pi-free 2.0.9 → 2.0.11

package/README.md

@@ -97,6 +97,7 @@ Want to see paid models too? Run the toggle command for your provider:
 /toggle-crofai    # Toggle CrofAI (💳 paid - needs API key with credits)
 /toggle-codestral # Toggle Codestral (💳 paid - free Experiment plan)
 /toggle-deepinfra # Toggle DeepInfra (💳 trial credit provider)
+/toggle-together  # Toggle Together AI (💳 trial credit provider)
 /toggle-sambanova # Toggle SambaNova (🔄 freemium)
 /toggle-llm7      # Toggle LLM7 (✅ free gateway)
 ```
@@ -412,6 +413,18 @@ AI inference cloud with 100+ open-source models:
 export DEEPINFRA_TOKEN="..."
 ```
 
+### Together AI ($1 trial credit)
+
+Fast inference on 200+ open-source models:
+
+- $1 one-time credit on signup (no credit card)
+- 138 chat models (Llama, DeepSeek, Qwen, Mixtral, etc.)
+- 60 RPM, 600 RPD (varies by model)
+
+```bash
+export TOGETHER_AI_API_KEY="..."
+```
+
 ### SambaNova Cloud (free tier)
 
 Fast inference on custom RDU hardware:
@@ -444,10 +457,13 @@ Each provider has toggle commands to switch between free and all models:
 | `/toggle-huggingface` | Toggle between free/all Hugging Face models (🔧 dynamic) |
 | `/toggle-codestral`   | Toggle Codestral (💳 paid)                               |
 | `/toggle-deepinfra`   | Toggle DeepInfra (💳 trial credit)                       |
+| `/toggle-together`    | Toggle Together AI (💳 trial credit)                     |
 | `/toggle-sambanova`   | Toggle SambaNova (🔄 freemium)                           |
 | `/toggle-llm7`        | Toggle LLM7 (✅ free gateway)                            |
 | `/toggle-zenmux`      | Toggle ZenMux (💳 paid)                                  |
 | `/toggle-crofai`      | Toggle CrofAI (💳 paid)                                  |
+| `/ollama-cloud-refresh` | Re-fetch Ollama Cloud models live (no restart needed)  |
+| `/probe-ollama`       | Test Ollama Cloud models for 403 errors (auto-hide)      |
 
 **The toggle command:**
 

package/banner.svg

@@ -106,10 +106,11 @@
     <rect x="0" y="0" width="250" height="170" rx="12" fill="url(#card1)" stroke="#7c3aed" stroke-width="0.5" stroke-opacity="0.2" filter="url(#shadow)"/>
     <rect x="0" y="0" width="250" height="170" rx="12" fill="none" stroke="url(#accent)" stroke-width="0.5" opacity="0.1"/>
     <text x="20" y="28" font-family="system-ui, sans-serif" font-size="13" font-weight="700" fill="#c4b5fd">Custom Providers</text>
-    <text x="20" y="55" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">Kilo · Cline</text>
-    <text x="20" y="78" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">NVIDIA</text>
-    <text x="20" y="101" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">Ollama Cloud · ZenMux</text>
-    <text x="20" y="124" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">CrofAI</text>
+    <text x="20" y="50" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">Kilo · Cline · NVIDIA</text>
+    <text x="20" y="70" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">Ollama Cloud · ZenMux</text>
+    <text x="20" y="90" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">CrofAI · Codestral · LLM7</text>
+    <text x="20" y="110" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">DeepInfra · SambaNova</text>
+    <text x="20" y="130" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">Together</text>
   </g>
 
