pi-io-provider 1.1.1 → 1.1.3

package/AGENTS.md

@@ -7,6 +7,7 @@ The following files are **idempotent** and regenerated by `scripts/update-models
 | File | Why it's auto-generated |
 |------|------------------------|
 | `models.json` | Built from the provider API. `update-models.js` fetches models, preserves curated data for known IDs, and writes this file. |
+| `deprecated-models.json` | Graveyard for models the API delisted. update-models.js stamps them with deprecatedAt and pi keeps serving them for a 2-week grace period, then evicts them. Never edit by hand. |
 | `README.md` (model table) | The table under `## Available Models` is replaced in-place by `update-models.js` after merging base models → patch → custom models. |
 
 ## Correct Files to Edit

package/README.md

@@ -16,35 +16,38 @@ A [pi](https://github.com/badlogic/pi-mono) extension that adds [IO Intelligence
 
 | Model | ID | Context | Max Output | Vision | Reasoning | Cache | Input $/M | Output $/M |
 |-------|----|---------|------------|--------|-----------|-------|-----------|------------|
+| Kimi K2.5 | `moonshotai/Kimi-K2.5` | 262K | 262K | ✅ | ✅ | ✅ | $0.56 | $2.82 |
+| Kimi K2.6 | `moonshotai/Kimi-K2.6` | 262K | 262K | ✅ | ✅ | ✅ | $0.84 | $3.76 |
+| Kimi K2.7 Code | `moonshotai/Kimi-K2.7-Code` | 262K | 262K | ✅ | ✅ | ✅ | $1.07 | $4.65 |
+| Kimi K3 | `moonshotai/Kimi-K3` | 1.0M | 1.0M | ✅ | ✅ | ✅ | $3.60 | $18.00 |
+| Qwen3.6 27B | `Qwen/Qwen3.6-27B` | 33K | 33K | ✅ | ✅ | ✅ | $0.37 | $3.19 |
+| Qwen3.6 35B A3B | `Qwen/Qwen3.6-35B-A3B` | 262K | 262K | ✅ | ✅ | ✅ | $0.17 | $1.12 |
 | DeepSeek R1 0528 | `deepseek-ai/DeepSeek-R1-0528` | 128K | 128K | ❌ | ✅ | ✅ | $0.57 | $2.28 |
+| DeepSeek V3.2 | `deepseek-ai/DeepSeek-V3.2` | 164K | 164K | ❌ | ✅ | ✅ | $0.90 | $1.76 |
+| DeepSeek V4 Flash | `deepseek-ai/DeepSeek-V4-Flash` | 33K | 33K | ❌ | ✅ | ✅ | $0.20 | $0.37 |
+| DeepSeek V4 Flash 0731 | `deepseek-ai/DeepSeek-V4-Flash-0731` | 262K | 66K | ❌ | ✅ | ✅ | $0.14 | $0.28 |
+| DeepSeek V4 Pro | `deepseek-ai/DeepSeek-V4-Pro` | 1.0M | 600K | ❌ | ✅ | ✅ | $1.52 | $3.04 |
+| Gemma 4 26B A4B | `google/gemma-4-26b-a4b-it` | 262K | 262K | ❌ | ✅ | ✅ | $0.13 | $0.43 |
+| MiniMax M2.5 | `MiniMaxAI/MiniMax-M2.5` | 197K | 197K | ❌ | ✅ | ✅ | $0.29 | $1.15 |
+| MiniMax M2.7 | `MiniMaxAI/MiniMax-M2.7` | 262K | 66K | ❌ | ✅ | ✅ | $0.44 | $1.72 |
 | Kimi K2 Thinking | `moonshotai/Kimi-K2-Thinking` | 262K | 262K | ❌ | ✅ | ✅ | $0.60 | $2.50 |
+| gpt-oss-120b | `openai/gpt-oss-120b` | 131K | 131K | ❌ | ✅ | ✅ | $0.19 | $0.70 |
+| gpt-oss-20b | `openai/gpt-oss-20b` | 64K | 64K | ❌ | ✅ | ✅ | $0.05 | $0.22 |
+| MiMo-V2.5 | `XiaomiMiMo/MiMo-V2.5` | 262K | 262K | ❌ | ✅ | ✅ | $0.20 | $0.64 |
+| GLM-4.5-Air | `zai-org/GLM-4.5-Air` | 131K | 131K | ❌ | ✅ | ✅ | $0.16 | $0.94 |
+| GLM 4.6 | `zai-org/GLM-4.6` | 131K | 131K | ❌ | ✅ | ✅ | $0.54 | $2.07 |
+| GLM 4.7 | `zai-org/GLM-4.7` | 203K | 203K | ❌ | ✅ | ✅ | $0.86 | $2.24 |
+| GLM 4.7 Flash | `zai-org/GLM-4.7-Flash` | 200K | 200K | ❌ | ✅ | ✅ | $0.08 | $0.42 |
+| GLM 5 | `zai-org/GLM-5` | 203K | 203K | ❌ | ✅ | ✅ | $0.85 | $2.62 |
