@oh-my-pi/pi-ai 8.0.16 → 8.1.0

package/package.json

@@ -1,17 +1,30 @@
 {
 	"name": "@oh-my-pi/pi-ai",
-	"version": "8.0.16",
+	"version": "8.1.0",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./src/index.ts",
 	"types": "./src/index.ts",
+	"exports": {
+		".": {
+			"types": "./src/index.ts",
+			"import": "./src/index.ts"
+		},
+		"./utils/*": {
+			"types": "./src/utils/*.ts",
+			"import": "./src/utils/*.ts"
+		},
+		"./*": {
+			"types": "./src/*",
+			"import": "./src/*"
+		}
+	},
 	"bin": {
 		"pi-ai": "./src/cli.ts"
 	},
 	"files": [
 		"src",
-		"README.md",
-		"tsconfig.json"
+		"README.md"
 	],
 	"scripts": {
 		"generate-models": "bun scripts/generate-models.ts",
@@ -19,7 +32,7 @@
 		"prepublishOnly": "cp tsconfig.publish.json tsconfig.json"
 	},
 	"dependencies": {
-		"@oh-my-pi/pi-utils": "8.0.16",
+		"@oh-my-pi/pi-utils": "workspace:*",
 		"@anthropic-ai/sdk": "0.71.2",
 		"@aws-sdk/client-bedrock-runtime": "^3.968.0",
 		"@bufbuild/protobuf": "^2.10.2",
@@ -57,12 +70,5 @@
 	},
 	"devDependencies": {
 		"@types/node": "^24.3.0"
-	},
-	"exports": {
-		".": {
-			"types": "./src/index.ts",
-			"import": "./src/index.ts"
-		},
-		"./*": "./src/*"
 	}
 }

package/src/cli.ts

@@ -2,12 +2,12 @@
 import { createInterface } from "readline";
 import { CliAuthStorage } from "./storage";
 import "./utils/migrate-env";
+import { getOAuthProviders } from "./utils/oauth";
 import { loginAnthropic } from "./utils/oauth/anthropic";
 import { loginCursor } from "./utils/oauth/cursor";
 import { loginGitHubCopilot } from "./utils/oauth/github-copilot";
 import { loginAntigravity } from "./utils/oauth/google-antigravity";
 import { loginGeminiCli } from "./utils/oauth/google-gemini-cli";
-import { getOAuthProviders } from "./utils/oauth/index";
 import { loginOpenAICodex } from "./utils/oauth/openai-codex";
 import type { OAuthCredentials, OAuthProvider } from "./utils/oauth/types";
 

package/src/index.ts

@@ -18,7 +18,8 @@ export * from "./usage/google-antigravity";
 export * from "./usage/openai-codex";
 export * from "./usage/zai";
 export * from "./utils/event-stream";
-export * from "./utils/oauth/index";
+export * from "./utils/oauth";
 export * from "./utils/overflow";
+export * from "./utils/retry";
 export * from "./utils/typebox-helpers";
 export * from "./utils/validation";

