@aliou/pi-neuralwatt 0.4.2 → 0.5.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aliou/pi-neuralwatt",
-  "version": "0.4.2",
+  "version": "0.5.1",
   "license": "MIT",
   "type": "module",
   "private": false,
@@ -28,23 +28,23 @@
   "files": [
     "src",
     "schema.json",
-    "README.md"
+    "README.md",
+    "!src/**/*.test.ts"
   ],
   "dependencies": {
     "@aliou/pi-utils-settings": "^0.15.0",
     "@aliou/pi-utils-ui": "^0.4.0"
   },
   "peerDependencies": {
-    "@earendil-works/pi-coding-agent": "0.74.0",
-    "@earendil-works/pi-tui": "0.74.0",
-    "@sinclair/typebox": ">=0.34.0"
+    "@earendil-works/pi-coding-agent": "*",
+    "@earendil-works/pi-tui": "*"
   },
   "devDependencies": {
     "@aliou/biome-plugins": "^0.8.1",
-    "@biomejs/biome": "^2.4.12",
+    "@biomejs/biome": "^2.4.15",
     "@changesets/cli": "^2.27.11",
-    "@earendil-works/pi-coding-agent": "0.74.0",
-    "@earendil-works/pi-tui": "0.74.0",
+    "@earendil-works/pi-coding-agent": "0.77.0",
+    "@earendil-works/pi-tui": "0.77.0",
     "@types/node": "^25.0.10",
     "husky": "^9.1.7",
     "ts-json-schema-generator": "^2.4.0",
@@ -57,9 +57,6 @@
     },
     "@earendil-works/pi-tui": {
       "optional": true
-    },
-    "@sinclair/typebox": {
-      "optional": true
     }
   },
   "scripts": {

package/src/extensions/provider/index.ts

@@ -8,45 +8,37 @@ import {
   registerNeuralwattSettings,
 } from "../../config";
 import { getNeuralwattApiKey } from "../../lib/env";
-import { fetchModels } from "../../lib/fetch-models";
 import type { NeuralwattQuotas } from "../../types/quota-api";
 import {
   NEURALWATT_QUOTAS_REQUEST_EVENT,
   NEURALWATT_QUOTAS_UPDATED_EVENT,
   type NeuralwattQuotasUpdatedPayload,
 } from "../../types/quota-events";
-import { isOffline } from "../../utils/is-offline";
 import { fetchQuotas } from "../../utils/quotas";
 import { normalizeNeuralwattContextOverflowError } from "./context-overflow";
-import type { NeuralwattModelConfig } from "./models";
-import { NEURALWATT_MODELS_CACHE } from "./models";
-import { buildModelsPayload } from "./provider-payload";
+import { NEURALWATT_MODELS } from "./models";
 import { buildQuotasFromHeaders, fetchRequestedQuotas } from "./quota-store";
 
 const HEADER_EMIT_THROTTLE_MS = 5_000;
 
-function registerNeuralwattProvider(
-  pi: ExtensionAPI,
-  models: NeuralwattModelConfig[],
-): void {
+function registerNeuralwattProvider(pi: ExtensionAPI): void {
   pi.registerProvider("neuralwatt", {
     baseUrl: "https://api.neuralwatt.com/v1",
-    apiKey: "NEURALWATT_API_KEY",
+    apiKey: "$NEURALWATT_API_KEY",
     api: "openai-completions",
     authHeader: true,
     headers: {
       Referer: "https://pi.dev",
       "X-Title": "npm:@aliou/pi-neuralwatt",
     },
-    models: buildModelsPayload(models),
+    models: NEURALWATT_MODELS,
   });
 }
 
 export default async function (pi: ExtensionAPI) {
   await configLoader.load();
 
-  // Register with hardcoded cache immediately so models are available on startup
-  registerNeuralwattProvider(pi, NEURALWATT_MODELS_CACHE);
+  registerNeuralwattProvider(pi);
 
   const loadedFeatures = new Set<NeuralwattFeatureId>();
 
@@ -110,28 +102,6 @@ export default async function (pi: ExtensionAPI) {
     pi.events.emit(NEURALWATT_EXTENSIONS_REQUEST_EVENT, undefined);
     emitConfigUpdated(pi);
 
