pi-usereq 0.6.0 → 0.10.0

package/pi.dev-src/pi-mono/ai/src/types.ts

@@ -0,0 +1,417 @@
+import type { AssistantMessageEventStream } from "./utils/event-stream.js";
+
+export type { AssistantMessageEventStream } from "./utils/event-stream.js";
+
+export type KnownApi =
+	| "openai-completions"
+	| "mistral-conversations"
+	| "openai-responses"
+	| "azure-openai-responses"
+	| "openai-codex-responses"
+	| "anthropic-messages"
+	| "bedrock-converse-stream"
+	| "google-generative-ai"
+	| "google-gemini-cli"
+	| "google-vertex";
+
+export type Api = KnownApi | (string & {});
+
+export type KnownProvider =
+	| "amazon-bedrock"
+	| "anthropic"
+	| "google"
+	| "google-gemini-cli"
+	| "google-antigravity"
+	| "google-vertex"
+	| "openai"
+	| "azure-openai-responses"
+	| "openai-codex"
+	| "github-copilot"
+	| "xai"
+	| "groq"
+	| "cerebras"
+	| "openrouter"
+	| "vercel-ai-gateway"
+	| "zai"
+	| "mistral"
+	| "minimax"
+	| "minimax-cn"
+	| "huggingface"
+	| "fireworks"
+	| "opencode"
+	| "opencode-go"
+	| "kimi-coding";
+export type Provider = KnownProvider | string;
+
+export type ThinkingLevel = "minimal" | "low" | "medium" | "high" | "xhigh";
+
+/** Token budgets for each thinking level (token-based providers only) */
+export interface ThinkingBudgets {
+	minimal?: number;
+	low?: number;
+	medium?: number;
+	high?: number;
+}
+
+// Base options all providers share
+export type CacheRetention = "none" | "short" | "long";
+
+export type Transport = "sse" | "websocket" | "auto";
+
+export interface ProviderResponse {
+	status: number;
+	headers: Record<string, string>;
+}
+
+export interface StreamOptions {
+	temperature?: number;
+	maxTokens?: number;
+	signal?: AbortSignal;
+	apiKey?: string;
+	/**
+	 * Preferred transport for providers that support multiple transports.
+	 * Providers that do not support this option ignore it.
+	 */
+	transport?: Transport;
+	/**
+	 * Prompt cache retention preference. Providers map this to their supported values.
+	 * Default: "short".
+	 */
+	cacheRetention?: CacheRetention;
+	/**
+	 * Optional session identifier for providers that support session-based caching.
+	 * Providers can use this to enable prompt caching, request routing, or other
+	 * session-aware features. Ignored by providers that don't support it.
+	 */
+	sessionId?: string;
+	/**
+	 * Optional callback for inspecting or replacing provider payloads before sending.
+	 * Return undefined to keep the payload unchanged.
+	 */
+	onPayload?: (payload: unknown, model: Model<Api>) => unknown | undefined | Promise<unknown | undefined>;
+	/**
+	 * Optional callback invoked after an HTTP response is received and before
+	 * its body stream is consumed.
+	 */
+	onResponse?: (response: ProviderResponse, model: Model<Api>) => void | Promise<void>;
+	/**
+	 * Optional custom HTTP headers to include in API requests.
+	 * Merged with provider defaults; can override default headers.
+	 * Not supported by all providers (e.g., AWS Bedrock uses SDK auth).
+	 */
+	headers?: Record<string, string>;
+	/**
+	 * Maximum delay in milliseconds to wait for a retry when the server requests a long wait.
+	 * If the server's requested delay exceeds this value, the request fails immediately
+	 * with an error containing the requested delay, allowing higher-level retry logic
+	 * to handle it with user visibility.
+	 * Default: 60000 (60 seconds). Set to 0 to disable the cap.
+	 */
+	maxRetryDelayMs?: number;
+	/**
+	 * Optional metadata to include in API requests.
+	 * Providers extract the fields they understand and ignore the rest.
+	 * For example, Anthropic uses `user_id` for abuse tracking and rate limiting.