package/src/models.generated.ts

@@ -1726,7 +1726,7 @@ export const MODELS = {
 				cacheRead: 0,
 				cacheWrite: 0,
 			},
-			contextWindow: 128000,
+			contextWindow: 64000,
 			maxTokens: 16384,
 		} satisfies Model<"openai-completions">,
 		"gpt-4o": {
@@ -1766,24 +1766,6 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 128000,
 		} satisfies Model<"openai-responses">,
-		"gpt-5-codex": {
-			id: "gpt-5-codex",
-			name: "GPT-5-Codex",
-			api: "openai-responses",
-			provider: "github-copilot",
-			baseUrl: "https://api.individual.githubcopilot.com",
-			headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
-			reasoning: true,
-			input: ["text", "image"],
-			cost: {
-				input: 0,
-				output: 0,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 128000,
-			maxTokens: 128000,
-		} satisfies Model<"openai-responses">,
 		"gpt-5-mini": {
 			id: "gpt-5-mini",
 			name: "GPT-5-mini",
@@ -3641,7 +3623,7 @@ export const MODELS = {
 			cost: {
 				input: 1.25,
 				output: 10,
-				cacheRead: 0.13,
+				cacheRead: 0.125,
 				cacheWrite: 0,
 			},
 			contextWindow: 400000,
@@ -3692,7 +3674,7 @@ export const MODELS = {
 			cost: {
 				input: 0.25,
 				output: 2,
-				cacheRead: 0.03,
+				cacheRead: 0.025,
 				cacheWrite: 0,
 			},
 			contextWindow: 400000,
@@ -3709,7 +3691,7 @@ export const MODELS = {
 			cost: {
 				input: 0.05,
 				output: 0.4,
-				cacheRead: 0.01,
+				cacheRead: 0.005,
 				cacheWrite: 0,
 			},
 			contextWindow: 400000,
@@ -4314,6 +4296,23 @@ export const MODELS = {
 			contextWindow: 204800,
 			maxTokens: 131072,
 		} satisfies Model<"openai-completions">,
+		"glm-4.7": {
+			id: "glm-4.7",
+			name: "GLM-4.7",
+			api: "openai-completions",
+			provider: "opencode",
+			baseUrl: "https://opencode.ai/zen/v1",
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0.6,
+				output: 2.2,
+				cacheRead: 0.1,
+				cacheWrite: 0,
+			},
+			contextWindow: 204800,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
 		"glm-4.7-free": {
 			id: "glm-4.7-free",
 			name: "GLM-4.7",
@@ -4615,7 +4614,7 @@ export const MODELS = {
 			input: ["text"],
 			cost: {
 				input: 0.09,
-				output: 0.39999999999999997,
+				output: 0.44999999999999996,
 				cacheRead: 0,
 				cacheWrite: 0,
 			},
@@ -5028,7 +5027,7 @@ export const MODELS = {
 				cacheWrite: 0,
 			},
 			contextWindow: 262144,
-			maxTokens: 16384,
+			maxTokens: 32768,
 		} satisfies Model<"openai-completions">,
 		"cohere/command-r-08-2024": {
 			id: "cohere/command-r-08-2024",
@@ -5297,7 +5296,7 @@ export const MODELS = {
 				input: 0.09999999999999999,
 				output: 0.39999999999999997,
 				cacheRead: 0.024999999999999998,
-				cacheWrite: 0.0833,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
 			maxTokens: 8192,
@@ -5311,10 +5310,10 @@ export const MODELS = {
 			reasoning: false,
 			input: ["text", "image"],
 			cost: {
-				input: 0,
-				output: 0,
-				cacheRead: 0,
-				cacheWrite: 0,
+				input: 0.09999999999999999,
+				output: 0.39999999999999997,
+				cacheRead: 0.024999999999999998,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
 			maxTokens: 8192,
@@ -5348,7 +5347,7 @@ export const MODELS = {
 				input: 0.3,
 				output: 2.5,
 				cacheRead: 0.03,
-				cacheWrite: 0.08333333333333334,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
 			maxTokens: 65535,
@@ -5365,7 +5364,7 @@ export const MODELS = {
 				input: 0.09999999999999999,
 				output: 0.39999999999999997,
 				cacheRead: 0.01,
-				cacheWrite: 0.0833,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
 			maxTokens: 65535,
@@ -5382,10 +5381,10 @@ export const MODELS = {
 				input: 0.09999999999999999,
 				output: 0.39999999999999997,