-    if (!isOffline()) {
-      const result = await fetchModels();
-      if (result.success) {
-        const cacheIds = new Set(NEURALWATT_MODELS_CACHE.map((m) => m.id));
-        const liveIds = new Set(result.models.map((m) => m.id));
-        const added = result.models.filter((m) => !cacheIds.has(m.id));
-        const removed = NEURALWATT_MODELS_CACHE.filter(
-          (m) => !liveIds.has(m.id),
-        );
-        if (added.length > 0 || removed.length > 0) {
-          const parts: string[] = [];
-          if (added.length > 0) parts.push(`${added.length} new`);
-          if (removed.length > 0) parts.push(`${removed.length} removed`);
-          ctx.ui.notify(
-            `Neuralwatt models updated (${parts.join(", ")})`,
-            "info",
-          );
-        }
-        registerNeuralwattProvider(pi, result.models);
-      }
-    }
-
     if (ctx.model?.provider !== "neuralwatt") return;
     const apiKey = await getNeuralwattApiKey(ctx.modelRegistry.authStorage);
     if (!apiKey) return;

package/src/extensions/provider/models.ts

@@ -1,27 +1,10 @@
 // Hardcoded models from Neuralwatt API
 // Source: https://api.neuralwatt.com/v1/models
-// Pricing: https://portal.neuralwatt.com/pricing
-// max_model_len from /v1/models, pricing from /pricing page
+// Pricing, capabilities, and limits from metadata fields in /v1/models
 
 import type { ProviderModelConfig } from "@earendil-works/pi-coding-agent";
 
-export interface NeuralwattModelConfig extends ProviderModelConfig {
-  /** Fast variant of a parent model (e.g. "glm-5-fast" is the fast variant of "zai-org/GLM-5.1-FP8"). */
-  fast?: boolean;
-}
-
-const NEURALWATT_BINARY_THINKING_LEVEL_MAP = {
-  minimal: null,
-  low: null,
-  medium: "medium",
-  high: null,
-  xhigh: null,
-} as const;
-
-/** Hardcoded model cache. Used as a fallback on startup before live models are fetched.
- * Updated from https://api.neuralwatt.com/v1/models and https://portal.neuralwatt.com/pricing
- */
-export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
+export const NEURALWATT_MODELS: ProviderModelConfig[] = [
   // Devstral Small 2 - Mistral
   {
     id: "mistralai/Devstral-Small-2-24B-Instruct-2512",
@@ -46,7 +29,6 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     id: "glm-5-fast",
     name: "GLM-5 Fast",
     reasoning: false,
-    fast: true,
     input: ["text"],
     cost: {
       input: 1.1,
@@ -75,10 +57,17 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     },
     contextWindow: 202736,
     maxTokens: 32768,
-    thinkingLevelMap: NEURALWATT_BINARY_THINKING_LEVEL_MAP,
+    thinkingLevelMap: {
+      minimal: null,
+      low: null,
+      medium: "medium",
+      high: null,
+      xhigh: null,
+    },
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
   // GLM-5.1 Fast - ZhipuAI
@@ -86,7 +75,6 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     id: "glm-5.1-fast",
     name: "GLM-5.1 Fast",
     reasoning: false,
-    fast: true,
     input: ["text"],
     cost: {
       input: 1.1,
@@ -125,6 +113,7 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