+	 */
+	metadata?: Record<string, unknown>;
+}
+
+export type ProviderStreamOptions = StreamOptions & Record<string, unknown>;
+
+// Unified options with reasoning passed to streamSimple() and completeSimple()
+export interface SimpleStreamOptions extends StreamOptions {
+	reasoning?: ThinkingLevel;
+	/** Custom token budgets for thinking levels (token-based providers only) */
+	thinkingBudgets?: ThinkingBudgets;
+}
+
+// Generic StreamFunction with typed options.
+//
+// Contract:
+// - Must return an AssistantMessageEventStream.
+// - Once invoked, request/model/runtime failures should be encoded in the
+//   returned stream, not thrown.
+// - Error termination must produce an AssistantMessage with stopReason
+//   "error" or "aborted" and errorMessage, emitted via the stream protocol.
+export type StreamFunction<TApi extends Api = Api, TOptions extends StreamOptions = StreamOptions> = (
+	model: Model<TApi>,
+	context: Context,
+	options?: TOptions,
+) => AssistantMessageEventStream;
+
+export interface TextSignatureV1 {
+	v: 1;
+	id: string;
+	phase?: "commentary" | "final_answer";
+}
+
+export interface TextContent {
+	type: "text";
+	text: string;
+	textSignature?: string; // e.g., for OpenAI responses, message metadata (legacy id string or TextSignatureV1 JSON)
+}
+
+export interface ThinkingContent {
+	type: "thinking";
+	thinking: string;
+	thinkingSignature?: string; // e.g., for OpenAI responses, the reasoning item ID
+	/** When true, the thinking content was redacted by safety filters. The opaque
+	 *  encrypted payload is stored in `thinkingSignature` so it can be passed back
+	 *  to the API for multi-turn continuity. */
+	redacted?: boolean;
+}
+
+export interface ImageContent {
+	type: "image";
+	data: string; // base64 encoded image data
+	mimeType: string; // e.g., "image/jpeg", "image/png"
+}
+
+export interface ToolCall {
+	type: "toolCall";
+	id: string;
+	name: string;
+	arguments: Record<string, any>;
+	thoughtSignature?: string; // Google-specific: opaque signature for reusing thought context
+}
+
+export interface Usage {
+	input: number;
+	output: number;
+	cacheRead: number;
+	cacheWrite: number;
+	totalTokens: number;
+	cost: {
+		input: number;
+		output: number;
+		cacheRead: number;
+		cacheWrite: number;
+		total: number;
+	};
+}
+
+export type StopReason = "stop" | "length" | "toolUse" | "error" | "aborted";
+
+export interface UserMessage {
+	role: "user";
+	content: string | (TextContent | ImageContent)[];
+	timestamp: number; // Unix timestamp in milliseconds
+}
+
+export interface AssistantMessage {
+	role: "assistant";
+	content: (TextContent | ThinkingContent | ToolCall)[];
+	api: Api;
+	provider: Provider;
+	model: string;
+	responseId?: string; // Provider-specific response/message identifier when the upstream API exposes one
+	usage: Usage;
+	stopReason: StopReason;
+	errorMessage?: string;
+	timestamp: number; // Unix timestamp in milliseconds
+}
+
+export interface ToolResultMessage<TDetails = any> {
+	role: "toolResult";
+	toolCallId: string;
+	toolName: string;
+	content: (TextContent | ImageContent)[]; // Supports text and images
+	details?: TDetails;
+	isError: boolean;
+	timestamp: number; // Unix timestamp in milliseconds
+}
+
+export type Message = UserMessage | AssistantMessage | ToolResultMessage;
+
+import type { TSchema } from "typebox";
+
+export interface Tool<TParameters extends TSchema = TSchema> {
+	name: string;
+	description: string;
+	parameters: TParameters;
+}
+
+export interface Context {
+	systemPrompt?: string;
+	messages: Message[];
+	tools?: Tool[];
+}
+
+/**
+ * Event protocol for AssistantMessageEventStream.
+ *
+ * Streams should emit `start` before partial updates, then terminate with either:
+ * - `done` carrying the final successful AssistantMessage, or
+ * - `error` carrying the final AssistantMessage with stopReason "error" or "aborted"
+ *   and errorMessage.