   <g transform="translate(970, 55)">
@@ -117,12 +118,13 @@
     <rect x="0" y="0" width="250" height="170" rx="12" fill="url(#card2)" stroke="#06b6d4" stroke-width="0.5" stroke-opacity="0.2" filter="url(#shadow)"/>
     <rect x="0" y="0" width="250" height="170" rx="12" fill="none" stroke="url(#accent2)" stroke-width="0.5" opacity="0.1"/>
     <text x="20" y="28" font-family="system-ui, sans-serif" font-size="13" font-weight="700" fill="#67e8f9">Features</text>
-    <text x="20" y="52" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">✦ Free model auto-detection</text>
-    <text x="20" y="72" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">✦ Per-provider toggles</text>
-    <text x="20" y="92" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">✦ OAuth flows (Kilo, Cline)</text>
-    <text x="20" y="112" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">✦ Coding Index (CI) scores</text>
-    <text x="20" y="132" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">✦ Model health probes</text>
-    <text x="20" y="152" font-family="system-ui, sans-serif" font-size="12" fill="#8b9aaa">✦ 404/403 auto-hide on probes</text>
+    <text x="20" y="50" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ Free model auto-detection</text>
+    <text x="20" y="68" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ Per-provider toggles</text>
+    <text x="20" y="86" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ OAuth flows (Kilo, Cline)</text>
+    <text x="20" y="104" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ Coding Index (CI) scores</text>
+    <text x="20" y="122" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ Model health probes</text>
+    <text x="20" y="140" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ Thinking level maps</text>
+    <text x="20" y="158" font-family="system-ui, sans-serif" font-size="11" fill="#8b9aaa">✦ 404/403 auto-hide on probe</text>
   </g>
 
   <!-- Bottom tagline -->

package/config.ts

@@ -28,14 +28,10 @@ interface PiFreeConfig {
 	zenmux_api_key?: string;
 	crofai_api_key?: string;
 	codestral_api_key?: string;
-	mistral_api_key?: string;
 	llm7_api_key?: string;
 	deepinfra_api_key?: string;
 	sambanova_api_key?: string;
-	groq_api_key?: string;
-	cerebras_api_key?: string;
-	xai_api_key?: string;
-	hf_token?: string;
+	together_api_key?: string;
 	kilo_free_only?: boolean;
 	hidden_models?: string[];
 	free_only?: boolean;
@@ -48,6 +44,7 @@ interface PiFreeConfig {
 	llm7_show_paid?: boolean;
 	deepinfra_show_paid?: boolean;
 	sambanova_show_paid?: boolean;
+	together_show_paid?: boolean;
 	openrouter_show_paid?: boolean;
 	opencode_show_paid?: boolean;
 }
@@ -58,14 +55,10 @@ const CONFIG_TEMPLATE: PiFreeConfig = {
 	zenmux_api_key: "",
 	crofai_api_key: "",
 	codestral_api_key: "",
-	mistral_api_key: "",
 	llm7_api_key: "",
 	deepinfra_api_key: "",
 	sambanova_api_key: "",
-	groq_api_key: "",
-	cerebras_api_key: "",
-	xai_api_key: "",
-	hf_token: "",
+	together_api_key: "",
 
 	kilo_free_only: false,
 	hidden_models: [],
@@ -79,6 +72,7 @@ const CONFIG_TEMPLATE: PiFreeConfig = {
 	llm7_show_paid: false,
 	deepinfra_show_paid: false,
 	sambanova_show_paid: false,
+	together_show_paid: false,
 	openrouter_show_paid: false,
 	opencode_show_paid: false,
 };
@@ -90,9 +84,21 @@ function ensureConfigFile(): void {
 	try {
 		mkdirSync(PI_DIR, { recursive: true });
 		if (existsSync(CONFIG_PATH)) {
-			const existing = JSON.parse(
-				readFileSync(CONFIG_PATH, "utf8"),
-			) as PiFreeConfig;
+			let existing: PiFreeConfig;
+			try {
+				existing = JSON.parse(
+					readFileSync(CONFIG_PATH, "utf8"),
+				) as PiFreeConfig;
+			} catch (_parseErr) {
+				// File exists but is corrupt — DO NOT overwrite it.
+				// The user needs to fix or delete it manually.
+				_logger.error(
+					"Config file exists but is corrupt — refusing to overwrite. Fix or delete ~/.pi/free.json.",
+					{ path: CONFIG_PATH },
+				);
+				return;
+			}
+			// Merge with template to add any missing keys, preserving existing values
 			const merged = { ...CONFIG_TEMPLATE, ...existing };
 			if (JSON.stringify(merged) !== JSON.stringify(existing)) {
 				writeFileSync(
@@ -120,7 +126,7 @@ export function loadConfigFile(): PiFreeConfig {
 	try {
 		return JSON.parse(readFileSync(CONFIG_PATH, "utf8")) as PiFreeConfig;
 	} catch (err) {
-		_logger.warn("Could not parse config file — returning empty config", {
+		_logger.error("Could not parse config file — returning empty config", {
 			path: CONFIG_PATH,
 			error: err instanceof Error ? err.message : String(err),
 		});
@@ -128,6 +134,18 @@ export function loadConfigFile(): PiFreeConfig {
 	}
 }
 
