pi-free 1.0.8 → 2.0.0

package/providers/cloudflare/cloudflare.ts

@@ -0,0 +1,368 @@
+/**
+ * Cloudflare Workers AI Provider Extension
+ *
+ * Provides access to Cloudflare's serverless GPU network with 18+ models.
+ * All models use Cloudflare's "Neurons" pricing system:
+ *   - 10,000 Neurons per day FREE (resets daily at 00:00 UTC)
+ *   - $0.011 per 1,000 Neurons beyond free allocation
+ *
+ * Setup:
+ *   1. Create API token at https://dash.cloudflare.com/profile/api-tokens
+ *      with "Cloudflare AI" > "Read" permission
+ *   2. Get Account ID from https://dash.cloudflare.com (right sidebar)
+ *   3. Add credentials to ~/.pi/agent/auth.json or set env vars
+ *
+ * Auth (in order of priority):
+ *   - Environment: CF_API_TOKEN and CF_ACCOUNT_ID
+ *   - Config file: ~/.pi/agent/auth.json
+ *     { "cloudflare-ai": { "access": "token", "account_id": "id" } }
+ *   - Legacy: CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID env vars
+ *
+ * Models can be customized via ~/.pi/cloudflare-models.json
+ */
+
+import { existsSync, readFileSync } from "node:fs";
+import { homedir } from "node:os";
+import { join } from "node:path";
+import type {
+	ExtensionAPI,
+	ProviderModelConfig,
+} from "@mariozechner/pi-coding-agent";
+import { createLogger } from "../../lib/logger.ts";
+
+const _logger = createLogger("cloudflare");
+
+// =============================================================================
+// Auth Resolution
+// =============================================================================
+
+interface CloudflareAuth {
+	token?: string;
+	accountId?: string;
+}
+
+function getCloudflareAuth(): CloudflareAuth {
+	const result: CloudflareAuth = {};
+
+	// Check new env var names first
+	if (process.env.CF_API_TOKEN) result.token = process.env.CF_API_TOKEN;
+	if (process.env.CF_ACCOUNT_ID) result.accountId = process.env.CF_ACCOUNT_ID;
+
+	// Check legacy env var names
+	if (!result.token && process.env.CLOUDFLARE_API_TOKEN) {
+		result.token = process.env.CLOUDFLARE_API_TOKEN;
+	}
+	if (!result.accountId && process.env.CLOUDFLARE_ACCOUNT_ID) {
+		result.accountId = process.env.CLOUDFLARE_ACCOUNT_ID;
+	}
+
+	if (result.token && result.accountId) return result;
+
+	// Check ~/.pi/free.json (pi-free config)
+	try {
+		const configPath = join(homedir(), ".pi", "free.json");
+		if (existsSync(configPath)) {
+			const config = JSON.parse(readFileSync(configPath, "utf-8"));
+			if (!result.token && config.cloudflare_api_token) {
+				result.token = config.cloudflare_api_token;
+			}
+			if (!result.accountId && config.cloudflare_account_id) {
+				result.accountId = config.cloudflare_account_id;
+			}
+		}
+	} catch {
+		// Ignore config file errors
+	}
+
+	return result;
+}
+
+// =============================================================================
+// Compatibility Settings
+// =============================================================================
+
+/**
+ * Cloudflare Workers AI compatibility settings.
+ * Prevents 413 Payload Too Large errors by disabling unsupported parameters.
+ */
+const CLOUDFLARE_COMPAT: {
+	supportsStore?: boolean;
+	supportsDeveloperRole?: boolean;
+	supportsReasoningEffort?: boolean;
+	supportsStrictMode?: boolean;
+	maxTokensField?: "max_tokens" | "max_completion_tokens";
+	requiresThinkingAsText?: boolean;
+} = {
+	supportsStore: false,
+	supportsDeveloperRole: false,
+	supportsReasoningEffort: false,
+	supportsStrictMode: false,
+	maxTokensField: "max_tokens",
+};
+
+// =============================================================================
+// Default Models (18 models from Cloudflare Workers AI)
+// =============================================================================
+
+interface ModelConfig extends ProviderModelConfig {
+	compat?: { requiresThinkingAsText?: boolean };
+	_remove?: boolean;
+}
+
+const DEFAULT_MODELS: ModelConfig[] = [
+	// Frontier models
+	{
+		id: "@cf/moonshotai/kimi-k2.5",
+		name: "Kimi K2.5",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: { input: 0.6, output: 3.0, cacheRead: 0.1, cacheWrite: 0 },
+		contextWindow: 256000,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/meta/llama-4-scout-17b-16e-instruct",
+		name: "Llama 4 Scout 17B",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: { input: 0.27, output: 0.85, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 131072,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/nvidia/nemotron-3-120b-a12b",
+		name: "Nemotron 3 Super 120B",
+		reasoning: true,
+		input: ["text"],
+		cost: { input: 0.5, output: 1.5, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 256000,
+		maxTokens: 8192,
+		compat: { requiresThinkingAsText: true },
+	},
+	{
+		id: "@cf/google/gemma-4-26b-a4b-it",
+		name: "Gemma 4 26B",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: { input: 0.1, output: 0.3, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 256000,
+		maxTokens: 8192,
+		compat: { requiresThinkingAsText: true },
+	},
+	{
+		id: "@cf/google/gemma-3-12b-it",
+		name: "Gemma 3 12B",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: { input: 0.345, output: 0.556, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 80000,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/qwen/qwen3-30b-a3b-fp8",
+		name: "Qwen3 30B A3B",
+		reasoning: true,
+		input: ["text"],
+		cost: { input: 0.051, output: 0.34, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 32768,
+		maxTokens: 8192,
+		compat: { requiresThinkingAsText: true },
+	},
+	{
+		id: "@cf/zai-org/glm-4.7-flash",
