@aliou/pi-neuralwatt 0.11.4 → 0.12.1

package/extensions/provider/index.ts

@@ -199,6 +199,13 @@ export default async function (pi: ExtensionAPI) {
     return { message: overflowMessage };
   });
 
+  // Inject the active Pi session id as a conversation id on every Neuralwatt
+  // request so the gateway can correlate requests within a session.
+  pi.on("before_provider_headers", (event, ctx) => {
+    if (ctx.model?.provider !== "neuralwatt") return;
+    event.headers["X-NW-Conversation-ID"] = ctx.sessionManager.getSessionId();
+  });
+
   pi.on("after_provider_response", (event, ctx) => {
     if (ctx.model?.provider !== "neuralwatt") return;
 

package/extensions/provider/models/public-models.ts

@@ -56,10 +56,26 @@ const DEEPSEEK_V4_FLASH: NeuralwattModelFamily = {
   },
 };
 
-// Google, served from NVIDIA's NVFP4 checkpoint.
+// Google, served from NVIDIA's NVFP4 checkpoint. Gemma 4's chat template
+// takes a boolean rather than an effort level, so it has a single reasoning
+// depth (`max`) plus thinking-off; every non-`none` value resolves to `max`.
+// It does not reason by default (`default_enabled: false`), but the model
+// can produce reasoning traces when asked. See
+// https://portal.neuralwatt.com/docs/api/chat-completions#reasoning-effort
+const GEMMA_4_THINKING: ThinkingLevelMap = {
+  off: "none",
+  minimal: null,
+  low: null,
+  medium: null,
+  high: null,
+  xhigh: null,
+  max: "max",
+};
+
 const GEMMA_4: NeuralwattModelFamily = {
   cost: { input: 0.144, output: 0.42, cacheRead: 0.0144 },
   vision: true,
+  thinkingLevelMap: GEMMA_4_THINKING,
 };
 
 // ZhipuAI.
@@ -132,7 +148,7 @@ const FAMILIES: [NeuralwattModelFamily, NeuralwattVariantSpec[]][] = [
         name: "Gemma 4 31B",
         contextWindow: 262128,
         maxOutputTokens: 16384,
-        reasoning: false,
+        reasoning: true,
       },
     ],
   ],
@@ -147,11 +163,14 @@ const FAMILIES: [NeuralwattModelFamily, NeuralwattVariantSpec[]][] = [
         reasoning: true,
       },
       {
+        // GLM-5.2 Fast pins thinking off by default, but keeps the parent's
+        // full reasoning contract (`high`/`max`/`none`): sending
+        // `reasoning_effort` re-enables thinking for that request.
         id: "glm-5.2-fast",
-        name: "GLM-5.2 Fast",
+        name: "GLM-5.2 (fast)",
         contextWindow: 1048560,
         maxOutputTokens: null,
-        reasoning: false,
+        reasoning: true,
       },
       {
         id: "glm-5.2-flex",
@@ -169,11 +188,13 @@ const FAMILIES: [NeuralwattModelFamily, NeuralwattVariantSpec[]][] = [
         reasoning: true,
       },
       {
+        // Short/fast: pins thinking off but keeps the parent reasoning
+        // contract, like glm-5.2-fast.
         id: "glm-5.2-short-fast",
-        name: "GLM-5.2 Short Fast",
+        name: "GLM-5.2 (short, fast)",
         contextWindow: 199984,
         maxOutputTokens: 32000,
-        reasoning: false,
+        reasoning: true,
       },
       {
         id: "glm-5.2-short-flex",
@@ -184,11 +205,13 @@ const FAMILIES: [NeuralwattModelFamily, NeuralwattVariantSpec[]][] = [
         costMultiplier: FLEX_COST_MULTIPLIER,
       },
       {
+        // Short/fast/flex: pins thinking off but keeps the parent reasoning
+        // contract, like glm-5.2-fast.
         id: "glm-5.2-short-fast-flex",
         name: "GLM-5.2 (short, fast, flex)",
         contextWindow: 199984,
         maxOutputTokens: 32000,
-        reasoning: false,
+        reasoning: true,
         costMultiplier: FLEX_COST_MULTIPLIER,
       },
     ],
@@ -210,6 +233,14 @@ const FAMILIES: [NeuralwattModelFamily, NeuralwattVariantSpec[]][] = [
         maxOutputTokens: null,
         reasoning: false,
       },
+      {
+        id: "kimi-k3-flex",
+        name: "Kimi K3 (flex)",
+        contextWindow: 1048560,
+        maxOutputTokens: null,
+        reasoning: true,
+        costMultiplier: FLEX_COST_MULTIPLIER,
+      },
     ],
   ],
   [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aliou/pi-neuralwatt",
-  "version": "0.11.4",
+  "version": "0.12.1",
   "license": "MIT",
   "type": "module",
   "private": false,
@@ -34,7 +34,7 @@
     "!extensions/**/*.test.ts"
   ],
   "dependencies": {
-    "@aliou/pi-utils-settings": "^0.18.0",
+    "@aliou/pi-utils-settings": "^0.19.1",
     "@aliou/pi-utils-ui": "^0.5.0"
   },
   "peerDependencies": {
@@ -67,8 +67,8 @@
     "format": "biome check --write",
     "test": "vitest run",
     "test:watch": "vitest",
-    "gen:schema": "ts-json-schema-generator --path src/config/types.ts --type NeuralwattConfig --no-type-check -o schema.json",
-    "check:schema": "ts-json-schema-generator --path src/config/types.ts --type NeuralwattConfig --no-type-check -o /tmp/schema-check.json && diff -q schema.json /tmp/schema-check.json",
+    "gen:schema": "pi-settings-schema -p src/config/types.ts -t NeuralwattConfig -o schema.json --version 0.12.0",
+    "check:schema": "pi-settings-schema -p src/config/types.ts -t NeuralwattConfig -o schema.json --version 0.12.0 --check",
     "check:lockfile": "pnpm install --frozen-lockfile --ignore-scripts",
     "prepare": "[ -d .git ] && husky || true",
     "changeset": "changeset",