+/**
+ * Read the raw config file content without merging with template.
+ * Returns the file content as string, or undefined if unreadable.
+ */
+function readRawConfigFile(): string | undefined {
+	try {
+		return readFileSync(CONFIG_PATH, "utf8");
+	} catch {
+		return undefined;
+	}
+}
+
 ensureConfigFile();
 
 // Resolve each value: env var takes priority over config file.
@@ -188,6 +206,10 @@ export function getSambanovaShowPaid(): boolean {
 	);
 }
 
+export function getTogetherShowPaid(): boolean {
+	return resolveBool("TOGETHER_SHOW_PAID", loadConfigFile().together_show_paid);
+}
+
 export function getOllamaShowPaid(): boolean {
 	return resolveBool("OLLAMA_SHOW_PAID", loadConfigFile().ollama_show_paid);
 }
@@ -247,28 +269,37 @@ export function getSambanovaApiKey(): string | undefined {
 	return resolve("SAMBANOVA_API_KEY", loadConfigFile().sambanova_api_key);
 }
 
+export function getTogetherApiKey(): string | undefined {
+	return resolve("TOGETHER_AI_API_KEY", loadConfigFile().together_api_key);
+}
+
 export function getOllamaApiKey(): string | undefined {
 	return resolve("OLLAMA_API_KEY", loadConfigFile().ollama_api_key);
 }
 
+/** Mistral is pi's built-in provider — key comes from env var only. */
 export function getMistralApiKey(): string | undefined {
-	return resolve("MISTRAL_API_KEY", loadConfigFile().mistral_api_key);
+	return process.env.MISTRAL_API_KEY;
 }
 
+/** Groq is pi's built-in provider — key comes from env var only. */
 export function getGroqApiKey(): string | undefined {
-	return resolve("GROQ_API_KEY", loadConfigFile().groq_api_key);
+	return process.env.GROQ_API_KEY;
 }
 
+/** Cerebras is pi's built-in provider — key comes from env var only. */
 export function getCerebrasApiKey(): string | undefined {
-	return resolve("CEREBRAS_API_KEY", loadConfigFile().cerebras_api_key);
+	return process.env.CEREBRAS_API_KEY;
 }
 
+/** xAI is pi's built-in provider — key comes from env var only. */
 export function getXaiApiKey(): string | undefined {
-	return resolve("XAI_API_KEY", loadConfigFile().xai_api_key);
+	return process.env.XAI_API_KEY;
 }
 
+/** HuggingFace is pi's built-in provider — token comes from env var only. */
 export function getHfToken(): string | undefined {
-	return resolve("HF_TOKEN", loadConfigFile().hf_token);
+	return process.env.HF_TOKEN;
 }
 
 /**
@@ -315,7 +346,42 @@ export function applyHidden<T extends { id: string }>(
 
 export function saveConfig(updates: Partial<PiFreeConfig>): void {
 	try {
-		const existing = loadConfigFile();
+		// Read the raw file content — never use loadConfigFile() here because
+		// if the file is unparseable, loadConfigFile() returns {} which would
+		// cause us to write a partial config and WIPE all existing keys.
+		const raw = readRawConfigFile();
+		if (raw === undefined) {
+			// File doesn't exist or can't be read — start from template
+			const merged = { ...CONFIG_TEMPLATE, ...updates };
+			writeFileSync(
+				CONFIG_PATH,
+				`${JSON.stringify(merged, null, 2)}\n`,
+				"utf8",
+			);
+			_logger.info("Config saved (new file)", {
+				path: CONFIG_PATH,
+				keys: Object.keys(updates),
+			});
+			return;
+		}
+
+		let existing: PiFreeConfig;
+		try {
+			existing = JSON.parse(raw) as PiFreeConfig;
+		} catch (parseErr) {
+			// File exists but is corrupt. REFUSE to overwrite it with a partial
+			// config — that would permanently destroy the user's keys.
+			_logger.error(
+				"REFUSING to save config — existing file is corrupt. Fix or delete ~/.pi/free.json manually.",
+				{
+					path: CONFIG_PATH,
+					error:
+						parseErr instanceof Error ? parseErr.message : String(parseErr),
+				},
+			);
+			return;
+		}
+
 		const merged = { ...existing, ...updates };
 		writeFileSync(CONFIG_PATH, `${JSON.stringify(merged, null, 2)}\n`, "utf8");
 		_logger.info("Config saved", {

