@oh-my-pi/pi-ai 7.0.0 → 8.0.1

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-ai",
-	"version": "7.0.0",
+	"version": "8.0.1",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"type": "module",
 	"main": "./src/index.ts",
@@ -10,14 +10,16 @@
 	},
 	"files": [
 		"src",
-		"README.md"
+		"README.md",
+		"tsconfig.json"
 	],
 	"scripts": {
 		"generate-models": "bun scripts/generate-models.ts",
-		"test": "bun test"
+		"test": "bun test",
+		"prepublishOnly": "cp tsconfig.publish.json tsconfig.json"
 	},
 	"dependencies": {
-		"@oh-my-pi/pi-utils": "7.0.0",
+		"@oh-my-pi/pi-utils": "8.0.0",
 		"@anthropic-ai/sdk": "0.71.2",
 		"@aws-sdk/client-bedrock-runtime": "^3.968.0",
 		"@bufbuild/protobuf": "^2.10.2",

package/src/providers/amazon-bedrock.ts

@@ -18,7 +18,7 @@ import {
 	ToolResultStatus,
 } from "@aws-sdk/client-bedrock-runtime";
 
-import { calculateCost } from "../models";
+import { calculateCost } from "$ai/models";
 import type {
 	Api,
 	AssistantMessage,
@@ -34,10 +34,10 @@ import type {
 	Tool,
 	ToolCall,
 	ToolResultMessage,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { parseStreamingJson } from "$ai/utils/json-parse";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import { transformMessages } from "./transform-messages";
 
 export interface BedrockOptions extends StreamOptions {

package/src/providers/anthropic.ts

@@ -4,8 +4,8 @@ import type {
 	MessageCreateParamsStreaming,
 	MessageParam,
 } from "@anthropic-ai/sdk/resources/messages";
-import { calculateCost } from "../models";
-import { getEnvApiKey, OUTPUT_FALLBACK_BUFFER } from "../stream";
+import { calculateCost } from "$ai/models";
+import { getEnvApiKey, OUTPUT_FALLBACK_BUFFER } from "$ai/stream";
 import type {
 	Api,
 	AssistantMessage,
@@ -21,11 +21,11 @@ import type {
 	Tool,
 	ToolCall,
 	ToolResultMessage,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { parseStreamingJson } from "$ai/utils/json-parse";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 
 import { transformMessages } from "./transform-messages";
 

package/src/providers/cursor.ts

@@ -4,7 +4,7 @@ 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 "../models";
+import { calculateCost } from "$ai/models";
 import type {
 	Api,
 	AssistantMessage,
@@ -23,10 +23,10 @@ import type {
 	Tool,
 	ToolCall,
 	ToolResultMessage,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { parseStreamingJson } from "$ai/utils/json-parse";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
 import type { McpToolDefinition } from "./cursor/gen/agent_pb";
 import {
 	AgentClientMessageSchema,

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

@@ -6,8 +6,8 @@ import type {
 	UsageProvider,
 	UsageReport,
 	UsageWindow,
-} from "../usage";
-import { refreshGoogleCloudToken } from "../utils/oauth/google-gemini-cli";
+} from "$ai/usage";
+import { refreshGoogleCloudToken } from "$ai/utils/oauth/google-gemini-cli";
 
 const DEFAULT_ENDPOINT = "https://cloudcode-pa.googleapis.com";
 const CACHE_TTL_MS = 60_000;

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

@@ -7,7 +7,7 @@
 import { createHash } from "node:crypto";
 import type { Content, ThinkingConfig } from "@google/genai";
 import { abortableSleep } from "@oh-my-pi/pi-utils";
-import { calculateCost } from "../models";
+import { calculateCost } from "$ai/models";
 import type {
 	Api,
 	AssistantMessage,
@@ -18,9 +18,9 @@ import type {
 	TextContent,
 	ThinkingContent,
 	ToolCall,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import {
 	convertMessages,
 	convertTools,

package/src/providers/google-shared.ts

@@ -3,8 +3,8 @@
  */
 
 import { type Content, FinishReason, FunctionCallingConfigMode, type Part, type Schema } from "@google/genai";
-import type { Context, ImageContent, Model, StopReason, TextContent, Tool } from "../types";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+import type { Context, ImageContent, Model, StopReason, TextContent, Tool } from "$ai/types";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import { transformMessages } from "./transform-messages";
 
 type GoogleApiType = "google-generative-ai" | "google-gemini-cli" | "google-vertex";

package/src/providers/google-vertex.ts

@@ -5,7 +5,7 @@ import {
 	type ThinkingConfig,
 	ThinkingLevel,
 } from "@google/genai";
-import { calculateCost } from "../models";
+import { calculateCost } from "$ai/models";
 import type {
 	Api,
 	AssistantMessage,
@@ -16,10 +16,10 @@ import type {
 	TextContent,
 	ThinkingContent,
 	ToolCall,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import type { GoogleThinkingLevel } from "./google-gemini-cli";
 import {
 	convertMessages,