+| GLM 5.1 | `zai-org/GLM-5.1` | 203K | 33K | ❌ | ✅ | ✅ | $1.27 | $4.13 |
+| GLM 5.2 | `zai-org/GLM-5.2` | 262K | 66K | ❌ | ✅ | ✅ | $1.96 | $6.16 |
 | Llama 3.2 90B Vision Instruct | `meta-llama/Llama-3.2-90B-Vision-Instruct` | 16K | 16K | ✅ | ❌ | ✅ | $0.34 | $0.34 |
 | Llama 4 Maverick 17B 128E Instruct FP8 | `meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8` | 430K | 430K | ✅ | ❌ | ✅ | $0.28 | $0.93 |
-| Kimi K2.5 | `moonshotai/Kimi-K2.5` | 262K | 262K | ✅ | ❌ | ✅ | $0.53 | $2.79 |
-| Kimi K2.6 | `moonshotai/Kimi-K2.6` | 262K | 262K | ✅ | ❌ | ✅ | $0.95 | $3.98 |
-| DeepSeek V3.2 | `deepseek-ai/DeepSeek-V3.2` | 164K | 164K | ❌ | ❌ | ✅ | $0.89 | $1.63 |
-| DeepSeek V4 Flash | `deepseek-ai/DeepSeek-V4-Flash` | 33K | 33K | ❌ | ❌ | ✅ | $0.20 | $0.37 |
-| DeepSeek V4 Pro | `deepseek-ai/DeepSeek-V4-Pro` | 1.0M | 600K | ❌ | ❌ | ✅ | $1.65 | $3.30 |
-| Gemma 4 26B A4B | `google/gemma-4-26b-a4b-it` | 262K | 262K | ❌ | ❌ | ✅ | $0.11 | $0.36 |
-| Qwen3 Coder 480B A35B Instruct INT4 Mixed AR | `Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar` | 106K | 106K | ❌ | ❌ | ✅ | $0.60 | $2.08 |
-| Llama 3.3 70B Instruct | `meta-llama/Llama-3.3-70B-Instruct` | 128K | 128K | ❌ | ❌ | ✅ | $0.61 | $1.04 |
-| MiniMax M2.5 | `MiniMaxAI/MiniMax-M2.5` | 197K | 197K | ❌ | ❌ | ✅ | $0.28 | $1.13 |
-| MiniMax M2.7 | `MiniMaxAI/MiniMax-M2.7` | 205K | 205K | ❌ | ❌ | ✅ | $0.35 | $1.33 |
-| Mistral Nemo Instruct 2407 | `mistralai/Mistral-Nemo-Instruct-2407` | 128K | 128K | ❌ | ❌ | ✅ | $0.06 | $0.10 |
+| Qwen3 Coder 480B A35B Instruct INT4 Mixed AR | `Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar` | 106K | 106K | ❌ | ❌ | ✅ | $0.57 | $2.13 |
+| Llama 3.3 70B Instruct | `meta-llama/Llama-3.3-70B-Instruct` | 128K | 128K | ❌ | ❌ | ✅ | $0.51 | $1.04 |
+| Mistral Nemo Instruct 2407 | `mistralai/Mistral-Nemo-Instruct-2407` | 128K | 128K | ❌ | ❌ | ✅ | $0.07 | $0.12 |
 | Kimi K2 Instruct 0905 | `moonshotai/Kimi-K2-Instruct-0905` | 262K | 262K | ❌ | ❌ | ✅ | $0.57 | $2.30 |
-| Kimi K2.7 Code | `moonshotai/Kimi-K2.7-Code` | 262K | 262K | ❌ | ❌ | ✅ | $0.90 | $3.90 |
-| gpt-oss-120b | `openai/gpt-oss-120b` | 131K | 131K | ❌ | ❌ | ✅ | $0.16 | $0.59 |
-| gpt-oss-20b | `openai/gpt-oss-20b` | 64K | 64K | ❌ | ❌ | ✅ | $0.06 | $0.23 |
 | Qwen3 Next 80B A3B Instruct | `Qwen/Qwen3-Next-80B-A3B-Instruct` | 262K | 262K | ❌ | ❌ | ✅ | $0.12 | $1.14 |
-| Qwen3.6 27B | `Qwen/Qwen3.6-27B` | 33K | 33K | ❌ | ❌ | ✅ | $0.43 | $3.21 |
-| Qwen3.6 35B A3B | `Qwen/Qwen3.6-35B-A3B` | 262K | 262K | ❌ | ❌ | ✅ | $0.19 | $1.12 |
-| GLM-4.5-Air | `zai-org/GLM-4.5-Air` | 131K | 131K | ❌ | ❌ | ✅ | $0.16 | $0.94 |
-| GLM 4.6 | `zai-org/GLM-4.6` | 131K | 131K | ❌ | ❌ | ✅ | $0.54 | $2.07 |
-| GLM 4.7 | `zai-org/GLM-4.7` | 203K | 203K | ❌ | ❌ | ✅ | $0.86 | $2.24 |
-| GLM 4.7 Flash | `zai-org/GLM-4.7-Flash` | 200K | 200K | ❌ | ❌ | ✅ | $0.08 | $0.42 |
-| GLM 5 | `zai-org/GLM-5` | 203K | 203K | ❌ | ❌ | ✅ | $0.84 | $2.69 |
-| GLM 5.1 | `zai-org/GLM-5.1` | 203K | 33K | ❌ | ❌ | ✅ | $1.37 | $4.47 |
-| GLM 5.2 | `zai-org/GLM-5.2` | 262K | 66K | ❌ | ❌ | ✅ | $1.64 | $5.13 |
 