   // Kimi K2.5 - MoonshotAI
@@ -141,10 +130,17 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     },
     contextWindow: 262128,
     maxTokens: 65536,
-    thinkingLevelMap: NEURALWATT_BINARY_THINKING_LEVEL_MAP,
+    thinkingLevelMap: {
+      minimal: null,
+      low: null,
+      medium: "medium",
+      high: null,
+      xhigh: null,
+    },
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
   // Kimi K2.5 Fast - MoonshotAI
@@ -152,7 +148,6 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     id: "kimi-k2.5-fast",
     name: "Kimi K2.5 Fast",
     reasoning: false,
-    fast: true,
     input: ["text", "image"],
     cost: {
       input: 0.52,
@@ -181,10 +176,17 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     },
     contextWindow: 262128,
     maxTokens: 65536,
-    thinkingLevelMap: NEURALWATT_BINARY_THINKING_LEVEL_MAP,
+    thinkingLevelMap: {
+      minimal: null,
+      low: null,
+      medium: "medium",
+      high: null,
+      xhigh: null,
+    },
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
   // Kimi K2.6 Fast - MoonshotAI
@@ -192,7 +194,6 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     id: "kimi-k2.6-fast",
     name: "Kimi K2.6 Fast",
     reasoning: false,
-    fast: true,
     input: ["text", "image"],
     cost: {
       input: 0.69,
@@ -221,10 +222,17 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     },
     contextWindow: 196592,
     maxTokens: 65536,
-    thinkingLevelMap: NEURALWATT_BINARY_THINKING_LEVEL_MAP,
+    thinkingLevelMap: {
+      minimal: null,
+      low: null,
+      medium: "medium",
+      high: null,
+      xhigh: null,
+    },
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
   // Qwen3.5 397B - Qwen
@@ -241,10 +249,17 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     },
     contextWindow: 262128,
     maxTokens: 65536,
-    thinkingLevelMap: NEURALWATT_BINARY_THINKING_LEVEL_MAP,
+    thinkingLevelMap: {
+      minimal: null,
+      low: null,
+      medium: "medium",
+      high: null,
+      xhigh: null,
+    },
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
   // Qwen3.5 397B Fast - Qwen
@@ -252,7 +267,6 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     id: "qwen3.5-397b-fast",
     name: "Qwen3.5 397B Fast",
     reasoning: false,
-    fast: true,
     input: ["text"],
     cost: {
       input: 0.69,
@@ -281,18 +295,24 @@ export const NEURALWATT_MODELS_CACHE: NeuralwattModelConfig[] = [
     },
     contextWindow: 131056,
     maxTokens: 32768,
-    thinkingLevelMap: NEURALWATT_BINARY_THINKING_LEVEL_MAP,
+    thinkingLevelMap: {
+      minimal: null,
+      low: null,
+      medium: "medium",
+      high: null,
+      xhigh: null,
+    },
     compat: {
       supportsDeveloperRole: false,
       maxTokensField: "max_tokens",
+      requiresReasoningContentOnAssistantMessages: true,
     },
   },
-  // Qwen3.6 35B Fast (qwen3.6-35b-fast) - Qwen
+  // Qwen3.6 35B Fast - Qwen
   {
     id: "qwen3.6-35b-fast",
     name: "Qwen3.6 35B Fast",
     reasoning: false,
-    fast: true,
     input: ["text", "image"],
     cost: {
       input: 0.29,