package/src/providers/google.ts

@@ -4,8 +4,8 @@ import {
 	GoogleGenAI,
 	type ThinkingConfig,
 } from "@google/genai";
-import { calculateCost } from "../models";
-import { getEnvApiKey } from "../stream";
+import { calculateCost } from "$ai/models";
+import { getEnvApiKey } from "$ai/stream";
 import type {
 	Api,
 	AssistantMessage,
@@ -16,10 +16,10 @@ import type {
 	TextContent,
 	ThinkingContent,
 	ToolCall,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import type { GoogleThinkingLevel } from "./google-gemini-cli";
 import {
 	convertMessages,

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

@@ -9,9 +9,8 @@ import type {
 	ResponseOutputMessage,
 	ResponseReasoningItem,
 } from "openai/resources/responses/responses";
-import packageJson from "../../package.json" with { type: "json" };
-import { calculateCost } from "../models";
-import { getEnvApiKey } from "../stream";
+import { calculateCost } from "$ai/models";
+import { getEnvApiKey } from "$ai/stream";
 import type {
 	Api,
 	AssistantMessage,
@@ -24,11 +23,12 @@ import type {
 	ThinkingContent,
 	Tool,
 	ToolCall,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { parseStreamingJson } from "$ai/utils/json-parse";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
+import packageJson from "../../package.json" with { type: "json" };
 import {
 	CODEX_BASE_URL,
 	JWT_CLAIM_PATH,

package/src/providers/openai-completions.ts

@@ -8,8 +8,8 @@ import type {
 	ChatCompletionMessageParam,
 	ChatCompletionToolMessageParam,
 } from "openai/resources/chat/completions";
-import { calculateCost } from "../models";
-import { getEnvApiKey } from "../stream";
+import { calculateCost } from "$ai/models";
+import { getEnvApiKey } from "$ai/stream";
 import type {
 	AssistantMessage,
 	Context,
@@ -23,11 +23,12 @@ import type {
 	ThinkingContent,
 	Tool,
 	ToolCall,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+	ToolResultMessage,
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { parseStreamingJson } from "$ai/utils/json-parse";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import { transformMessages } from "./transform-messages";
 
 /**
@@ -464,7 +465,7 @@ function maybeAddOpenRouterAnthropicCacheControl(
 	}
 }
 
-function convertMessages(
+export function convertMessages(
 	model: Model<"openai-completions">,
 	context: Context,
 	compat: Required<OpenAICompat>,
@@ -481,7 +482,8 @@ function convertMessages(
 
 	let lastRole: string | null = null;
 
-	for (const msg of transformedMessages) {
+	for (let i = 0; i < transformedMessages.length; i++) {
+		const msg = transformedMessages[i];
 		// Some providers (e.g. Mistral/Devstral) don't allow user messages directly after tool results
 		// Insert a synthetic assistant message to bridge the gap
 		if (compat.requiresAssistantAfterToolResult && lastRole === "toolResult" && msg.role === "user") {
@@ -605,55 +607,73 @@ function convertMessages(
 			}
 			params.push(assistantMsg);
 		} else if (msg.role === "toolResult") {
-			// Extract text and image content
-			const textResult = msg.content
-				.filter((c) => c.type === "text")
-				.map((c) => (c as any).text)
-				.join("\n");
-			const hasImages = msg.content.some((c) => c.type === "image");
-
-			// Always send tool result with text (or placeholder if only images)
-			const hasText = textResult.length > 0;
-			// Some providers (e.g. Mistral) require the 'name' field in tool results
-			const toolResultMsg: ChatCompletionToolMessageParam = {
-				role: "tool",
-				content: sanitizeSurrogates(hasText ? textResult : "(see attached image)"),
-				tool_call_id: normalizeMistralToolId(msg.toolCallId, compat.requiresMistralToolIds),
-			};
-			if (compat.requiresToolResultName && msg.toolName) {
-				(toolResultMsg as any).name = msg.toolName;
+			// Batch consecutive tool results and collect all images
+			const imageBlocks: Array<{ type: "image_url"; image_url: { url: string } }> = [];
+			let j = i;
+
+			for (; j < transformedMessages.length && transformedMessages[j].role === "toolResult"; j++) {
+				const toolMsg = transformedMessages[j] as ToolResultMessage;
+
+				// Extract text and image content
+				const textResult = toolMsg.content
+					.filter((c) => c.type === "text")
+					.map((c) => (c as any).text)
+					.join("\n");
+				const hasImages = toolMsg.content.some((c) => c.type === "image");
+
+				// Always send tool result with text (or placeholder if only images)
+				const hasText = textResult.length > 0;
+				// Some providers (e.g. Mistral) require the 'name' field in tool results
+				const toolResultMsg: ChatCompletionToolMessageParam = {
+					role: "tool",
+					content: sanitizeSurrogates(hasText ? textResult : "(see attached image)"),
+					tool_call_id: normalizeMistralToolId(toolMsg.toolCallId, compat.requiresMistralToolIds),
+				};
+				if (compat.requiresToolResultName && toolMsg.toolName) {
+					(toolResultMsg as any).name = toolMsg.toolName;
+				}
+				params.push(toolResultMsg);
+
+				if (hasImages && model.input.includes("image")) {
+					for (const block of toolMsg.content) {
+						if (block.type === "image") {
+							imageBlocks.push({
+								type: "image_url",
+								image_url: {
+									url: `data:${(block as any).mimeType};base64,${(block as any).data}`,
+								},
+							});
+						}
+					}
+				}
 			}
-			params.push(toolResultMsg);
-
-			// If there are images and model supports them, send a follow-up user message with images
-			if (hasImages && model.input.includes("image")) {
-				const contentBlocks: Array<
-					{ type: "text"; text: string } | { type: "image_url"; image_url: { url: string } }
-				> = [];
-
-				// Add text prefix
-				contentBlocks.push({
-					type: "text",
-					text: "Attached image(s) from tool result:",
-				});
 