package/schema.json

@@ -1,12 +1,11 @@
 {
-  "$ref": "#/definitions/NeuralwattConfig",
   "$schema": "http://json-schema.org/draft-07/schema#",
+  "$ref": "#/definitions/NeuralwattConfig",
   "definitions": {
     "NeuralwattConfig": {
-      "additionalProperties": false,
+      "type": "object",
       "properties": {
         "$schema": {
-          "description": "$schema URL for editor autocomplete.",
           "type": "string"
         },
         "provider": {
@@ -24,57 +23,70 @@
         "subBarIntegration": {
           "$ref": "#/definitions/NeuralwattSubBarIntegrationConfig",
           "description": "Sub-bar/status-bar integration feature."
+        },
+        "version": {
+          "anyOf": [
+            {
+              "type": "integer",
+              "minimum": 0
+            },
+            {
+              "type": "string",
+              "pattern": "^\\d{1,15}(\\.\\d{1,15})?(\\.\\d{1,15})?$"
+            }
+          ],
+          "description": "Config schema version, stamped by migrations. Current version: 0.12.0."
         }
       },
-      "type": "object"
+      "additionalProperties": false
     },
     "NeuralwattProviderConfig": {
-      "additionalProperties": false,
+      "type": "object",
       "properties": {
+        "includeLegacyModelIds": {
+          "type": "boolean",
+          "description": "Include legacy Neuralwatt model IDs in the model picker."
+        },
         "includeAliasedModelIds": {
-          "description": "Include alternate creator-scoped Neuralwatt model IDs in the model picker.",
-          "type": "boolean"
+          "type": "boolean",
+          "description": "Include alternate creator-scoped Neuralwatt model IDs in the model picker."
         },
         "includeEarlyAccessModels": {
-          "description": "Include early-access Neuralwatt models discovered via the authenticated API.",
-          "type": "boolean"
-        },
-        "includeLegacyModelIds": {
-          "description": "Include legacy Neuralwatt model IDs in the model picker.",
-          "type": "boolean"
+          "type": "boolean",
+          "description": "Include early-access Neuralwatt models discovered via the authenticated API."
         }
       },
-      "type": "object"
+      "additionalProperties": false
     },
     "NeuralwattQuotaCommandConfig": {
-      "additionalProperties": false,
+      "type": "object",
       "properties": {
         "enabled": {
-          "description": "Show the quota command (/neuralwatt:quota).",
-          "type": "boolean"
+          "type": "boolean",
+          "description": "Show the quota command (/neuralwatt:quota)."
         }
       },
-      "type": "object"
+      "additionalProperties": false
     },
     "NeuralwattQuotaWarningsConfig": {
-      "additionalProperties": false,
+      "type": "object",
       "properties": {
         "enabled": {
-          "description": "Show quota warnings when credits or energy are low.",
-          "type": "boolean"
+          "type": "boolean",
+          "description": "Show quota warnings when credits or energy are low."
         }
       },
-      "type": "object"
+      "additionalProperties": false
     },
     "NeuralwattSubBarIntegrationConfig": {
-      "additionalProperties": false,
+      "type": "object",
       "properties": {
         "enabled": {
-          "description": "Show usage in the sub-bar / status bar.",
-          "type": "boolean"
+          "type": "boolean",
+          "description": "Show usage in the sub-bar / status bar."
         }
       },
-      "type": "object"
+      "additionalProperties": false
     }
   }
-}
+}

package/src/config/migration/01-disable-legacy-model-ids-by-default.ts

@@ -29,6 +29,7 @@ function isPreviousConfigWithoutLegacyDefault(
 export const disableLegacyModelIdsByDefaultMigration: Migration<NeuralwattConfig> =
   {
     name: "disable-legacy-model-ids-by-default",
+    version: "0.8.0",
     shouldRun: isPreviousConfigWithoutLegacyDefault,
     message:
       "[neuralwatt] legacy model IDs (ids including the provider and the quantization) are disabled by default. You can enable them with /neuralwatt:settings.",

package/src/config/migration/02-flat-to-nested-config.ts

@@ -80,6 +80,7 @@ export async function backupConfig(filePath: string): Promise<void> {
 
 export const flatToNestedConfigMigration: Migration<NeuralwattConfig> = {
   name: "flat-to-nested-config",
+  version: "0.8.1",
   shouldRun: isPreviousConfig,
   message: FLAT_CONFIG_MIGRATION_MESSAGE,
   run: async (config, filePath) => {

package/src/config/migration/03-rename-hidden-to-early-access.ts

@@ -25,6 +25,7 @@ function previousProvider(
  */
 export const renameHiddenToEarlyAccessMigration: Migration<NeuralwattConfig> = {
   name: "rename-hidden-models-to-early-access",
+  version: "0.10.6",
   shouldRun: (config) =>
     previousProvider(config)?.includeHiddenModels !== undefined,
   message:

package/src/config/migration/04-enable-aliases-for-legacy-users.ts

@@ -25,6 +25,7 @@ function previousProvider(
 export const enableAliasesForLegacyUsersMigration: Migration<NeuralwattConfig> =
   {
     name: "enable-alias-model-ids-for-legacy-users",
+    version: "0.11.0",
     shouldRun: (config) => {
       const provider = previousProvider(config);
       return (