 				cacheRead: 0.01,
-				cacheWrite: 0.0833,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
-			maxTokens: 65536,
+			maxTokens: 65535,
 		} satisfies Model<"openai-completions">,
 		"google/gemini-2.5-flash-preview-09-2025": {
 			id: "google/gemini-2.5-flash-preview-09-2025",
@@ -5398,8 +5397,8 @@ export const MODELS = {
 			cost: {
 				input: 0.3,
 				output: 2.5,
-				cacheRead: 0.075,
-				cacheWrite: 0.0833,
+				cacheRead: 0.03,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
 			maxTokens: 65535,
@@ -5432,7 +5431,7 @@ export const MODELS = {
 			cost: {
 				input: 1.25,
 				output: 10,
-				cacheRead: 0.31,
+				cacheRead: 0.125,
 				cacheWrite: 0.375,
 			},
 			contextWindow: 1048576,
@@ -5449,7 +5448,7 @@ export const MODELS = {
 			cost: {
 				input: 1.25,
 				output: 10,
-				cacheRead: 0.31,
+				cacheRead: 0.125,
 				cacheWrite: 0.375,
 			},
 			contextWindow: 1048576,
@@ -5467,7 +5466,7 @@ export const MODELS = {
 				input: 0.5,
 				output: 3,
 				cacheRead: 0.049999999999999996,
-				cacheWrite: 0,
+				cacheWrite: 0.08333333333333333,
 			},
 			contextWindow: 1048576,
 			maxTokens: 65535,
@@ -5754,12 +5753,12 @@ export const MODELS = {
 			input: ["text"],
 			cost: {
 				input: 0.27,
-				output: 1.12,
+				output: 1.1,
 				cacheRead: 0,
 				cacheWrite: 0,
 			},
 			contextWindow: 196608,
-			maxTokens: 65536,
+			maxTokens: 196608,
 		} satisfies Model<"openai-completions">,
 		"mistralai/codestral-2508": {
 			id: "mistralai/codestral-2508",
@@ -6048,7 +6047,7 @@ export const MODELS = {
 				cacheWrite: 0,
 			},
 			contextWindow: 131072,
-			maxTokens: 131072,
+			maxTokens: 16384,
 		} satisfies Model<"openai-completions">,
 		"mistralai/mistral-saba": {
 			id: "mistralai/mistral-saba",
@@ -7869,6 +7868,23 @@ export const MODELS = {
 			contextWindow: 262144,
 			maxTokens: 4096,
 		} satisfies Model<"openai-completions">,
+		"qwen/qwen3-vl-235b-a22b-thinking": {
+			id: "qwen/qwen3-vl-235b-a22b-thinking",
+			name: "Qwen: Qwen3 VL 235B A22B Thinking",
+			api: "openai-completions",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			reasoning: true,
+			input: ["text", "image"],
+			cost: {
+				input: 0.44999999999999996,
+				output: 3.5,
+				cacheRead: 0,
+				cacheWrite: 0,
+			},
+			contextWindow: 262144,
+			maxTokens: 262144,
+		} satisfies Model<"openai-completions">,
 		"qwen/qwen3-vl-30b-a3b-instruct": {
 			id: "qwen/qwen3-vl-30b-a3b-instruct",
 			name: "Qwen: Qwen3 VL 30B A3B Instruct",
@@ -8430,6 +8446,23 @@ export const MODELS = {
 			contextWindow: 202752,
 			maxTokens: 65535,
 		} satisfies Model<"openai-completions">,
+		"z-ai/glm-4.7-flash": {
+			id: "z-ai/glm-4.7-flash",
+			name: "Z.AI: GLM 4.7 Flash",
+			api: "openai-completions",
+			provider: "openrouter",
+			baseUrl: "https://openrouter.ai/api/v1",
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0.07,
+				output: 0.39999999999999997,
+				cacheRead: 0.01,
+				cacheWrite: 0,
+			},
+			contextWindow: 200000,
+			maxTokens: 131072,
+		} satisfies Model<"openai-completions">,
 	},
 	"vercel-ai-gateway": {
 		"alibaba/qwen-3-14b": {
@@ -8562,7 +8595,7 @@ export const MODELS = {
 			cost: {
 				input: 1,
 				output: 5,
-				cacheRead: 0,
+				cacheRead: 0.19999999999999998,
 				cacheWrite: 0,
 			},
 			contextWindow: 1000000,
@@ -8619,23 +8652,6 @@ export const MODELS = {
 			contextWindow: 200000,
 			maxTokens: 4096,
 		} satisfies Model<"anthropic-messages">,
-		"anthropic/claude-3-opus": {
-			id: "anthropic/claude-3-opus",
-			name: "Claude 3 Opus",
-			api: "anthropic-messages",
-			provider: "vercel-ai-gateway",
-			baseUrl: "https://ai-gateway.vercel.sh",