-				// Add images
-				for (const block of msg.content) {
-					if (block.type === "image") {
-						contentBlocks.push({
-							type: "image_url",
-							image_url: {
-								url: `data:${(block as any).mimeType};base64,${(block as any).data}`,
-							},
-						});
-					}
+			i = j - 1;
+
+			// After all consecutive tool results, add a single user message with all images
+			if (imageBlocks.length > 0) {
+				if (compat.requiresAssistantAfterToolResult) {
+					params.push({
+						role: "assistant",
+						content: "I have processed the tool results.",
+					});
 				}
 
 				params.push({
 					role: "user",
-					content: contentBlocks,
+					content: [
+						{
+							type: "text",
+							text: "Attached image(s) from tool result:",
+						},
+						...imageBlocks,
+					],
 				});
+				lastRole = "user";
+			} else {
+				lastRole = "toolResult";
 			}
+			continue;
 		}
 
 		lastRole = msg.role;

package/src/providers/openai-responses.ts

@@ -10,8 +10,8 @@ import type {
 	ResponseOutputMessage,
 	ResponseReasoningItem,
 } from "openai/resources/responses/responses";
-import { calculateCost } from "../models";
-import { getEnvApiKey } from "../stream";
+import { calculateCost } from "$ai/models";
+import { getEnvApiKey } from "$ai/stream";
 import type {
 	Api,
 	AssistantMessage,
@@ -24,11 +24,11 @@ import type {
 	ThinkingContent,
 	Tool,
 	ToolCall,
-} from "../types";
-import { AssistantMessageEventStream } from "../utils/event-stream";
-import { parseStreamingJson } from "../utils/json-parse";
-import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode";
+} from "$ai/types";
+import { AssistantMessageEventStream } from "$ai/utils/event-stream";
+import { parseStreamingJson } from "$ai/utils/json-parse";
+import { formatErrorMessageWithRetryAfter } from "$ai/utils/retry-after";
+import { sanitizeSurrogates } from "$ai/utils/sanitize-unicode";
 import { transformMessages } from "./transform-messages";
 
 /** Fast deterministic hash to shorten long strings */
@@ -507,6 +507,15 @@ function convertMessages(
 			}
 		} else if (msg.role === "assistant") {
 			const output: ResponseInput = [];
+			const assistantMsg = msg as AssistantMessage;
+
+			// Check if this message is from a different model (same provider, different model ID).
+			// For such messages, tool call IDs with fc_ prefix need to be stripped to avoid
+			// OpenAI's reasoning/function_call pairing validation errors.
+			const isDifferentModel =
+				assistantMsg.model !== model.id &&
+				assistantMsg.provider === model.provider &&
+				assistantMsg.api === model.api;
 