 *Costs are per million tokens. Cache read/write pricing available on most models.*
 

package/deprecated-models.json

@@ -0,0 +1 @@
+{}

package/index.ts

@@ -40,6 +40,7 @@ import { getAgentDir, type ExtensionAPI, type ModelRegistry } from "@earendil-wo
 import modelsData from "./models.json" with { type: "json" };
 import customModelsData from "./custom-models.json" with { type: "json" };
 import patchData from "./patch.json" with { type: "json" };
+import deprecatedData from "./deprecated-models.json" with { type: "json" };
 import fs from "fs";
 import path from "path";
 
@@ -58,12 +59,14 @@ interface JsonModel {
   };
   contextWindow: number;
   maxTokens: number;
+  thinkingLevelMap?: Record<string, string | null>;
   compat?: {
     supportsDeveloperRole?: boolean;
     supportsStore?: boolean;
     maxTokensField?: "max_completion_tokens" | "max_tokens";
     thinkingFormat?: "openai" | "zai" | "qwen" | "qwen-chat-template";
     supportsReasoningEffort?: boolean;
+    requiresReasoningContentOnAssistantMessages?: boolean;
   };
 }
 
@@ -79,6 +82,7 @@ interface PatchEntry {
   };
   contextWindow?: number;
   maxTokens?: number;
+  thinkingLevelMap?: Record<string, string | null>;
   compat?: Record<string, unknown>;
 }
 
@@ -94,6 +98,7 @@ function applyPatch(model: JsonModel, patch: PatchEntry): JsonModel {
   if (patch.input !== undefined) result.input = patch.input;
   if (patch.contextWindow !== undefined) result.contextWindow = patch.contextWindow;
   if (patch.maxTokens !== undefined) result.maxTokens = patch.maxTokens;
+  if (patch.thinkingLevelMap !== undefined) result.thinkingLevelMap = { ...patch.thinkingLevelMap };
 
   if (patch.cost) {
     result.cost = {
@@ -110,6 +115,9 @@ function applyPatch(model: JsonModel, patch: PatchEntry): JsonModel {
   if (!result.reasoning && result.compat?.thinkingFormat) {
     delete result.compat.thinkingFormat;
   }
+  if (!result.reasoning && result.thinkingLevelMap) {
+    delete result.thinkingLevelMap;
+  }
   if (result.compat && Object.keys(result.compat).length === 0) {
     delete result.compat;
   }
@@ -121,7 +129,10 @@ function applyPatch(model: JsonModel, patch: PatchEntry): JsonModel {
 function buildModels(base: JsonModel[], custom: JsonModel[], patch: PatchData): JsonModel[] {
   const modelMap = new Map<string, JsonModel>();
 
-  for (const model of base) {
+  // Seed with the base list plus grace-period deprecated models so patch.json
+  // entries apply to deprecated models exactly as while the model was live
+  // (withDeprecated keeps live data on id conflicts).
+  for (const model of withDeprecated(base)) {
     modelMap.set(model.id, model);
   }
 
@@ -160,7 +171,7 @@ const LIVE_FETCH_TIMEOUT_MS = 8000;
 