-			reasoning: false,
-			input: ["text", "image"],
-			cost: {
-				input: 15,
-				output: 75,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 200000,
-			maxTokens: 8192,
-		} satisfies Model<"anthropic-messages">,
 		"anthropic/claude-3.5-haiku": {
 			id: "anthropic/claude-3.5-haiku",
 			name: "Claude 3.5 Haiku",
@@ -8786,7 +8802,7 @@ export const MODELS = {
 				cacheRead: 0.3,
 				cacheWrite: 3.75,
 			},
-			contextWindow: 200000,
+			contextWindow: 1000000,
 			maxTokens: 64000,
 		} satisfies Model<"anthropic-messages">,
 		"anthropic/claude-sonnet-4.5": {
@@ -8803,7 +8819,7 @@ export const MODELS = {
 				cacheRead: 0.3,
 				cacheWrite: 3.75,
 			},
-			contextWindow: 200000,
+			contextWindow: 1000000,
 			maxTokens: 64000,
 		} satisfies Model<"anthropic-messages">,
 		"bytedance/seed-1.6": {
@@ -8925,40 +8941,6 @@ export const MODELS = {
 			contextWindow: 128000,
 			maxTokens: 64000,
 		} satisfies Model<"anthropic-messages">,
-		"google/gemini-2.0-flash": {
-			id: "google/gemini-2.0-flash",
-			name: "Gemini 2.0 Flash",
-			api: "anthropic-messages",
-			provider: "vercel-ai-gateway",
-			baseUrl: "https://ai-gateway.vercel.sh",
-			reasoning: false,
-			input: ["text", "image"],
-			cost: {
-				input: 0.09999999999999999,
-				output: 0.39999999999999997,
-				cacheRead: 0.024999999999999998,
-				cacheWrite: 0,
-			},
-			contextWindow: 1000000,
-			maxTokens: 8192,
-		} satisfies Model<"anthropic-messages">,
-		"google/gemini-2.0-flash-lite": {
-			id: "google/gemini-2.0-flash-lite",
-			name: "Gemini 2.0 Flash Lite",
-			api: "anthropic-messages",
-			provider: "vercel-ai-gateway",
-			baseUrl: "https://ai-gateway.vercel.sh",
-			reasoning: false,
-			input: ["text", "image"],
-			cost: {
-				input: 0.075,
-				output: 0.3,
-				cacheRead: 0,
-				cacheWrite: 0,
-			},
-			contextWindow: 1048576,
-			maxTokens: 8192,
-		} satisfies Model<"anthropic-messages">,
 		"google/gemini-2.5-flash": {
 			id: "google/gemini-2.5-flash",
 			name: "Gemini 2.5 Flash",
@@ -8966,15 +8948,15 @@ export const MODELS = {
 			provider: "vercel-ai-gateway",
 			baseUrl: "https://ai-gateway.vercel.sh",
 			reasoning: true,
-			input: ["text", "image"],
+			input: ["text"],
 			cost: {
 				input: 0.3,
 				output: 2.5,
-				cacheRead: 0.03,
+				cacheRead: 0,
 				cacheWrite: 0,
 			},
 			contextWindow: 1000000,
-			maxTokens: 64000,
+			maxTokens: 65536,
 		} satisfies Model<"anthropic-messages">,
 		"google/gemini-2.5-flash-lite": {
 			id: "google/gemini-2.5-flash-lite",
@@ -9034,11 +9016,11 @@ export const MODELS = {
 			provider: "vercel-ai-gateway",
 			baseUrl: "https://ai-gateway.vercel.sh",
 			reasoning: true,
-			input: ["text", "image"],
+			input: ["text"],
 			cost: {
 				input: 1.25,
 				output: 10,
-				cacheRead: 0.125,
+				cacheRead: 0,
 				cacheWrite: 0,
 			},
 			contextWindow: 1048576,
@@ -10449,7 +10431,7 @@ export const MODELS = {
 			cost: {
 				input: 0.19999999999999998,
 				output: 1.1,
-				cacheRead: 0,
+				cacheRead: 0.03,
 				cacheWrite: 0,
 			},
 			contextWindow: 128000,
@@ -10540,6 +10522,23 @@ export const MODELS = {
 			contextWindow: 202752,
 			maxTokens: 120000,
 		} satisfies Model<"anthropic-messages">,
+		"zai/glm-4.7-flashx": {
+			id: "zai/glm-4.7-flashx",
+			name: "GLM 4.7 FlashX",
+			api: "anthropic-messages",
+			provider: "vercel-ai-gateway",
+			baseUrl: "https://ai-gateway.vercel.sh",
+			reasoning: true,
+			input: ["text"],
+			cost: {
+				input: 0.06,
+				output: 0.39999999999999997,
+				cacheRead: 0.01,
+				cacheWrite: 0,
+			},
+			contextWindow: 200000,
+			maxTokens: 128000,
+		} satisfies Model<"anthropic-messages">,
 	},
 	"xai": {
 		"grok-2": {