package/constants.ts

@@ -21,6 +21,7 @@ export const PROVIDER_CODESTRAL = "codestral";
 export const PROVIDER_LLM7 = "llm7";
 export const PROVIDER_DEEPINFRA = "deepinfra";
 export const PROVIDER_SAMBANOVA = "sambanova";
+export const PROVIDER_TOGETHER = "together";
 
 export const ALL_UNIQUE_PROVIDERS = [
 	PROVIDER_KILO,
@@ -36,6 +37,7 @@ export const ALL_UNIQUE_PROVIDERS = [
 	PROVIDER_LLM7,
 	PROVIDER_DEEPINFRA,
 	PROVIDER_SAMBANOVA,
+	PROVIDER_TOGETHER,
 ] as const;
 
 // =============================================================================
@@ -56,6 +58,7 @@ export const BASE_URL_CODESTRAL = "https://codestral.mistral.ai/v1";
 export const BASE_URL_LLM7 = "https://api.llm7.io/v1";
 export const BASE_URL_DEEPINFRA = "https://api.deepinfra.com/v1/openai";
 export const BASE_URL_SAMBANOVA = "https://api.sambanova.ai/v1";
+export const BASE_URL_TOGETHER = "https://api.together.xyz/v1";
 
 /** Cline fetches free models from OpenRouter */
 export const BASE_URL_OPENROUTER = "https://openrouter.ai/api/v1";

package/index.ts

@@ -13,6 +13,7 @@
  * - Codestral: Mistral's code-focused model via codestral.mistral.ai (free tier)
  * - DeepInfra: AI inference cloud ($5 trial credit)
  * - SambaNova: Fast inference on RDU hardware (free tier, no credit card)
+ * - Together: Fast inference on 200+ open-source models ($1 trial credit)
  * - LLM7: AI gateway (free default/fast selectors)
  */
 
@@ -38,6 +39,7 @@ import kilo from "./providers/kilo/kilo.ts";
 import llm7 from "./providers/llm7/llm7.ts";
 import deepinfra from "./providers/deepinfra/deepinfra.ts";
 import sambanova from "./providers/sambanova/sambanova.ts";
+import together from "./providers/together/together.ts";
 import nvidia from "./providers/nvidia/nvidia.ts";
 import ollama from "./providers/ollama/ollama.ts";
 import zenmux from "./providers/zenmux/zenmux.ts";
@@ -207,6 +209,7 @@ export default async function piFreeEntry(pi: ExtensionAPI) {
 		llm7(pi),
 		deepinfra(pi),
 		sambanova(pi),
+		together(pi),
 	]);
 
 	// Setup dynamic built-in providers (Mistral, Groq, Cerebras, xAI, Hugging Face)

package/lib/util.ts

@@ -384,12 +384,32 @@ export interface OpenAIModelDefaults {
 
 /**
  * Generic model shape returned by OpenAI-compatible /v1/models endpoints.
+ *
+ * Some providers (SambaNova, DeepInfra) return extended fields beyond
+ * the standard OpenAI format. We accept them loosely and use what's
+ * available, falling back to defaults otherwise.
  */
 export interface OpenAIModelEntry {
 	id: string;
 	object?: string;
 	created?: number;
 	owned_by?: string;
+	/** Extended: per-model reasoning capability (some providers expose this) */
+	reasoning?: boolean;
+	/** Extended: input modalities (some providers expose this) */
+	input_modalities?: string[];
+	/** Extended: per-model context length (SambaNova, etc.) */
+	context_length?: number;
+	/** Extended: alternate field name for context length */
+	max_context_length?: number;
+	/** Extended: alternate field name for context length (snake_case) */
+	context_window?: number;
+	/** Extended: per-model max completion tokens (SambaNova, etc.) */
+	max_completion_tokens?: number;
+	/** Extended: alternate field name for max tokens */
+	max_tokens?: number;
+	/** Extended: per-model pricing (SambaNova, etc.) */
+	pricing?: { prompt?: string | number; completion?: string | number };
 }
 
 /**
@@ -426,8 +446,10 @@ export async function fetchOpenAICompatibleModels(
 			throw new Error(`${providerId} API error: ${response.status}`);
 		}
 
