Max",
+    "api": "anthropic-messages",
+    "baseUrl": "https://opencode.ai/zen/go",
+    "reasoning": true,
+    "input": [
+      "text",
+      "image"
+    ],
+    "cost": {
+      "input": 2,
+      "output": 6,
+      "cacheRead": 0.25,
+      "cacheWrite": 2.5
+    },
+    "contextWindow": 1000000,
+    "maxTokens": 131072
+  },
   {
     "id": "mimo-v2.5",
     "name": "MiMo V2.5",
@@ -201,6 +220,25 @@
     "contextWindow": 1000000,
     "maxTokens": 128000
   },
+  {
+    "id": "gpt-5.6-luna",
+    "name": "GPT-5.6 Luna (2x usage)",
+    "api": "openai-completions",
+    "baseUrl": "https://opencode.ai/zen/go/v1",
+    "reasoning": true,
+    "input": [
+      "text",
+      "image"
+    ],
+    "cost": {
+      "input": 0.1,
+      "output": 0.6,
+      "cacheRead": 0.01,
+      "cacheWrite": 0.125
+    },
+    "contextWindow": 1050000,
+    "maxTokens": 128000
+  },
   {
     "id": "grok-4.5",
     "name": "Grok 4.5",
@@ -241,7 +279,7 @@
   },
   {
     "id": "kimi-k3",
-    "name": "Kimi K3 (2x usage)",
+    "name": "Kimi K3",
     "api": "openai-completions",
     "baseUrl": "https://opencode.ai/zen/go/v1",
     "reasoning": true,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-opencode-go-provider",
-  "version": "1.0.6",
+  "version": "1.0.8",
   "description": "Opencode Go provider extension for pi - Fast, efficient GLM, Kimi, and MiniMax models through the opencode.ai API",
   "type": "module",
   "main": "index.ts",

package/patch.json

@@ -1,17 +1,176 @@
 {
+  "glm-5.1": {
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
   "deepseek-v4-flash": {
-    "compat": { "requiresReasoningContentOnAssistantMessages": true, "thinkingFormat": "deepseek" },
-    "thinkingLevelMap": { "minimal": null, "low": null, "medium": null, "high": "high", "max": "max" }
+    "name": "DeepSeek V4 Flash",
+    "thinkingLevelMap": {
+      "minimal": null,
+      "low": null,
+      "medium": null,
+      "high": "high",
+      "max": "max"
+    },
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "requiresReasoningContentOnAssistantMessages": true,
+      "thinkingFormat": "deepseek"
+    }
   },
-  "deepseek-v4-pro": {
-    "compat": { "requiresReasoningContentOnAssistantMessages": true, "thinkingFormat": "deepseek" },
-    "thinkingLevelMap": { "minimal": null, "low": null, "medium": null, "high": "high", "max": "max" }
+  "minimax-m2.7": {
+    "api": "openai-completions",
+    "baseUrl": "https://opencode.ai/zen/go/v1",
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
+  "glm-5.2": {
+    "thinkingLevelMap": {
+      "off": null,
+      "minimal": null,
+      "low": null,
+      "medium": null,
+      "high": "high",
+      "xhigh": null,
+      "max": "max"
+    },
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   "kimi-k2.6": {
-    "compat": { "thinkingFormat": "deepseek", "supportsReasoningEffort": false },
-    "thinkingLevelMap": { "minimal": null, "low": null, "medium": null }
+    "thinkingLevelMap": {
+      "minimal": null,
+      "low": null,
+      "medium": null
+    },
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "thinkingFormat": "deepseek",
+      "supportsReasoningEffort": false,
+      "maxTokensField": "max_tokens",
+      "supportsLongCacheRetention": false
+    }
+  },
+  "hy3": {
+    "thinkingLevelMap": {
+      "off": "none",
+      "minimal": null,
+      "low": "low",
+      "medium": null,
+      "high": "high",
+      "xhigh": null,
+      "max": null
+    },
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
+  "deepseek-v4-pro": {
+    "thinkingLevelMap": {
+      "minimal": null,
+      "low": null,
+      "medium": null,
+      "high": "high",
+      "max": "max"
+    },
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "requiresReasoningContentOnAssistantMessages": true,
+      "thinkingFormat": "deepseek"
+    }
+  },
+  "mimo-v2.5": {
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
+  "grok-4.5": {
+    "api": "openai-responses",
+    "thinkingLevelMap": {
+      "off": null,
+      "minimal": null,
+      "low": "low",
+      "medium": "medium",
+      "high": "high",
+      "xhigh": null,
+      "max": null
+    },
+    "compat": {
+      "sessionAffinityFormat": "openai-nosession"
+    }
+  },
+  "kimi-k2.7-code": {
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
+  "kimi-k3": {
+    "name": "Kimi K3 (2x usage)",
+    "thinkingLevelMap": {
+      "off": null,
+      "minimal": null,
+      "low": null,
+      "medium": null,
+      "high": null,
+      "xhigh": null,
+      "max": "max"
+    },
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
+  "mimo-v2.5-pro": {
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
+  },
+  "gpt-5.6-luna": {
+    "api": "openai-responses",
+    "thinkingLevelMap": {
+      "off": null,
+      "minimal": null,
+      "low": "low",
+      "medium": "medium",
+      "high": "high",
+      "xhigh": "xhigh",
+      "max": "max"
+    },
+    "compat": {
+      "sessionAffinityFormat": "openai-nosession"
+    }
   },
   "qwen3.6-plus": {
-    "compat": { "thinkingFormat": "qwen" }
+    "api": "openai-completions",
+    "baseUrl": "https://opencode.ai/zen/go/v1",
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "thinkingFormat": "qwen",
+      "maxTokensField": "max_tokens"
+    }
   }
 }