package/src/providers/anthropic.ts

@@ -727,6 +727,8 @@ function convertMessages(
 	isOAuthToken: boolean,
 ): MessageParam[] {
 	const params: MessageParam[] = [];
+	// Track tool call IDs from skipped assistant messages to also skip their results
+	let skippedToolCallIds: string[] | null = null;
 
 	// Transform messages for cross-provider compatibility
 	const transformedMessages = transformMessages(messages, model);
@@ -779,6 +781,31 @@ function convertMessages(
 			// Skip messages with undefined/null content
 			if (!msg.content || !Array.isArray(msg.content)) continue;
 
+			// When interleaved thinking is enabled, Anthropic requires the last assistant
+			// message to start with a thinking block. If the first content block is a thinking
+			// block with a missing/invalid signature (e.g., from aborted stream), we must skip
+			// the entire message to avoid API rejection. Checking the first non-empty block.
+			const firstContentBlock = msg.content.find(
+				(b) =>
+					(b.type === "text" && b.text.trim().length > 0) ||
+					(b.type === "thinking" && b.thinking.trim().length > 0) ||
+					b.type === "toolCall",
+			);
+			if (
+				firstContentBlock?.type === "thinking" &&
+				(!firstContentBlock.thinkingSignature || firstContentBlock.thinkingSignature.trim().length === 0)
+			) {
+				// Skip this assistant message - it has corrupt thinking that would break the API.
+				// Also track any tool calls in this message so we can skip their results.
+				for (const block of msg.content) {
+					if (block.type === "toolCall") {
+						skippedToolCallIds ??= [];
+						skippedToolCallIds.push(block.id);
+					}
+				}
+				continue;
+			}
+
 			const blocks: Array<ContentBlockParam & CacheControlBlock> = [];
 
 			for (const block of msg.content) {
@@ -824,23 +851,27 @@ function convertMessages(
 			const toolResults: Array<ContentBlockParam & CacheControlBlock> = [];
 