package/src/extensions/provider/models.test.ts

@@ -1,347 +0,0 @@
-import { describe, expect, it } from "vitest";
-import type {
-  ApiModel as FullApiModel,
-  ApiResponse as FullApiResponse,
-} from "../../lib/fetch-models";
-import { mapApiModel } from "../../lib/fetch-models";
-import { NEURALWATT_MODELS_CACHE } from "./models";
-
-interface Discrepancy {
-  model: string;
-  field: string;
-  hardcoded: unknown;
-  api: unknown;
-}
-
-async function fetchApiModels(): Promise<FullApiModel[]> {
-  const apiKey = process.env.NEURALWATT_API_KEY;
-  const headers: Record<string, string> = {
-    "Content-Type": "application/json",
-    Referer: "https://github.com/aliou/pi-neuralwatt",
-  };
-  if (apiKey) {
-    headers.Authorization = `Bearer ${apiKey}`;
-  }
-
-  const response = await fetch("https://api.neuralwatt.com/v1/models", {
-    headers,
-  });
-
-  if (!response.ok) {
-    throw new Error(
-      `API request failed: ${response.status} ${response.statusText}`,
-    );
-  }
-
-  const data: FullApiResponse = await response.json();
-  return data.data;
-}
-
-function compareModels(
-  apiModels: FullApiModel[],
-  hardcodedModels: typeof NEURALWATT_MODELS_CACHE,
-): Discrepancy[] {
-  const discrepancies: Discrepancy[] = [];
-
-  for (const hardcoded of hardcodedModels) {
-    const apiModel = apiModels.find((m) => m.id === hardcoded.id);
-
-    if (!apiModel) {
-      discrepancies.push({
-        model: hardcoded.id,
-        field: "exists",
-        hardcoded: true,
-        api: false,
-      });
-      continue;
-    }
-
-    // Check context window
-    if (apiModel.max_model_len !== hardcoded.contextWindow) {
-      discrepancies.push({
-        model: hardcoded.id,
-        field: "contextWindow",
-        hardcoded: hardcoded.contextWindow,
-        api: apiModel.max_model_len,
-      });
-    }
-
-    // Check metadata-driven fields if available
-    const meta = apiModel.metadata;
-    if (meta) {
-      // Check reasoning
-      if (meta.capabilities.reasoning !== hardcoded.reasoning) {
-        discrepancies.push({
-          model: hardcoded.id,
-          field: "reasoning",
-          hardcoded: hardcoded.reasoning,
-          api: meta.capabilities.reasoning,
-        });
-      }
-
-      // Check pricing
-      if (meta.pricing.input_per_million !== hardcoded.cost.input) {
-        discrepancies.push({
-          model: hardcoded.id,
-          field: "cost.input",
-          hardcoded: hardcoded.cost.input,
-          api: meta.pricing.input_per_million,
-        });
-      }
-      if (meta.pricing.output_per_million !== hardcoded.cost.output) {
-        discrepancies.push({
-          model: hardcoded.id,
-          field: "cost.output",
-          hardcoded: hardcoded.cost.output,
-          api: meta.pricing.output_per_million,
-        });
-      }
-
-      // Check vision
-      const hasVision = hardcoded.input.includes("image");
-      if (meta.capabilities.vision !== hasVision) {
-        discrepancies.push({
-          model: hardcoded.id,
-          field: "input (vision)",
-          hardcoded: hasVision,
-          api: meta.capabilities.vision,
-        });
-      }
-    }
-  }
-
-  // Check for API models not in hardcoded list
-  for (const apiModel of apiModels) {
-    if (apiModel.metadata?.deprecated || apiModel.metadata?.pricing.pricing_tbd)
-      continue;
-    const hardcoded = hardcodedModels.find((m) => m.id === apiModel.id);
-    if (!hardcoded) {
-      discrepancies.push({
-        model: apiModel.id,
-        field: "exists",
-        hardcoded: false,
-        api: true,
-      });
-    }
-  }
-
-  return discrepancies;
-}
-
-describe("Neuralwatt models", () => {
-  it("should match API model definitions", { timeout: 30000 }, async () => {
-    const apiModels = await fetchApiModels();
-    const discrepancies = compareModels(apiModels, NEURALWATT_MODELS_CACHE);
-
-    if (discrepancies.length > 0) {
-      console.error("\nModel discrepancies found:");
-      console.error("==========================");
-      for (const d of discrepancies) {
-        if (d.field === "exists") {
-          if (d.hardcoded) {
-            console.error(`  ${d.model}: Missing from API`);
-          } else {
-            console.error(`  ${d.model}: Missing from hardcoded models (NEW)`);
-          }
-        } else {
-          console.error(`  ${d.model}.${d.field}:`);
-          console.error(`    hardcoded: ${JSON.stringify(d.hardcoded)}`);
-          console.error(`    api:       ${JSON.stringify(d.api)}`);
-        }
-      }
-      console.error("==========================\n");
-    }
-
-    expect(discrepancies).toHaveLength(0);
-  });
-
-  it("should map API models with metadata correctly", () => {
-    // Simulate a reasoning model with reasoning_effort support (like gpt-oss-20b)
-    const apiModelWithEffort: FullApiModel = {
-      id: "openai/gpt-oss-20b",
-      object: "model",
-      created: 1777467968,
-      owned_by: "vllm",
-      root: "openai/gpt-oss-20b",
-      parent: null,
-      max_model_len: 16384,
-      metadata: {
-        display_name: "GPT-OSS 20B",
-        description: "OpenAI GPT-OSS 20B",
-        provider: "OpenAI",