-		const data = (await response.json()) as { data?: OpenAIModelEntry[] };
-		const models = data.data ?? [];
+		const body = (await response.json()) as
+			| OpenAIModelEntry[]
+			| { data?: OpenAIModelEntry[] };
+		const models = Array.isArray(body) ? body : (body.data ?? []);
 
 		logger.info(`[${providerId}] Fetched ${models.length} models`);
 
@@ -435,19 +457,61 @@ export async function fetchOpenAICompatibleModels(
 			.filter((m) => m.id)
 			.map((m): PiProviderModelConfig => {
 				const name = m.id.split("/").pop() || m.id;
+
+				// Use per-model context length if the API provides it (try multiple field names)
+				const contextWindow =
+					m.context_length ??
+					m.max_context_length ??
+					m.context_window ??
+					defaults.contextWindow ??
+					128_000;
+
+				// Use per-model max tokens if the API provides it (try multiple field names)
+				const maxTokens =
+					m.max_completion_tokens ??
+					m.max_tokens ??
+					defaults.maxTokens ??
+					4_096;
+
+				// Use per-model reasoning flag if the API provides it
+				const reasoning =
+					m.reasoning ?? isLikelyReasoningModel({ id: m.id, name });
+
+				// Use per-model input_modalities if the API provides it
+				const hasVision = m.input_modalities?.includes("image") ?? false;
+				const input =
+					(defaults.input as PiProviderModelConfig["input"]) ??
+					(hasVision ? ["text", "image"] : ["text"]);
+
+				// Use per-model pricing if the API provides it, otherwise use defaults
+				const inputCost =
+					(typeof m.pricing?.prompt === "number" ||
+					typeof m.pricing?.prompt === "string"
+						? Number(m.pricing.prompt)
+						: undefined) ??
+					defaults.cost?.input ??
+					0;
+				const outputCost =
+					(typeof m.pricing?.completion === "number" ||
+					typeof m.pricing?.completion === "string"
+						? Number(m.pricing.completion)
+						: undefined) ??
+					defaults.cost?.output ??
+					0;
+
 				return {
 					id: m.id,
 					name,
-					reasoning: isLikelyReasoningModel({ id: m.id, name }),
-					input: (defaults.input as PiProviderModelConfig["input"]) ?? ["text"],
+					reasoning,
+					input,
 					cost: {
-						input: defaults.cost?.input ?? 0,
-						output: defaults.cost?.output ?? 0,
+						input: inputCost,
+						output: outputCost,
 						cacheRead: 0,
 						cacheWrite: 0,
 					},
-					contextWindow: defaults.contextWindow ?? 128_000,
-					maxTokens: defaults.maxTokens ?? 4_096,
+					contextWindow,
+					maxTokens,
 					compat: getProxyModelCompat({ id: m.id, name }),
 				};
 			});

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "pi-free",
-	"version": "2.0.9",
+	"version": "2.0.11",
 	"type": "module",
 	"description": "AI model providers for Pi with free model filtering. Shows only $0 cost models by default. Supports Kilo (free OAuth), Cline (free), NVIDIA (freemium), ZenMux, CrofAI, Ollama Cloud, and more.",
 	"keywords": [

package/providers/crofai/crofai.ts

@@ -1,30 +1,128 @@
 /**
  * CrofAI Provider Extension
  *
- * Provides access to CrofAI API - OpenAI-compatible LLM inference service.
+ * Provides access to CrofAI API - OpenAI-compatible LLM inference service
+ * hosting DeepSeek, Qwen, and other open-source models.
+ *
+ * NOTE: CrofAI's /v1/models returns per-model context_length, max_completion_tokens,
+ * name, custom_reasoning, and reasoning_effort. Pricing is per-MILLION tokens.
  *
  * Setup:
  *   1. Get API key from https://ai.nahcrof.com
  *   2. Set CROFAI_API_KEY env var or add to ~/.pi/free.json
  *
- * Responds to global free-only filter.
- *
  * Usage:
  *   pi install git:github.com/apmantza/pi-free
  *   # Set CROFAI_API_KEY env var
  *   # Models appear in /model selector
  */
 