 			// Add the current tool result
-			toolResults.push({
-				type: "tool_result",
-				tool_use_id: sanitizeToolCallId(msg.toolCallId),
-				content: convertContentBlocks(msg.content),
-				is_error: msg.isError,
-			});
+			if (!skippedToolCallIds?.includes(msg.toolCallId)) {
+				toolResults.push({
+					type: "tool_result",
+					tool_use_id: sanitizeToolCallId(msg.toolCallId),
+					content: convertContentBlocks(msg.content),
+					is_error: msg.isError,
+				});
+			}
 
 			// Look ahead for consecutive toolResult messages
 			let j = i + 1;
 			while (j < transformedMessages.length && transformedMessages[j].role === "toolResult") {
 				const nextMsg = transformedMessages[j] as ToolResultMessage; // We know it's a toolResult
-				toolResults.push({
-					type: "tool_result",
-					tool_use_id: sanitizeToolCallId(nextMsg.toolCallId),
-					content: convertContentBlocks(nextMsg.content),
-					is_error: nextMsg.isError,
-				});
+				if (!skippedToolCallIds?.includes(nextMsg.toolCallId)) {
+					toolResults.push({
+						type: "tool_result",
+						tool_use_id: sanitizeToolCallId(nextMsg.toolCallId),
+						content: convertContentBlocks(nextMsg.content),
+						is_error: nextMsg.isError,
+					});
+				}
 				j++;
 			}
 

package/src/providers/cursor.ts

@@ -3,7 +3,6 @@ import { appendFile } from "node:fs/promises";
 import http2 from "node:http2";
 import { create, fromBinary, fromJson, type JsonValue, toBinary, toJson } from "@bufbuild/protobuf";
 import { ValueSchema } from "@bufbuild/protobuf/wkt";
-import JSON5 from "json5";
 import { calculateCost } from "@oh-my-pi/pi-ai/models";
 import type {
 	Api,
@@ -27,6 +26,7 @@ import type {
 import { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream";
 import { parseStreamingJson } from "@oh-my-pi/pi-ai/utils/json-parse";
 import { formatErrorMessageWithRetryAfter } from "@oh-my-pi/pi-ai/utils/retry-after";
+import JSON5 from "json5";
 import type { McpToolDefinition } from "./cursor/gen/agent_pb";
 import {
 	AgentClientMessageSchema,

package/src/providers/google-gemini-cli.ts

@@ -6,7 +6,6 @@
 
 import { createHash } from "node:crypto";
 import type { Content, ThinkingConfig } from "@google/genai";
-import { abortableSleep } from "@oh-my-pi/pi-utils";
 import { calculateCost } from "@oh-my-pi/pi-ai/models";
 import type {
 	Api,
@@ -21,6 +20,7 @@ import type {
 } from "@oh-my-pi/pi-ai/types";
 import { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream";
 import { sanitizeSurrogates } from "@oh-my-pi/pi-ai/utils/sanitize-unicode";
+import { abortableSleep } from "@oh-my-pi/pi-utils";
 import {
 	convertMessages,
 	convertTools,

package/src/providers/openai-codex-responses.ts

@@ -1,14 +1,4 @@
 import os from "node:os";
-import { abortableSleep } from "@oh-my-pi/pi-utils";
-import type {
-	ResponseFunctionToolCall,
-	ResponseInput,
-	ResponseInputContent,
-	ResponseInputImage,
-	ResponseInputText,
-	ResponseOutputMessage,
-	ResponseReasoningItem,
-} from "openai/resources/responses/responses";
 import { calculateCost } from "@oh-my-pi/pi-ai/models";
 import { getEnvApiKey } from "@oh-my-pi/pi-ai/stream";
 import type {
@@ -28,6 +18,16 @@ import { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream"
 import { parseStreamingJson } from "@oh-my-pi/pi-ai/utils/json-parse";
 import { formatErrorMessageWithRetryAfter } from "@oh-my-pi/pi-ai/utils/retry-after";
 import { sanitizeSurrogates } from "@oh-my-pi/pi-ai/utils/sanitize-unicode";
+import { abortableSleep } from "@oh-my-pi/pi-utils";
+import type {
+	ResponseFunctionToolCall,
+	ResponseInput,
+	ResponseInputContent,
+	ResponseInputImage,
+	ResponseInputText,
+	ResponseOutputMessage,
+	ResponseReasoningItem,
+} from "openai/resources/responses/responses";
 import packageJson from "../../package.json" with { type: "json" };
 import {
 	CODEX_BASE_URL,