-        huggingface_id: null,
-        pricing: {
-          input_per_million: 0.03,
-          output_per_million: 0.16,
-          cached_input_per_million: null,
-          cached_output_per_million: null,
-          currency: "USD",
-          pricing_tbd: false,
-        },
-        capabilities: {
-          tools: true,
-          json_mode: true,
-          vision: false,
-          reasoning: true,
-          reasoning_effort: true,
-          streaming: true,
-          system_role: true,
-          developer_role: false,
-        },
-        limits: {
-          max_context_length: 16384,
-          max_output_tokens: 4096,
-          max_images: null,
-        },
-        deprecated: false,
-        deprecated_message: null,
-      },
-    };
-
-    const result = mapApiModel(apiModelWithEffort);
-    expect(result.id).toBe("openai/gpt-oss-20b");
-    expect(result.name).toBe("GPT-OSS 20B");
-    expect(result.reasoning).toBe(true);
-    expect(result.contextWindow).toBe(16384);
-    expect(result.maxTokens).toBe(4096);
-    expect(result.input).toEqual(["text"]);
-    expect(result.cost.input).toBe(0.03);
-    expect(result.cost.output).toBe(0.16);
-    expect(result.thinkingLevelMap).toEqual({
-      minimal: "low",
-      low: "low",
-      medium: "medium",
-      high: "high",
-      xhigh: null,
-    });
-    expect(result.fast).toBeUndefined();
-  });
-
-  it("should map fast variants correctly", () => {
-    // Simulate a fast variant (owned by "neuralwatt")
-    const fastModel: FullApiModel = {
-      id: "qwen3.6-35b-fast",
-      object: "model",
-      created: 0,
-      owned_by: "neuralwatt",
-      max_model_len: 131072,
-      metadata: {
-        display_name: "Qwen3.6 35B Fast",
-        description: "Fast variant",
-        provider: "Qwen",
-        huggingface_id: null,
-        pricing: {
-          input_per_million: 0.05,
-          output_per_million: 0.1,
-          cached_input_per_million: null,
-          cached_output_per_million: null,
-          currency: "USD",
-          pricing_tbd: false,
-        },
-        capabilities: {
-          tools: true,
-          json_mode: true,
-          vision: false,
-          reasoning: false,
-          reasoning_effort: false,
-          streaming: true,
-          system_role: true,
-          developer_role: false,
-        },
-        limits: {
-          max_context_length: 131072,
-          max_output_tokens: null,
-          max_images: null,
-        },
-        deprecated: false,
-        deprecated_message: null,
-      },
-    };
-
-    const result = mapApiModel(fastModel);
-    expect(result.id).toBe("qwen3.6-35b-fast");
-    expect(result.fast).toBe(true);
-    expect(result.reasoning).toBe(false);
-    expect(
-      (result.compat as Record<string, unknown>)?.supportsReasoningEffort,
-    ).toBeUndefined();
-  });
-
-  it("should map vision models correctly", () => {
-    const visionModel: FullApiModel = {
-      id: "moonshotai/Kimi-K2.6",
-      object: "model",
-      created: 1777467965,
-      owned_by: "vllm",
-      root: "moonshotai/Kimi-K2.6",
-      parent: null,
-      max_model_len: 262144,
-      metadata: {
-        display_name: "Kimi K2.6",
-        description: "Moonshot Kimi K2.6",
-        provider: "MoonshotAI",
-        huggingface_id: null,
-        pricing: {
-          input_per_million: 0.69,
-          output_per_million: 3.22,
-          cached_input_per_million: null,
-          cached_output_per_million: null,
-          currency: "USD",
-          pricing_tbd: false,
-        },
-        capabilities: {
-          tools: true,
-          json_mode: true,
-          vision: true,
-          reasoning: true,
-          reasoning_effort: false,
-          streaming: true,
-          system_role: true,
-          developer_role: false,
-        },
-        limits: {
-          max_context_length: 262144,
-          max_output_tokens: null,
-          max_images: 20,
-        },
-        deprecated: false,
-        deprecated_message: null,
-      },
-    };
-
-    const result = mapApiModel(visionModel);
-    expect(result.input).toEqual(["text", "image"]);
-    expect(result.reasoning).toBe(true);
-    expect(result.thinkingLevelMap).toEqual({
-      minimal: null,
-      low: null,
-      medium: "medium",
-      high: null,
-      xhigh: null,
-    });
-  });
-
-  it("should use defaults when metadata is missing", () => {
-    const bareModel: FullApiModel = {
-      id: "test/model",
-      object: "model",
-      created: 0,
-      owned_by: "vllm",
-      max_model_len: 8192,
-    };
-
-    const result = mapApiModel(bareModel);
-    expect(result.id).toBe("test/model");
-    expect(result.name).toBe("test/model");
-    expect(result.reasoning).toBe(false);
-    expect(result.contextWindow).toBe(8192);
-    expect(result.maxTokens).toBe(65536);
-    expect(result.input).toEqual(["text"]);
-    expect(result.cost.input).toBe(0);
-    expect(result.cost.output).toBe(0);
-    expect(result.fast).toBeUndefined();
-    expect(
-      (result.compat as Record<string, unknown>)?.supportsReasoningEffort,
-    ).toBeUndefined();
-  });
-});