+		name: "GLM-4.7 Flash",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.06, output: 0.4, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 131072,
+		maxTokens: 8192,
+	},
+	// Popular models
+	{
+		id: "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
+		name: "Llama 3.3 70B (Fast)",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.5, output: 0.5, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 131072,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/meta/llama-3.1-8b-instruct",
+		name: "Llama 3.1 8B",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.1, output: 0.1, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 131072,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/meta/llama-3.1-70b-instruct",
+		name: "Llama 3.1 70B",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.5, output: 0.5, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 131072,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/meta/llama-3.1-405b-instruct",
+		name: "Llama 3.1 405B",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 2.0, output: 2.0, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 131072,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/deepseek-ai/deepseek-r1-distill-qwen-32b",
+		name: "DeepSeek R1 Distill Qwen 32B",
+		reasoning: true,
+		input: ["text"],
+		cost: { input: 0.3, output: 0.3, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 32768,
+		maxTokens: 8192,
+		compat: { requiresThinkingAsText: true },
+	},
+	{
+		id: "@cf/deepseek-ai/deepseek-math-7b-instruct",
+		name: "DeepSeek Math 7B",
+		reasoning: true,
+		input: ["text"],
+		cost: { input: 0.1, output: 0.1, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 16384,
+		maxTokens: 4096,
+	},
+	// Mistral models
+	{
+		id: "@cf/mistral/mistral-small-3.1-24b-instruct",
+		name: "Mistral Small 3.1 24B",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: { input: 0.3, output: 0.3, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 32768,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/mistral/mistral-7b-instruct-v0.2-lora",
+		name: "Mistral 7B Instruct",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.1, output: 0.1, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 32768,
+		maxTokens: 4096,
+	},
+	{
+		id: "@cf/mistral/mixtral-8x7b-instruct-v0.1-awq",
+		name: "Mixtral 8x7B Instruct",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.3, output: 0.3, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 32768,
+		maxTokens: 4096,
+	},
+	// Qwen and Gemma
+	{
+		id: "@cf/qwen/qwen1.5-14b-chat-awq",
+		name: "Qwen 1.5 14B Chat",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.2, output: 0.2, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 32768,
+		maxTokens: 8192,
+	},
+	{
+		id: "@cf/google/gemma-2b-it-lora",
+		name: "Gemma 2B",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.05, output: 0.05, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 8192,
+		maxTokens: 2048,
+	},
+	{
+		id: "@cf/google/gemma-7b-it-lora",
+		name: "Gemma 7B",
+		reasoning: false,
+		input: ["text"],
+		cost: { input: 0.1, output: 0.1, cacheRead: 0, cacheWrite: 0 },
+		contextWindow: 8192,
+		maxTokens: 2048,
+	},
+];
+
+// =============================================================================
+// Model Customization (user overrides)
+// =============================================================================
+
+function getModels(): ProviderModelConfig[] {
+	// Apply Cloudflare compat settings to all default models
+	const defaults = DEFAULT_MODELS.map((m) => ({
+		...m,
+		compat: { ...CLOUDFLARE_COMPAT, ...m.compat },
+	})) as ProviderModelConfig[];
+
+	// Check for user overrides
+	const overridePath = join(homedir(), ".pi", "cloudflare-models.json");
+	if (!existsSync(overridePath)) return defaults;
+
+	try {
+		const override = JSON.parse(
+			readFileSync(overridePath, "utf-8"),
+		) as ModelConfig[];
+		const modelMap = new Map<string, any>(defaults.map((m) => [m.id, m]));
+
+		for (const model of override) {
+			if (model._remove) {
+				modelMap.delete(model.id);
+			} else {
+				// Apply Cloudflare compat settings to user overrides
+				model.compat = { ...CLOUDFLARE_COMPAT, ...model.compat };
+				modelMap.set(model.id, model);
+			}
+		}
+
+		return Array.from(modelMap.values()) as ProviderModelConfig[];
+	} catch (e) {
+		_logger.warn(
+			`[cloudflare] Failed to load ~/.pi/cloudflare-models.json: ${e instanceof Error ? e.message : String(e)}`,
+		);
+		return defaults;
+	}
+}
+
+// =============================================================================
+// Extension Entry Point
+// =============================================================================
+
+export default async function cloudflareProvider(pi: ExtensionAPI) {
+	const { token: apiToken, accountId } = getCloudflareAuth();
+
+	if (!apiToken) {
+		_logger.info(
+			"[cloudflare] CF_API_TOKEN or CLOUDFLARE_API_TOKEN not set. Provider will not be available.",
+		);
+		return;
+	}
+
+	if (!accountId) {
+		_logger.info(
+			"[cloudflare] CF_ACCOUNT_ID or CLOUDFLARE_ACCOUNT_ID not set. Provider will not be available.",
+		);
+		return;
+	}
+
+	const models = getModels();
+
+	pi.registerProvider("cloudflare", {
+		baseUrl: `https://api.cloudflare.com/client/v4/accounts/${accountId}/ai/v1`,
+		apiKey: apiToken,
+		api: "openai-completions",
+		authHeader: true,
+		models,
+	});
+
+	_logger.info(
+		`[cloudflare] Provider registered with account: ${accountId.slice(0, 8)}... (${models.length} models)`,
+	);
+}