 /** Transform a model from the IO Intelligence /v1/models API to JsonModel format. */
 function transformApiModel(apiModel: any): JsonModel | null {
-  const hasVision = apiModel.supports_images_input === true;
+  const hasVision = apiModel.supports_images_input === true || (apiModel.input_modalities || []).includes("image");
   // IO returns per-token pricing, convert to per-million. Round to 6 decimals to
   // normalize float noise from the ×1e6 multiply and preserve sub-cent cache prices.
   const toPerM = (v: any) => Math.round((typeof v === "number" ? v * 1_000_000 : 0) * 1e6) / 1e6;
@@ -177,13 +188,14 @@ function transformApiModel(apiModel: any): JsonModel | null {
       cacheWrite: toPerM(apiModel.cache_write_token_price),
     },
     contextWindow: apiModel.context_window || 131072,
-    maxTokens: apiModel.max_tokens || 0,
+    maxTokens: apiModel.max_tokens || apiModel.context_window || 131072,
+    compat: {
+      supportsStore: false,
+      supportsDeveloperRole: false,
+      maxTokensField: "max_tokens",
+      ...(hasReasoning ? { supportsReasoningEffort: true } : {}),
+    },
   };
-  if (hasReasoning) {
-    model.compat = {
-      supportsReasoningEffort: true,
-    };
-  }
   return model;
 }
 
@@ -258,6 +270,35 @@ function mergeWithEmbedded(liveModels: JsonModel[], embeddedModels: JsonModel[])
   return result;
 }
 
+// Grace period for delisted models. When the provider API stops listing a
+// model, update-models.js moves its last-known definition into
+// deprecated-models.json (stamped with deprecatedAt) instead of dropping it.
+// For 14 days the model keeps working here so in-flight sessions and saved
+// model settings do not break; afterwards it is evicted permanently.
+const DEPRECATED_MODEL_TTL_MS = 14 * 24 * 60 * 60 * 1000;
+
+// Grace-period deprecated models with deprecation metadata stripped.
+function activeDeprecatedModels(): JsonModel[] {
+  const now = Date.now();
+  const result: JsonModel[] = [];
+  for (const entry of Object.values(deprecatedData as Record<string, JsonModel & { deprecatedAt?: string }>)) {
+    if (!entry?.id) continue;
+    const removedAt = Date.parse(entry.deprecatedAt ?? "");
+    if (Number.isNaN(removedAt) || now - removedAt > DEPRECATED_MODEL_TTL_MS) continue;
+    const model = { ...entry } as JsonModel & { deprecatedAt?: string };
+    delete model.deprecatedAt;
+    result.push(model);
+  }
+  return result;
+}
+
+// Append grace-period deprecated models the list does not already have (live data wins).
+function withDeprecated(models: JsonModel[]): JsonModel[] {
+  const seen = new Set(models.map((m) => m.id));
+  const extras = activeDeprecatedModels().filter((m) => !seen.has(m.id));
+  return extras.length > 0 ? [...models, ...extras] : models;
+}
+
 function loadStaleModels(embeddedModels: JsonModel[]): JsonModel[] {
   const cached = loadCachedModels();
   if (!cached || cached.length === 0) return embeddedModels;