package/src/extensions/provider/provider-payload.ts

@@ -1,12 +0,0 @@
-import type { NeuralwattModelConfig } from "./models";
-
-export function buildModelsPayload(models: NeuralwattModelConfig[]) {
-  return models.map(({ fast: _fast, ...model }) => ({
-    ...model,
-    compat: {
-      supportsDeveloperRole: false,
-      maxTokensField: "max_tokens" as const,
-      ...model.compat,
-    },
-  }));
-}

package/src/lib/fetch-models.ts

@@ -1,187 +0,0 @@
-import type { NeuralwattModelConfig } from "../extensions/provider/models";
-
-const FETCH_TIMEOUT_MS = 15_000;
-
-const NEURALWATT_BINARY_THINKING_LEVEL_MAP = {
-  minimal: null,
-  low: null,
-  medium: "medium",
-  high: null,
-  xhigh: null,
-} as const;
-
-const GPT_OSS_THINKING_LEVEL_MAP = {
-  minimal: "low",
-  low: "low",
-  medium: "medium",
-  high: "high",
-  xhigh: null,
-} as const;
-
-export interface ApiModelMetadata {
-  display_name: string;
-  description: string | null;
-  provider: string;
-  huggingface_id: string | null;
-  pricing: {
-    input_per_million: number;
-    output_per_million: number;
-    cached_input_per_million: number | null;
-    cached_output_per_million: number | null;
-    currency: string;
-    pricing_tbd: boolean;
-  };
-  capabilities: {
-    tools: boolean;
-    json_mode: boolean;
-    vision: boolean;
-    reasoning: boolean;
-    reasoning_effort: boolean;
-    streaming: boolean;
-    system_role: boolean;
-    developer_role: boolean;
-  };
-  limits: {
-    max_context_length: number;
-    max_output_tokens: number | null;
-    max_images: number | null;
-  };
-  deprecated: boolean;
-  deprecated_message: string | null;
-}
-
-export interface ApiModel {
-  id: string;
-  object: string;
-  created: number;
-  owned_by: string;
-  root?: string;
-  parent?: string | null;
-  max_model_len: number;
-  metadata?: ApiModelMetadata;
-}
-
-export interface ApiResponse {
-  object: "list";
-  data: ApiModel[];
-}
-
-/** Identify fast variants by their owned_by field or naming convention. */
-function isFastModel(model: ApiModel): boolean {
-  if (model.owned_by === "neuralwatt") return true;
-  return model.id.endsWith("-fast");
-}
-
-/** Map API model data to NeuralwattModelConfig. */
-export function mapApiModel(model: ApiModel): NeuralwattModelConfig {
-  const meta = model.metadata;
-  const fast = isFastModel(model);
-
-  // Base fields from top-level API data
-  const result: NeuralwattModelConfig = {
-    id: model.id,
-    name: meta?.display_name ?? model.id,
-    reasoning: meta?.capabilities.reasoning ?? false,
-    contextWindow: model.max_model_len,
-    maxTokens: 65536, // sensible default
-    cost: {
-      input: meta?.pricing.input_per_million ?? 0,
-      output: meta?.pricing.output_per_million ?? 0,
-      cacheRead: meta?.pricing.cached_input_per_million ?? 0,
-      cacheWrite: meta?.pricing.cached_output_per_million ?? 0,
-    },
-    input: meta?.capabilities.vision ? ["text", "image"] : ["text"],
-    compat: {
-      supportsDeveloperRole: false,
-      maxTokensField: "max_tokens",
-    },
-  };
-
-  if (fast) {
-    result.fast = true;
-  }
-
-  // Override maxTokens from limits if available
-  if (meta?.limits.max_output_tokens) {
-    result.maxTokens = meta.limits.max_output_tokens;
-  }
-
-  if (result.reasoning) {
-    result.thinkingLevelMap =
-      model.id === "openai/gpt-oss-20b"
-        ? GPT_OSS_THINKING_LEVEL_MAP
-        : NEURALWATT_BINARY_THINKING_LEVEL_MAP;
-  }
-
-  return result;
-}
-
-export type FetchModelsResult =
-  | { success: true; models: NeuralwattModelConfig[] }
-  | {
-      success: false;
-      error: { message: string; kind: "timeout" | "network" | "cancelled" };
-    };
-
-/**
- * Fetch live model definitions from the Neuralwatt /v1/models endpoint.
- *
- * When the API returns metadata (pricing, capabilities, limits), those values
- * are used directly. Fields not exposed by the API fall back to sensible
- * defaults.
- */
-export async function fetchModels(
-  signal?: AbortSignal,
-): Promise<FetchModelsResult> {
-  const signals: AbortSignal[] = [AbortSignal.timeout(FETCH_TIMEOUT_MS)];
-  if (signal) signals.push(signal);
-  const combined = AbortSignal.any(signals);
-
-  try {
-    const response = await fetch("https://api.neuralwatt.com/v1/models", {
-      headers: {
-        Referer: "https://pi.dev",
-        "X-Title": "npm:@aliou/pi-neuralwatt",
-      },
-      signal: combined,
-    });
-
-    if (!response.ok) {
-      return {
-        success: false,
-        error: {
-          message: `Failed to fetch models: ${response.status} ${response.statusText}`,
-          kind: "network",
-        },
-      };
-    }
-
-    const data: ApiResponse = await response.json();
-
-    // Filter out deprecated models
-    const active = data.data.filter(
-      (m) => !m.metadata?.deprecated && !m.metadata?.pricing.pricing_tbd,
-    );
-
-    const models = active.map(mapApiModel);
-    return { success: true, models };
-  } catch (err: unknown) {
-    if (err instanceof DOMException && err.name === "AbortError") {
-      if (
-        combined.reason instanceof DOMException &&
-        combined.reason.name === "TimeoutError"
-      ) {
-        return {
-          success: false,
-          error: { message: "Fetch models timed out", kind: "timeout" },
-        };
-      }
-      return {
-        success: false,
-        error: { message: "Fetch models cancelled", kind: "cancelled" },
-      };
-    }
-    const message = err instanceof Error ? err.message : "Unknown error";
-    return { success: false, error: { message, kind: "network" } };
-  }
-}