package/src/providers/openai-completions.ts

@@ -1,13 +1,3 @@
-import OpenAI from "openai";
-import type {
-	ChatCompletionAssistantMessageParam,
-	ChatCompletionChunk,
-	ChatCompletionContentPart,
-	ChatCompletionContentPartImage,
-	ChatCompletionContentPartText,
-	ChatCompletionMessageParam,
-	ChatCompletionToolMessageParam,
-} from "openai/resources/chat/completions";
 import { calculateCost } from "@oh-my-pi/pi-ai/models";
 import { getEnvApiKey } from "@oh-my-pi/pi-ai/stream";
 import type {
@@ -29,6 +19,16 @@ import { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream"
 import { parseStreamingJson } from "@oh-my-pi/pi-ai/utils/json-parse";
 import { formatErrorMessageWithRetryAfter } from "@oh-my-pi/pi-ai/utils/retry-after";
 import { sanitizeSurrogates } from "@oh-my-pi/pi-ai/utils/sanitize-unicode";
+import OpenAI from "openai";
+import type {
+	ChatCompletionAssistantMessageParam,
+	ChatCompletionChunk,
+	ChatCompletionContentPart,
+	ChatCompletionContentPartImage,
+	ChatCompletionContentPartText,
+	ChatCompletionMessageParam,
+	ChatCompletionToolMessageParam,
+} from "openai/resources/chat/completions";
 import { transformMessages } from "./transform-messages";
 
 /**

package/src/providers/openai-responses.ts

@@ -1,15 +1,3 @@
-import OpenAI from "openai";
-import type {
-	Tool as OpenAITool,
-	ResponseCreateParamsStreaming,
-	ResponseFunctionToolCall,
-	ResponseInput,
-	ResponseInputContent,
-	ResponseInputImage,
-	ResponseInputText,
-	ResponseOutputMessage,
-	ResponseReasoningItem,
-} from "openai/resources/responses/responses";
 import { calculateCost } from "@oh-my-pi/pi-ai/models";
 import { getEnvApiKey } from "@oh-my-pi/pi-ai/stream";
 import type {
@@ -29,6 +17,18 @@ import { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream"
 import { parseStreamingJson } from "@oh-my-pi/pi-ai/utils/json-parse";
 import { formatErrorMessageWithRetryAfter } from "@oh-my-pi/pi-ai/utils/retry-after";
 import { sanitizeSurrogates } from "@oh-my-pi/pi-ai/utils/sanitize-unicode";
+import OpenAI from "openai";
+import type {
+	Tool as OpenAITool,
+	ResponseCreateParamsStreaming,
+	ResponseFunctionToolCall,
+	ResponseInput,
+	ResponseInputContent,
+	ResponseInputImage,
+	ResponseInputText,
+	ResponseOutputMessage,
+	ResponseReasoningItem,
+} from "openai/resources/responses/responses";
 import { transformMessages } from "./transform-messages";
 