+ */
+export type AssistantMessageEvent =
+	| { type: "start"; partial: AssistantMessage }
+	| { type: "text_start"; contentIndex: number; partial: AssistantMessage }
+	| { type: "text_delta"; contentIndex: number; delta: string; partial: AssistantMessage }
+	| { type: "text_end"; contentIndex: number; content: string; partial: AssistantMessage }
+	| { type: "thinking_start"; contentIndex: number; partial: AssistantMessage }
+	| { type: "thinking_delta"; contentIndex: number; delta: string; partial: AssistantMessage }
+	| { type: "thinking_end"; contentIndex: number; content: string; partial: AssistantMessage }
+	| { type: "toolcall_start"; contentIndex: number; partial: AssistantMessage }
+	| { type: "toolcall_delta"; contentIndex: number; delta: string; partial: AssistantMessage }
+	| { type: "toolcall_end"; contentIndex: number; toolCall: ToolCall; partial: AssistantMessage }
+	| { type: "done"; reason: Extract<StopReason, "stop" | "length" | "toolUse">; message: AssistantMessage }
+	| { type: "error"; reason: Extract<StopReason, "aborted" | "error">; error: AssistantMessage };
+
+/**
+ * Compatibility settings for OpenAI-compatible completions APIs.
+ * Use this to override URL-based auto-detection for custom providers.
+ */
+export interface OpenAICompletionsCompat {
+	/** Whether the provider supports the `store` field. Default: auto-detected from URL. */
+	supportsStore?: boolean;
+	/** Whether the provider supports the `developer` role (vs `system`). Default: auto-detected from URL. */
+	supportsDeveloperRole?: boolean;
+	/** Whether the provider supports `reasoning_effort`. Default: auto-detected from URL. */
+	supportsReasoningEffort?: boolean;
+	/** Optional mapping from pi-ai reasoning levels to provider/model-specific `reasoning_effort` values. */
+	reasoningEffortMap?: Partial<Record<ThinkingLevel, string>>;
+	/** Whether the provider supports `stream_options: { include_usage: true }` for token usage in streaming responses. Default: true. */
+	supportsUsageInStreaming?: boolean;
+	/** Which field to use for max tokens. Default: auto-detected from URL. */
+	maxTokensField?: "max_completion_tokens" | "max_tokens";
+	/** Whether tool results require the `name` field. Default: auto-detected from URL. */
+	requiresToolResultName?: boolean;
+	/** Whether a user message after tool results requires an assistant message in between. Default: auto-detected from URL. */
+	requiresAssistantAfterToolResult?: boolean;
+	/** Whether thinking blocks must be converted to text blocks with <thinking> delimiters. Default: auto-detected from URL. */
+	requiresThinkingAsText?: boolean;
+	/** Format for reasoning/thinking parameter. "openai" uses reasoning_effort, "openrouter" uses reasoning: { effort }, "zai" uses top-level enable_thinking: boolean, "qwen" uses top-level enable_thinking: boolean, and "qwen-chat-template" uses chat_template_kwargs.enable_thinking. Default: "openai". */
+	thinkingFormat?: "openai" | "openrouter" | "zai" | "qwen" | "qwen-chat-template";
+	/** OpenRouter-specific routing preferences. Only used when baseUrl points to OpenRouter. */
+	openRouterRouting?: OpenRouterRouting;
+	/** Vercel AI Gateway routing preferences. Only used when baseUrl points to Vercel AI Gateway. */
+	vercelGatewayRouting?: VercelGatewayRouting;
+	/** Whether z.ai supports top-level `tool_stream: true` for streaming tool call deltas. Default: false. */
+	zaiToolStream?: boolean;
+	/** Whether the provider supports the `strict` field in tool definitions. Default: true. */
+	supportsStrictMode?: boolean;
+	/** Cache control convention for prompt caching. "anthropic" applies Anthropic-style `cache_control` markers to the system prompt, last tool definition, and last user/assistant text content. */
+	cacheControlFormat?: "anthropic";
+	/** Whether to send known session-affinity headers (`session_id`, `x-client-request-id`, `x-session-affinity`) from `options.sessionId` when caching is enabled. Default: false. */
+	sendSessionAffinityHeaders?: boolean;
+}
+
+/** Compatibility settings for OpenAI Responses APIs. */
+export interface OpenAIResponsesCompat {
+	// Reserved for future use
+}
+
+/**
+ * OpenRouter provider routing preferences.