package/models.json

@@ -1,88 +1,188 @@
 [
+  {
+    "id": "deepseek-ai/DeepSeek-V4-Flash-0731",
+    "name": "DeepSeek V4 Flash 0731",
+    "reasoning": true,
+    "input": [
+      "text"
+    ],
+    "cost": {
+      "input": 0.14,
+      "output": 0.28,
+      "cacheRead": 0.07,
+      "cacheWrite": 0
+    },
+    "contextWindow": 262100,
+    "maxTokens": 65536,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
+  },
+  {
+    "id": "moonshotai/Kimi-K3",
+    "name": "Kimi K3",
+    "reasoning": true,
+    "input": [
+      "text",
+      "image"
+    ],
+    "cost": {
+      "input": 3.6,
+      "output": 18,
+      "cacheRead": 1.8,
+      "cacheWrite": 0
+    },
+    "contextWindow": 1048576,
+    "maxTokens": 1048576,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
+  },
+  {
+    "id": "XiaomiMiMo/MiMo-V2.5",
+    "name": "MiMo-V2.5",
+    "reasoning": true,
+    "input": [
+      "text"
+    ],
+    "cost": {
+      "input": 0.1976,
+      "output": 0.6352,
+      "cacheRead": 0.0988,
+      "cacheWrite": 0
+    },
+    "contextWindow": 262144,
+    "maxTokens": 262144,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
+  },
   {
     "id": "zai-org/GLM-5.2",
     "name": "GLM 5.2",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 1.638,
-      "output": 5.13,
-      "cacheRead": 0.819,
+      "input": 1.96,
+      "output": 6.16,
+      "cacheRead": 0.98,
       "cacheWrite": 0
     },
     "contextWindow": 262144,
-    "maxTokens": 65536
+    "maxTokens": 65536,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "moonshotai/Kimi-K2.7-Code",
     "name": "Kimi K2.7 Code",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
-      "text"
+      "text",
+      "image"
     ],
     "cost": {
-      "input": 0.9,
-      "output": 3.9,
-      "cacheRead": 0.45,
+      "input": 1.072,
+      "output": 4.65,
+      "cacheRead": 0.536,
       "cacheWrite": 0
     },
     "contextWindow": 262144,
-    "maxTokens": 262144
+    "maxTokens": 262144,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "Qwen/Qwen3.6-35B-A3B",
     "name": "Qwen3.6 35B A3B",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
-      "text"
+      "text",
+      "image"
     ],
     "cost": {
-      "input": 0.1932,
-      "output": 1.12275,
-      "cacheRead": 0.0966,
+      "input": 0.1672,
+      "output": 1.11675,
+      "cacheRead": 0.0836,
       "cacheWrite": 0
     },
     "contextWindow": 262140,
-    "maxTokens": 262140
+    "maxTokens": 262140,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "Qwen/Qwen3.6-27B",
     "name": "Qwen3.6 27B",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
-      "text"
+      "text",
+      "image"
     ],
     "cost": {
-      "input": 0.4266,
-      "output": 3.21,
-      "cacheRead": 0.2133,
+      "input": 0.373,
+      "output": 3.19,
+      "cacheRead": 0.1865,
       "cacheWrite": 0
     },
     "contextWindow": 32768,
-    "maxTokens": 32768
+    "maxTokens": 32768,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "MiniMaxAI/MiniMax-M2.7",
     "name": "MiniMax M2.7",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.348,
-      "output": 1.332,
-      "cacheRead": 0.174,
+      "input": 0.436,
+      "output": 1.72,
+      "cacheRead": 0.218,
       "cacheWrite": 0
     },
-    "contextWindow": 204800,
-    "maxTokens": 204800
+    "contextWindow": 262100,
+    "maxTokens": 65536,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "deepseek-ai/DeepSeek-V4-Flash",
     "name": "DeepSeek V4 Flash",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
@@ -93,121 +193,169 @@
       "cacheWrite": 0
     },
     "contextWindow": 32768,
-    "maxTokens": 32768
+    "maxTokens": 32768,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "deepseek-ai/DeepSeek-V4-Pro",
     "name": "DeepSeek V4 Pro",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 1.652,
-      "output": 3.304,
-      "cacheRead": 0.826,
+      "input": 1.52044,
+      "output": 3.04088,
+      "cacheRead": 0.76022,
       "cacheWrite": 0
     },
     "contextWindow": 1048576,
-    "maxTokens": 600000
+    "maxTokens": 600000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "moonshotai/Kimi-K2.6",
     "name": "Kimi K2.6",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text",
       "image"
     ],
     "cost": {
-      "input": 0.95,
-      "output": 3.98,
-      "cacheRead": 0.475,
+      "input": 0.842,
+      "output": 3.762,
+      "cacheRead": 0.421,
       "cacheWrite": 0
     },
     "contextWindow": 262142,
-    "maxTokens": 262142
+    "maxTokens": 262142,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "zai-org/GLM-5.1",
     "name": "GLM 5.1",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 1.368,
-      "output": 4.468,
-      "cacheRead": 0.684,
+      "input": 1.2732,
+      "output": 4.1272,
+      "cacheRead": 0.6366,
       "cacheWrite": 0
     },
     "contextWindow": 202750,
-    "maxTokens": 32768
+    "maxTokens": 32768,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "MiniMaxAI/MiniMax-M2.5",
     "name": "MiniMax M2.5",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.281,
-      "output": 1.128,
-      "cacheRead": 0.1405,
+      "input": 0.287,
+      "output": 1.152,
+      "cacheRead": 0.1435,
       "cacheWrite": 0
     },
     "contextWindow": 196600,
-    "maxTokens": 196600
+    "maxTokens": 196600,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "moonshotai/Kimi-K2.5",
     "name": "Kimi K2.5",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text",
       "image"
     ],
     "cost": {
-      "input": 0.528,
-      "output": 2.79,
-      "cacheRead": 0.264,
+      "input": 0.564,
+      "output": 2.82,
+      "cacheRead": 0.282,
       "cacheWrite": 1.1
     },
     "contextWindow": 262144,
-    "maxTokens": 262144
+    "maxTokens": 262144,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "zai-org/GLM-5",
     "name": "GLM 5",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.84,
-      "output": 2.688,
-      "cacheRead": 0.42,
+      "input": 0.85,
+      "output": 2.622,
+      "cacheRead": 0.425,
       "cacheWrite": 0
     },