package/src/utils/is-offline.test.ts

@@ -1,60 +0,0 @@
-import { describe, expect, it } from "vitest";
-import { isOffline } from "./is-offline";
-
-describe("isOffline", () => {
-  it("returns true when PI_OFFLINE is 1", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "1";
-    expect(isOffline()).toBe(true);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns true when PI_OFFLINE is true", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "true";
-    expect(isOffline()).toBe(true);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns true when PI_OFFLINE is yes", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "yes";
-    expect(isOffline()).toBe(true);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns false when PI_OFFLINE is unset", () => {
-    const original = process.env.PI_OFFLINE;
-    delete process.env.PI_OFFLINE;
-    expect(isOffline()).toBe(false);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns false when PI_OFFLINE is 0", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "0";
-    expect(isOffline()).toBe(false);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns false when PI_OFFLINE is false", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "false";
-    expect(isOffline()).toBe(false);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns false when PI_OFFLINE is no", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "no";
-    expect(isOffline()).toBe(false);
-    process.env.PI_OFFLINE = original;
-  });
-
-  it("returns false for other values", () => {
-    const original = process.env.PI_OFFLINE;
-    process.env.PI_OFFLINE = "maybe";
-    expect(isOffline()).toBe(false);
-    process.env.PI_OFFLINE = original;
-  });
-});

package/src/utils/is-offline.ts

@@ -1,4 +0,0 @@
-export function isOffline(): boolean {
-  const value = process.env.PI_OFFLINE;
-  return value === "1" || value === "true" || value === "yes";
-}