+ * Controls which upstream providers OpenRouter routes requests to.
+ * Sent as the `provider` field in the OpenRouter API request body.
+ * @see https://openrouter.ai/docs/guides/routing/provider-selection
+ */
+export interface OpenRouterRouting {
+	/** Whether to allow backup providers to serve requests. Default: true. */
+	allow_fallbacks?: boolean;
+	/** Whether to filter providers to only those that support all parameters in the request. Default: false. */
+	require_parameters?: boolean;
+	/** Data collection setting. "allow" (default): allow providers that may store/train on data. "deny": only use providers that don't collect user data. */
+	data_collection?: "deny" | "allow";
+	/** Whether to restrict routing to only ZDR (Zero Data Retention) endpoints. */
+	zdr?: boolean;
+	/** Whether to restrict routing to only models that allow text distillation. */
+	enforce_distillable_text?: boolean;
+	/** An ordered list of provider names/slugs to try in sequence, falling back to the next if unavailable. */
+	order?: string[];
+	/** List of provider names/slugs to exclusively allow for this request. */
+	only?: string[];
+	/** List of provider names/slugs to skip for this request. */
+	ignore?: string[];
+	/** A list of quantization levels to filter providers by (e.g., ["fp16", "bf16", "fp8", "fp6", "int8", "int4", "fp4", "fp32"]). */
+	quantizations?: string[];
+	/** Sorting strategy. Can be a string (e.g., "price", "throughput", "latency") or an object with `by` and `partition`. */
+	sort?:
+		| string
+		| {
+				/** The sorting metric: "price", "throughput", "latency". */
+				by?: string;
+				/** Partitioning strategy: "model" (default) or "none". */
+				partition?: string | null;
+		  };
+	/** Maximum price per million tokens (USD). */
+	max_price?: {
+		/** Price per million prompt tokens. */
+		prompt?: number | string;
+		/** Price per million completion tokens. */
+		completion?: number | string;
+		/** Price per image. */
+		image?: number | string;
+		/** Price per audio unit. */
+		audio?: number | string;
+		/** Price per request. */
+		request?: number | string;
+	};
+	/** Preferred minimum throughput (tokens/second). Can be a number (applies to p50) or an object with percentile-specific cutoffs. */
+	preferred_min_throughput?:
+		| number
+		| {
+				/** Minimum tokens/second at the 50th percentile. */
+				p50?: number;
+				/** Minimum tokens/second at the 75th percentile. */
+				p75?: number;
+				/** Minimum tokens/second at the 90th percentile. */
+				p90?: number;
+				/** Minimum tokens/second at the 99th percentile. */
+				p99?: number;
+		  };
+	/** Preferred maximum latency (seconds). Can be a number (applies to p50) or an object with percentile-specific cutoffs. */
+	preferred_max_latency?:
+		| number
+		| {
+				/** Maximum latency in seconds at the 50th percentile. */
+				p50?: number;
+				/** Maximum latency in seconds at the 75th percentile. */
+				p75?: number;
+				/** Maximum latency in seconds at the 90th percentile. */
+				p90?: number;
+				/** Maximum latency in seconds at the 99th percentile. */
+				p99?: number;
+		  };
+}
+
+/**
+ * Vercel AI Gateway routing preferences.
+ * Controls which upstream providers the gateway routes requests to.