     "contextWindow": 202752,
-    "maxTokens": 202752
+    "maxTokens": 202752,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "deepseek-ai/DeepSeek-V3.2",
     "name": "DeepSeek V3.2",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.8934,
-      "output": 1.6336,
-      "cacheRead": 0.4467,
+      "input": 0.90054,
+      "output": 1.75646,
+      "cacheRead": 0.45027,
       "cacheWrite": 0.5
     },
     "contextWindow": 163840,
-    "maxTokens": 163840
+    "maxTokens": 163840,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "moonshotai/Kimi-K2-Thinking",
@@ -223,12 +371,18 @@
       "cacheWrite": 0.64
     },
     "contextWindow": 262144,
-    "maxTokens": 262144
+    "maxTokens": 262144,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "zai-org/GLM-4.5-Air",
     "name": "GLM-4.5-Air",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
@@ -239,28 +393,40 @@
       "cacheWrite": 0
     },
     "contextWindow": 131070,
-    "maxTokens": 131070
+    "maxTokens": 131070,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "google/gemma-4-26b-a4b-it",
     "name": "Gemma 4 26B A4B",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.11,
-      "output": 0.358,
-      "cacheRead": 0.055,
+      "input": 0.13,
+      "output": 0.43,
+      "cacheRead": 0.065,
       "cacheWrite": 0
     },
     "contextWindow": 262142,
-    "maxTokens": 262142
+    "maxTokens": 262142,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "zai-org/GLM-4.7-Flash",
     "name": "GLM 4.7 Flash",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
@@ -271,12 +437,18 @@
       "cacheWrite": 0.14
     },
     "contextWindow": 200000,
-    "maxTokens": 200000
+    "maxTokens": 200000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "zai-org/GLM-4.7",
     "name": "GLM 4.7",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
@@ -287,7 +459,13 @@
       "cacheWrite": 0.6
     },
     "contextWindow": 202752,
-    "maxTokens": 202752
+    "maxTokens": 202752,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "moonshotai/Kimi-K2-Instruct-0905",
@@ -303,7 +481,12 @@
       "cacheWrite": 0.78
     },
     "contextWindow": 262144,
-    "maxTokens": 262144
+    "maxTokens": 262144,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   {
     "id": "meta-llama/Llama-3.2-90B-Vision-Instruct",
@@ -320,23 +503,34 @@
       "cacheWrite": 0.7
     },
     "contextWindow": 16000,
-    "maxTokens": 16000
+    "maxTokens": 16000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   {
     "id": "openai/gpt-oss-120b",
     "name": "gpt-oss-120b",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.158,
-      "output": 0.592,
-      "cacheRead": 0.079,
+      "input": 0.188,
+      "output": 0.7,
+      "cacheRead": 0.094,
       "cacheWrite": 0.04
     },
     "contextWindow": 131072,
-    "maxTokens": 131072
+    "maxTokens": 131072,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "deepseek-ai/DeepSeek-R1-0528",
@@ -352,12 +546,18 @@
       "cacheWrite": 0.8
     },
     "contextWindow": 128000,
-    "maxTokens": 128000
+    "maxTokens": 128000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "zai-org/GLM-4.6",
     "name": "GLM 4.6",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
@@ -368,7 +568,13 @@
       "cacheWrite": 0.7
     },
     "contextWindow": 131072,
-    "maxTokens": 131072
+    "maxTokens": 131072,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
@@ -384,7 +590,12 @@
       "cacheWrite": 0.12
     },
     "contextWindow": 262144,
-    "maxTokens": 262144
+    "maxTokens": 262144,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   {
     "id": "Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar",
@@ -394,13 +605,18 @@
       "text"
     ],
     "cost": {
-      "input": 0.601,
-      "output": 2.085,
-      "cacheRead": 0.3005,
+      "input": 0.569,
+      "output": 2.135,
+      "cacheRead": 0.2845,
       "cacheWrite": 0.44
     },
     "contextWindow": 106000,
-    "maxTokens": 106000
+    "maxTokens": 106000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   {
     "id": "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8",
@@ -417,7 +633,12 @@
       "cacheWrite": 0.3
     },
     "contextWindow": 430000,
-    "maxTokens": 430000
+    "maxTokens": 430000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   {
     "id": "mistralai/Mistral-Nemo-Instruct-2407",
@@ -427,29 +648,40 @@
       "text"
     ],
     "cost": {
-      "input": 0.05675,
-      "output": 0.095,
-      "cacheRead": 0.028375,
+      "input": 0.069667,
+      "output": 0.116667,
+      "cacheRead": 0.034834,
       "cacheWrite": 0.04
     },
     "contextWindow": 128000,
-    "maxTokens": 128000
+    "maxTokens": 128000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   },
   {
     "id": "openai/gpt-oss-20b",
     "name": "gpt-oss-20b",
-    "reasoning": false,
+    "reasoning": true,
     "input": [
       "text"
     ],
     "cost": {
-      "input": 0.063,
-      "output": 0.226,
-      "cacheRead": 0.0315,
+      "input": 0.053,
+      "output": 0.216,
+      "cacheRead": 0.0265,
       "cacheWrite": 0.03
     },
     "contextWindow": 64000,
-    "maxTokens": 64000
+    "maxTokens": 64000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens",
+      "supportsReasoningEffort": true
+    }
   },
   {
     "id": "meta-llama/Llama-3.3-70B-Instruct",
@@ -459,12 +691,17 @@
       "text"
     ],
     "cost": {
-      "input": 0.6066,
-      "output": 1.0386,
-      "cacheRead": 0.3033,
+      "input": 0.5126,
+      "output": 1.0446,
+      "cacheRead": 0.2563,
       "cacheWrite": 0.2
     },
     "contextWindow": 128000,
-    "maxTokens": 128000
+    "maxTokens": 128000,
+    "compat": {
+      "supportsStore": false,
+      "supportsDeveloperRole": false,
+      "maxTokensField": "max_tokens"
+    }
   }
 ]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-io-provider",
-  "version": "1.1.1",
+  "version": "1.1.3",
   "description": "IO Intelligence provider extension for pi - Access DeepSeek, Kimi, GLM, Llama, Qwen, Mistral, and more through the IO Intelligence API",
   "type": "module",
   "main": "index.ts",