 			for (const block of msg.content) {
 				// Do not submit thinking blocks if the completion had an error (i.e. abort)
@@ -535,11 +544,19 @@ function convertMessages(
 				} else if (block.type === "toolCall" && msg.stopReason !== "error") {
 					const toolCall = block as ToolCall;
 					const normalized = normalizeResponsesToolCallId(toolCall.id);
+					const callId = normalized.callId;
+					// For different-model messages, set id to undefined to avoid pairing validation.
+					// OpenAI tracks which fc_xxx IDs were paired with rs_xxx reasoning items.
+					// By omitting the id, we avoid triggering that validation (like cross-provider does).
+					let itemId: string | undefined = normalized.itemId;
+					if (isDifferentModel && itemId?.startsWith("fc_")) {
+						itemId = undefined;
+					}
 					knownCallIds.add(normalized.callId);
 					output.push({
 						type: "function_call",
-						id: normalized.itemId,
-						call_id: normalized.callId,
+						id: itemId,
+						call_id: callId,
 						name: toolCall.name,
 						arguments: JSON.stringify(toolCall.arguments),
 					});

package/src/providers/transform-messages.ts

@@ -1,4 +1,4 @@
-import type { Api, AssistantMessage, Message, Model, ToolCall, ToolResultMessage } from "../types";
+import type { Api, AssistantMessage, Message, Model, ToolCall, ToolResultMessage } from "$ai/types";
 
 /**
  * Normalize tool call ID for cross-provider compatibility.

package/src/usage/claude.ts

@@ -7,7 +7,7 @@ import type {
 	UsageReport,
 	UsageStatus,
 	UsageWindow,
-} from "../usage";
+} from "$ai/usage";
 
 const DEFAULT_ENDPOINT = "https://api.anthropic.com/api/oauth";
 const DEFAULT_CACHE_TTL_MS = 60_000;

package/src/usage/github-copilot.ts

@@ -14,7 +14,7 @@ import type {
 	UsageReport,
 	UsageStatus,
 	UsageWindow,
-} from "../usage";
+} from "$ai/usage";
 
 const COPILOT_HEADERS = {
 	"User-Agent": "GitHubCopilotChat/0.35.0",

package/src/usage/google-antigravity.ts

@@ -7,8 +7,8 @@ import type {
 	UsageReport,
 	UsageStatus,
 	UsageWindow,
-} from "../usage";
-import { refreshAntigravityToken } from "../utils/oauth/google-antigravity";
+} from "$ai/usage";
+import { refreshAntigravityToken } from "$ai/utils/oauth/google-antigravity";
 
 interface AntigravityQuotaInfo {
 	remainingFraction?: number;

package/src/usage/openai-codex.ts

@@ -1,5 +1,5 @@
 import { Buffer } from "node:buffer";
-import { CODEX_BASE_URL } from "../providers/openai-codex/constants";
+import { CODEX_BASE_URL } from "$ai/providers/openai-codex/constants";
 import type {
 	UsageAmount,
 	UsageCache,
@@ -9,7 +9,7 @@ import type {
 	UsageProvider,
 	UsageReport,
 	UsageWindow,
-} from "../usage";
+} from "$ai/usage";
 
 const CODEX_USAGE_PATH = "wham/usage";
 const DEFAULT_CACHE_TTL_MS = 60_000;

package/src/usage/zai.ts

@@ -7,7 +7,7 @@ import type {
 	UsageReport,
 	UsageStatus,
 	UsageWindow,
-} from "../usage";
+} from "$ai/usage";
 
 const DEFAULT_ENDPOINT = "https://api.z.ai";
 const QUOTA_PATH = "/api/monitor/usage/quota/limit";

package/src/utils/event-stream.ts

@@ -1,4 +1,4 @@
-import type { AssistantMessage, AssistantMessageEvent } from "../types";
+import type { AssistantMessage, AssistantMessageEvent } from "$ai/types";
 
 // Generic event stream class for async iteration
 export class EventStream<T, R = T> implements AsyncIterable<T> {

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

@@ -3,7 +3,7 @@
  */
 
 import { abortableSleep } from "@oh-my-pi/pi-utils";
-import { getModels } from "../../models";
+import { getModels } from "$ai/models";
 import type { OAuthCredentials } from "./types";
 
 const decode = (s: string) => atob(s);

package/src/utils/overflow.ts

@@ -1,4 +1,4 @@
-import type { AssistantMessage } from "../types";
+import type { AssistantMessage } from "$ai/types";
 
 /**
  * Regex patterns to detect context overflow errors from different providers.

package/src/utils/validation.ts

@@ -5,7 +5,7 @@ import addFormatsModule from "ajv-formats";
 const Ajv = (AjvModule as any).default || AjvModule;
 const addFormats = (addFormatsModule as any).default || addFormatsModule;
 
-import type { Tool, ToolCall } from "../types";
+import type { Tool, ToolCall } from "$ai/types";
 
 // ============================================================================
 // Type Coercion Utilities

package/tsconfig.json

@@ -0,0 +1,25 @@
+{
+	"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": {
+			"$ai/*": ["./src/*"]
+		}
+	},
+	"include": ["src/**/*.ts"],
+	"exclude": ["node_modules", "dist", "**/*.test.ts", "test/**"]
+}