 /** Fast deterministic hash to shorten long strings */

package/src/utils/oauth/github-copilot.ts

@@ -2,8 +2,8 @@
  * GitHub Copilot OAuth flow
  */
 
-import { abortableSleep } from "@oh-my-pi/pi-utils";
 import { getModels } from "@oh-my-pi/pi-ai/models";
+import { abortableSleep } from "@oh-my-pi/pi-utils";
 import type { OAuthCredentials } from "./types";
 
 const decode = (s: string) => atob(s);

package/src/utils/retry.ts

@@ -0,0 +1,78 @@
+type ErrorLike = {
+	message?: string;
+	name?: string;
+	status?: number;
+	statusCode?: number;
+	response?: { status?: number };
+	cause?: unknown;
+};
+
+const TRANSIENT_MESSAGE_PATTERN =
+	/overloaded|rate.?limit|usage.?limit|too many requests|service.?unavailable|server error|internal error|connection.?error|fetch failed/i;
+
+const VALIDATION_MESSAGE_PATTERN =
+	/invalid|validation|bad request|unsupported|schema|missing required|not found|unauthorized|forbidden/i;
+
+/**
+ * Identify errors that should be retried (timeouts, 5xx, 408, 429, transient network failures).
+ */
+export function isRetryableError(error: unknown): boolean {
+	const info = error as ErrorLike | null;
+	const message = info?.message ?? "";
+	const name = info?.name ?? "";
+	if (name === "AbortError" || /timeout|timed out|aborted/i.test(message)) return true;
+
+	const status = extractHttpStatusFromError(error);
+	if (status !== undefined) {
+		if (status >= 500) return true;
+		if (status === 408 || status === 429) return true;
+		if (status >= 400 && status < 500) return false;
+	}
+
+	if (VALIDATION_MESSAGE_PATTERN.test(message)) return false;
+
+	return TRANSIENT_MESSAGE_PATTERN.test(message);
+}
+
+export function extractHttpStatusFromError(error: unknown, depth = 0): number | undefined {
+	if (!error || typeof error !== "object" || depth > 3) return undefined;
+	const info = error as ErrorLike;
+	const status =
+		info.status ??
+		info.statusCode ??
+		(info.response && typeof info.response === "object" ? info.response.status : undefined);
+	if (typeof status === "number" && status >= 100 && status <= 599) {
+		return status;
+	}
+
+	if (info.message) {
+		const extracted = extractStatusFromMessage(info.message);
+		if (extracted !== undefined) return extracted;
+	}
+
+	if (info.cause) {
+		return extractHttpStatusFromError(info.cause, depth + 1);
+	}
+
+	return undefined;
+}
+
+function extractStatusFromMessage(message: string): number | undefined {
+	const patterns = [
+		/error\s*\((\d{3})\)/i,
+		/status\s*[:=]?\s*(\d{3})/i,
+		/\bhttp\s*(\d{3})\b/i,
+		/\b(\d{3})\s*(?:status|error)\b/i,
+	];
+
+	for (const pattern of patterns) {
+		const match = pattern.exec(message);
+		if (!match) continue;
+		const value = Number(match[1]);
+		if (Number.isFinite(value) && value >= 100 && value <= 599) {
+			return value;
+		}
+	}
+
+	return undefined;
+}

package/tsconfig.json

@@ -1,42 +0,0 @@
-{
-	"compilerOptions": {
-		"target": "ES2024",
-		"module": "ESNext",
-		"lib": [
-			"ES2024"
-		],
-		"strict": true,
-		"esModuleInterop": true,
-		"skipLibCheck": true,
-		"forceConsistentCasingInFileNames": true,
-		"moduleResolution": "Bundler",
-		"resolveJsonModule": true,
-		"allowImportingTsExtensions": true,
-		"experimentalDecorators": true,
-		"emitDecoratorMetadata": true,
-		"useDefineForClassFields": false,
-		"types": [
-			"bun",
-			"node"
-		],
-		"noEmit": true,
-		"baseUrl": ".",
-		"paths": {
-			"@oh-my-pi/pi-ai": [
-				"./src/index.ts"
-			],
-			"@oh-my-pi/pi-ai/*": [
-				"./src/*"
-			]
-		}
-	},
-	"include": [
-		"src/**/*.ts"
-	],
-	"exclude": [
-		"node_modules",
-		"dist",
-		"**/*.test.ts",
-		"test/**"
-	]
-}