package/scripts/update-models.js

@@ -81,9 +81,17 @@ function convertModel(apiModel, existingModelsMap) {
     if (priceOut > 0) existing.cost.output = Math.round(priceOut * 1e6) / 1e6;
     if (cacheRead > 0) existing.cost.cacheRead = Math.round(cacheRead * 1e6) / 1e6;
     if (cacheWrite > 0) existing.cost.cacheWrite = Math.round(cacheWrite * 1e6) / 1e6;
-    if (apiModel.supports_images_input && !existing.input.includes('image')) {
-      existing.input = ['text', 'image'];
-    }
+    const hasVision = apiModel.supports_images_input === true || (apiModel.input_modalities || []).includes('image');
+    const hasReasoning = apiModel.capabilities?.reasoning === true || apiModel.supports_reasoning === true;
+    existing.input = hasVision ? ['text', 'image'] : ['text'];
+    existing.reasoning = hasReasoning;
+    existing.compat = {
+      ...(existing.compat || {}),
+      supportsStore: false,
+      supportsDeveloperRole: false,
+      maxTokensField: 'max_tokens',
+      ...(hasReasoning ? { supportsReasoningEffort: true } : {}),
+    };
     return existing;
   }
 
@@ -91,7 +99,7 @@ function convertModel(apiModel, existingModelsMap) {
   const ctx = apiModel.context_window || 0;
   const maxTok = apiModel.max_tokens || ctx;
   const input = ['text'];
-  if (apiModel.supports_images_input) input.push('image');
+  if (apiModel.supports_images_input === true || (apiModel.input_modalities || []).includes('image')) input.push('image');
 
   const priceIn = (apiModel.input_token_price || 0) * 1_000_000;
   const priceOut = (apiModel.output_token_price || 0) * 1_000_000;
@@ -101,7 +109,7 @@ function convertModel(apiModel, existingModelsMap) {
   return {
     id,
     name: cleanName(apiModel.name, id),
-    reasoning: false,
+    reasoning: apiModel.capabilities?.reasoning === true || apiModel.supports_reasoning === true,
     input,
     cost: {
       input: Math.round(priceIn * 1e6) / 1e6,
@@ -111,6 +119,14 @@ function convertModel(apiModel, existingModelsMap) {
     },
     contextWindow: ctx,
     maxTokens: maxTok,
+    compat: {
+      supportsStore: false,
+      supportsDeveloperRole: false,
+      maxTokensField: 'max_tokens',
+      ...((apiModel.capabilities?.reasoning === true || apiModel.supports_reasoning === true)
+        ? { supportsReasoningEffort: true }
+        : {}),
+    },
   };
 }
 