-import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
+import type {
+	ExtensionAPI,
+	ProviderModelConfig,
+} from "@earendil-works/pi-coding-agent";
 import { getCrofaiApiKey, getCrofaiShowPaid } from "../../config.ts";
-import { BASE_URL_CROFAI, PROVIDER_CROFAI } from "../../constants.ts";
+import {
+	BASE_URL_CROFAI,
+	DEFAULT_FETCH_TIMEOUT_MS,
+	PROVIDER_CROFAI,
+} from "../../constants.ts";
 import { createLogger } from "../../lib/logger.ts";
+import {
+	getProxyModelCompat,
+	isLikelyReasoningModel,
+} from "../../lib/provider-compat.ts";
 import { isFreeModel, registerWithGlobalToggle } from "../../lib/registry.ts";
-import { fetchOpenAICompatibleModels } from "../../lib/util.ts";
+import { fetchWithRetry } from "../../lib/util.ts";
 import { createReRegister, setupProvider } from "../../provider-helper.ts";
 
 const _logger = createLogger("crofai");
 
+// =============================================================================
+// Types
+// =============================================================================
+
+interface CrofaiModel {
+	id: string;
+	name?: string;
+	context_length?: number;
+	max_completion_tokens?: number;
+	custom_reasoning?: boolean;
+	reasoning_effort?: boolean;
+	pricing?: {
+		prompt?: string;
+		completion?: string;
+		cache_prompt?: string;
+	};
+}
+
+// =============================================================================
+// Fetch
+// =============================================================================
+
+function parseCrofaiPrice(priceStr: string | undefined): number {
+	if (priceStr === undefined) return 0;
+	const num = Number.parseFloat(priceStr);
+	if (Number.isNaN(num)) return 0;
+	// CrofAI pricing is per-MILLION tokens. Divide to get per-token (Pi convention).
+	return num / 1_000_000;
+}
+
+async function fetchCrofaiModels(
+	apiKey: string,
+): Promise<ProviderModelConfig[]> {
+	const response = await fetchWithRetry(
+		`${BASE_URL_CROFAI}/models`,
+		{
+			headers: {
+				Authorization: `Bearer ${apiKey}`,
+				"Content-Type": "application/json",
+			},
+		},
+		3,
+		1000,
+		DEFAULT_FETCH_TIMEOUT_MS,
+	);
+
+	if (!response.ok) {
+		throw new Error(
+			`CrofAI API error: ${response.status} ${response.statusText}`,
+		);
+	}
+
+	// CrofAI returns { data: [...] }
+	const json = (await response.json()) as {
+		data?: CrofaiModel[];
+	};
+	const models = json.data ?? [];
+
+	_logger.info(`[crofai] Fetched ${models.length} models`);
+
+	return models
+		.filter((m) => m.id)
+		.map((m): ProviderModelConfig => {
+			const name = m.name || m.id;
+			const reasoning =
+				m.custom_reasoning ?? isLikelyReasoningModel({ id: m.id, name });
+
+			return {
+				id: m.id,
+				name,
+				reasoning,
+				input: ["text"],
+				cost: {
+					input: parseCrofaiPrice(m.pricing?.prompt),
+					output: parseCrofaiPrice(m.pricing?.completion),
+					cacheRead: parseCrofaiPrice(m.pricing?.cache_prompt),
+					cacheWrite: 0,
+				},
+				contextWindow: m.context_length ?? 128_000,
+				maxTokens: m.max_completion_tokens ?? 16_384,
+				compat: getProxyModelCompat({ id: m.id, name }),
+			};
+		});
+}
+
 // =============================================================================
 // Extension Entry Point
 // =============================================================================
@@ -39,21 +137,14 @@ export default async function crofaiProvider(pi: ExtensionAPI) {
 		return;
 	}
 
-	// Fetch models via shared OpenAI-compatible helper
-	const allModels = await fetchOpenAICompatibleModels(
-		"crofai",
-		BASE_URL_CROFAI,
-		apiKey,
-	);
+	// Fetch models
+	const allModels = await fetchCrofaiModels(apiKey);
 
 	if (allModels.length === 0) {
 		_logger.warn("[crofai] No models available");
 		return;
 	}
 
-	// Use isFreeModel with allModels for proper detection
-	// CrofAI doesn't expose pricing (all costs are $0), so Route B will be used:
-	// FREE only if "free" in name
 	const freeModels = allModels.filter((m) =>
 		isFreeModel({ ...m, provider: PROVIDER_CROFAI }, allModels),
 	);