package/scripts/update-models.js

@@ -101,6 +101,46 @@ function convertModel(model) {
   };
 }
 
+// Load a layered JSON model file.
+function loadJson(fileName) {
+  try {
+    return JSON.parse(fs.readFileSync(path.join(process.cwd(), fileName), 'utf8'));
+  } catch {
+    return fileName === 'patch.json' ? {} : [];
+  }
+}
+
+// Deep-merge patch overrides into a model for README documentation. The
+// generated models.json remains API-derived; only the displayed model map uses patches.
+function applyPatch(model, patch) {
+  if (!patch) return model;
+  const result = { ...model };
+  for (const [key, value] of Object.entries(patch)) {
+    if (key === 'cost' || key === 'compat' || key === 'thinkingLevelMap') {
+      result[key] = { ...(result[key] || {}), ...value };
+    } else {
+      result[key] = value;
+    }
+  }
+  if (!result.reasoning) {
+    delete result.thinkingLevelMap;
+    if (result.compat?.thinkingFormat) delete result.compat.thinkingFormat;
+  }
+  return result;
+}
+
+// Merge API models, patches, and custom models in the same order as index.ts.
+function buildModels(baseModels, customModels, patch) {
+  const byId = new Map(baseModels.map(model => [model.id, model]));
+  for (const [id, entry] of Object.entries(patch)) {
+    if (byId.has(id)) byId.set(id, applyPatch(byId.get(id), entry));
+  }
+  for (const model of customModels) {
+    byId.set(model.id, applyPatch(model, patch[model.id]));
+  }
+  return Array.from(byId.values());
+}
+
 // Generate README model table row
 function generateReadmeRow(model) {
   const cost = model.cost || {};
@@ -238,8 +278,10 @@ async function main() {
     fs.writeFileSync(modelsPath, JSON.stringify(models, null, 2) + '\n');
     console.log(`  Saved ${models.length} models to models.json`);
 
-    // Update README
-    updateReadme(models);
+    // Apply layered overrides for documentation: models.json → patch.json → custom-models.json.
+    const patch = loadJson('patch.json');
+    const customModels = Array.isArray(loadJson('custom-models.json')) ? loadJson('custom-models.json') : [];
+    updateReadme(buildModels(models, customModels, patch));
 
     console.log('\nDone!');
   } catch (error) {