@@ -123,6 +139,7 @@ function applyPatch(model, patch) {
   if (patch.input !== undefined) result.input = patch.input;
   if (patch.contextWindow !== undefined) result.contextWindow = patch.contextWindow;
   if (patch.maxTokens !== undefined) result.maxTokens = patch.maxTokens;
+  if (patch.thinkingLevelMap !== undefined) result.thinkingLevelMap = { ...patch.thinkingLevelMap };
   if (patch.cost) {
     result.cost = {
       input: patch.cost.input ?? result.cost.input,
@@ -137,6 +154,9 @@ function applyPatch(model, patch) {
   if (!result.reasoning && result.compat?.thinkingFormat) {
     delete result.compat.thinkingFormat;
   }
+  if (!result.reasoning && result.thinkingLevelMap) {
+    delete result.thinkingLevelMap;
+  }
   if (result.compat && Object.keys(result.compat).length === 0) {
     delete result.compat;
   }
@@ -328,6 +348,67 @@ MIT
 
 // ─── Main ────────────────────────────────────────────────────────────────────
 
+// Grace period for delisted models: update-models.js moves models the API no
+// longer lists into deprecated-models.json (stamped with deprecatedAt) instead
+// of dropping them; the runtime appends them back so sessions and saved model
+// settings keep working, and after 14 days they are evicted permanently.
+const DEPRECATED_MODEL_TTL_MS = 14 * 24 * 60 * 60 * 1000;
+
+/**
+ * Reconcile deprecated-models.json against the freshly fetched model list.
+ * - in old models.json but not the API: moved into the deprecated file
+ *   (deprecatedAt = now; preserved on repeat runs so the grace clock is not reset)
+ * - back in the API: resurrected (dropped from the deprecated file)
+ * - deprecatedAt older than 14 days: evicted permanently
+ * Must run BEFORE the new models.json is written; it reads the old file itself.
+ */
+function updateDeprecatedModels(modelsJsonPath, newModels) {
+  const deprecatedPath = path.join(path.dirname(modelsJsonPath), 'deprecated-models.json');
+
+  let oldModels = [];
+  try {
+    const parsed = JSON.parse(fs.readFileSync(modelsJsonPath, 'utf8'));
+    if (Array.isArray(parsed)) oldModels = parsed;
+  } catch { /* first run: no previous models.json */ }
+
+  let deprecated = {};
+  try {
+    const parsed = JSON.parse(fs.readFileSync(deprecatedPath, 'utf8'));
+    if (parsed && typeof parsed === 'object' && !Array.isArray(parsed)) deprecated = parsed;
+  } catch { /* no graveyard yet */ }
+
+  const currentIds = new Set(newModels.map(m => m.id));
+  const now = new Date().toISOString();
+  const added = [];
+  const resurrected = [];
+  const evicted = [];
+
+  for (const old of oldModels) {
+    if (old && old.id && !currentIds.has(old.id) && !deprecated[old.id]) {
+      deprecated[old.id] = { ...old, deprecatedAt: now };
+      added.push(old.id);
+    }
+  }
+
+  for (const [id, entry] of Object.entries(deprecated)) {
+    if (currentIds.has(id)) {
+      delete deprecated[id];
+      resurrected.push(id);
+      continue;
+    }
+    const removedAt = Date.parse(entry && entry.deprecatedAt ? entry.deprecatedAt : '');
+    if (Number.isNaN(removedAt) || Date.now() - removedAt > DEPRECATED_MODEL_TTL_MS) {
+      delete deprecated[id];
+      evicted.push(id);
+    }
+  }
+
+  if (added.length > 0 || resurrected.length > 0 || evicted.length > 0) {
+    fs.writeFileSync(deprecatedPath, JSON.stringify(deprecated, null, 2) + '\n');
+    console.log('Updated deprecated-models.json ' + JSON.stringify({ added, resurrected, evicted }));
+  }
+}
+
 async function main() {
   const apiKey = process.env.IOINTELLIGENCE_API_KEY;
   if (!apiKey) {
@@ -362,6 +443,8 @@ async function main() {
     console.log(`Converted ${models.length} models`);
 
     // Save models.json (pure API output, no patch/custom baked in)
+    // Move delisted models to deprecated-models.json BEFORE models.json is overwritten
+    updateDeprecatedModels(MODELS_PATH, models);
     fs.writeFileSync(MODELS_PATH, JSON.stringify(models, null, 2) + '\n');
     console.log(`✓ Saved ${models.length} models to models.json`);