+ * @see https://vercel.com/docs/ai-gateway/models-and-providers/provider-options
+ */
+export interface VercelGatewayRouting {
+	/** List of provider slugs to exclusively use for this request (e.g., ["bedrock", "anthropic"]). */
+	only?: string[];
+	/** List of provider slugs to try in order (e.g., ["anthropic", "openai"]). */
+	order?: string[];
+}
+
+// Model interface for the unified model system
+export interface Model<TApi extends Api> {
+	id: string;
+	name: string;
+	api: TApi;
+	provider: Provider;
+	baseUrl: string;
+	reasoning: boolean;
+	input: ("text" | "image")[];
+	cost: {
+		input: number; // $/million tokens
+		output: number; // $/million tokens
+		cacheRead: number; // $/million tokens
+		cacheWrite: number; // $/million tokens
+	};
+	contextWindow: number;
+	maxTokens: number;
+	headers?: Record<string, string>;
+	/** Compatibility overrides for OpenAI-compatible APIs. If not set, auto-detected from baseUrl. */
+	compat?: TApi extends "openai-completions"
+		? OpenAICompletionsCompat
+		: TApi extends "openai-responses"
+			? OpenAIResponsesCompat
+			: never;
+}

package/pi.dev-src/pi-mono/ai/src/utils/event-stream.ts

@@ -0,0 +1,87 @@
+import type { AssistantMessage, AssistantMessageEvent } from "../types.js";
+
+// Generic event stream class for async iteration
+export class EventStream<T, R = T> implements AsyncIterable<T> {
+	private queue: T[] = [];
+	private waiting: ((value: IteratorResult<T>) => void)[] = [];
+	private done = false;
+	private finalResultPromise: Promise<R>;
+	private resolveFinalResult!: (result: R) => void;
+
+	constructor(
+		private isComplete: (event: T) => boolean,
+		private extractResult: (event: T) => R,
+	) {
+		this.finalResultPromise = new Promise((resolve) => {
+			this.resolveFinalResult = resolve;
+		});
+	}
+
+	push(event: T): void {
+		if (this.done) return;
+
+		if (this.isComplete(event)) {
+			this.done = true;
+			this.resolveFinalResult(this.extractResult(event));
+		}
+
+		// Deliver to waiting consumer or queue it
+		const waiter = this.waiting.shift();
+		if (waiter) {
+			waiter({ value: event, done: false });
+		} else {
+			this.queue.push(event);
+		}
+	}
+
+	end(result?: R): void {
+		this.done = true;
+		if (result !== undefined) {
+			this.resolveFinalResult(result);
+		}
+		// Notify all waiting consumers that we're done
+		while (this.waiting.length > 0) {
+			const waiter = this.waiting.shift()!;
+			waiter({ value: undefined as any, done: true });
+		}
+	}
+
+	async *[Symbol.asyncIterator](): AsyncIterator<T> {
+		while (true) {
+			if (this.queue.length > 0) {
+				yield this.queue.shift()!;
+			} else if (this.done) {
+				return;
+			} else {
+				const result = await new Promise<IteratorResult<T>>((resolve) => this.waiting.push(resolve));
+				if (result.done) return;
+				yield result.value;
+			}
+		}
+	}
+
+	result(): Promise<R> {
+		return this.finalResultPromise;
+	}
+}
+
+export class AssistantMessageEventStream extends EventStream<AssistantMessageEvent, AssistantMessage> {
+	constructor() {
+		super(
+			(event) => event.type === "done" || event.type === "error",
+			(event) => {
+				if (event.type === "done") {
+					return event.message;
+				} else if (event.type === "error") {
+					return event.error;
+				}
+				throw new Error("Unexpected event type for final result");
+			},
+		);
+	}
+}
+
+/** Factory function for AssistantMessageEventStream (for use in extensions) */
+export function createAssistantMessageEventStream(): AssistantMessageEventStream {
+	return new AssistantMessageEventStream();
+}

package/pi.dev-src/pi-mono/ai/src/utils/hash.ts

@@ -0,0 +1,13 @@
+/** Fast deterministic hash to shorten long strings */
+export function shortHash(str: string): string {
+	let h1 = 0xdeadbeef;
+	let h2 = 0x41c6ce57;
+	for (let i = 0; i < str.length; i++) {
+		const ch = str.charCodeAt(i);
+		h1 = Math.imul(h1 ^ ch, 2654435761);
+		h2 = Math.imul(h2 ^ ch, 1597334677);
+	}
+	h1 = Math.imul(h1 ^ (h1 >>> 16), 2246822507) ^ Math.imul(h2 ^ (h2 >>> 13), 3266489909);
+	h2 = Math.imul(h2 ^ (h2 >>> 16), 2246822507) ^ Math.imul(h1 ^ (h1 >>> 13), 3266489909);
+	return (h2 >>> 0).toString(36) + (h1 >>> 0).toString(36);
+}

package/pi.dev-src/pi-mono/ai/src/utils/headers.ts

@@ -0,0 +1,7 @@
+export function headersToRecord(headers: Headers): Record<string, string> {
+	const result: Record<string, string> = {};
+	for (const [key, value] of headers.entries()) {
+		result[key] = value;
+	}
+	return result;
+}

package/pi.dev-src/pi-mono/ai/src/utils/json-parse.ts

@@ -0,0 +1,124 @@
+import { parse as partialParse } from "partial-json";
+
+const VALID_JSON_ESCAPES = new Set(['"', "\\", "/", "b", "f", "n", "r", "t", "u"]);
+
+function isControlCharacter(char: string): boolean {
+	const codePoint = char.codePointAt(0);
+	return codePoint !== undefined && codePoint >= 0x00 && codePoint <= 0x1f;
+}
+
+function escapeControlCharacter(char: string): string {
+	switch (char) {
+		case "\b":
+			return "\\b";
+		case "\f":
+			return "\\f";
+		case "\n":
+			return "\\n";
+		case "\r":
+			return "\\r";
+		case "\t":
+			return "\\t";
+		default:
+			return `\\u${char.codePointAt(0)?.toString(16).padStart(4, "0") ?? "0000"}`;
+	}
+}
+
+/**
+ * Repairs malformed JSON string literals by:
+ * - escaping raw control characters inside strings
+ * - doubling backslashes before invalid escape characters
+ */
+export function repairJson(json: string): string {
+	let repaired = "";
+	let inString = false;
+
+	for (let index = 0; index < json.length; index++) {
+		const char = json[index];
+
+		if (!inString) {
+			repaired += char;
+			if (char === '"') {
+				inString = true;
+			}
+			continue;
+		}
+
+		if (char === '"') {
+			repaired += char;
+			inString = false;
+			continue;
+		}
+
+		if (char === "\\") {
+			const nextChar = json[index + 1];
+			if (nextChar === undefined) {
+				repaired += "\\\\";
+				continue;
+			}
+
+			if (nextChar === "u") {
+				const unicodeDigits = json.slice(index + 2, index + 6);
+				if (/^[0-9a-fA-F]{4}$/.test(unicodeDigits)) {
+					repaired += `\\u${unicodeDigits}`;
+					index += 5;
+					continue;
+				}
+			}
+
+			if (VALID_JSON_ESCAPES.has(nextChar)) {
+				repaired += `\\${nextChar}`;
+				index += 1;
+				continue;
+			}
+
+			repaired += "\\\\";
+			continue;
+		}
+
+		repaired += isControlCharacter(char) ? escapeControlCharacter(char) : char;
+	}
+
+	return repaired;
+}
+
+export function parseJsonWithRepair<T>(json: string): T {
+	try {
+		return JSON.parse(json) as T;
+	} catch (error) {
+		const repairedJson = repairJson(json);
+		if (repairedJson !== json) {
+			return JSON.parse(repairedJson) as T;
+		}
+		throw error;
+	}
+}
+
+/**
+ * Attempts to parse potentially incomplete JSON during streaming.
+ * Always returns a valid object, even if the JSON is incomplete.
+ *
+ * @param partialJson The partial JSON string from streaming
+ * @returns Parsed object or empty object if parsing fails
+ */
+export function parseStreamingJson<T = Record<string, unknown>>(partialJson: string | undefined): T {
+	if (!partialJson || partialJson.trim() === "") {
+		return {} as T;
+	}
+
+	try {
+		return parseJsonWithRepair<T>(partialJson);
+	} catch {
+		try {
+			const result = partialParse(partialJson);
+			return (result ?? {}) as T;
+		} catch {
+			try {
+				const result = partialParse(repairJson(partialJson));
+				return (result ?? {}) as T;
+			} catch {
+				return {} as T;
+			}
+		}
+	}
+}