@xiaozhiclaw/provider-core 0.1.0

package/dist/transports/openai-media.d.ts

@@ -0,0 +1,24 @@
+/**
+ * OpenAI Media Transport 鈥?Images API (gpt-image-2) + Audio Speech API (tts-1).
+ *
+ * Image:  POST /v1/images/generations (sync response, returns URLs)
+ * TTS:    POST /v1/audio/speech      (sync response, returns raw audio bytes)
+ * Auth: Authorization: Bearer $OPENAI_API_KEY
+ * Docs: https://platform.openai.com/docs/api-reference/images/create
+ *       https://platform.openai.com/docs/api-reference/audio/createSpeech
+ */
+import type { MediaTransport, MediaRequest, MediaResult, MediaType } from "../media-transport.js";
+export interface OpenAIMediaConfig {
+    /** Base URL, e.g. "https://api.openai.com" */
+    baseUrl: string;
+    timeoutMs?: number;
+}
+export declare class OpenAIMediaTransport implements MediaTransport {
+    readonly supportedTypes: readonly MediaType[];
+    private baseUrl;
+    private timeoutMs;
+    constructor(config: OpenAIMediaConfig);
+    generate(request: MediaRequest, apiKey: string, signal?: AbortSignal): Promise<MediaResult>;
+    private generateImage;
+    private generateTTS;
+}

package/dist/transports/openai-media.js

@@ -0,0 +1,118 @@
+/**
+ * OpenAI Media Transport 鈥?Images API (gpt-image-2) + Audio Speech API (tts-1).
+ *
+ * Image:  POST /v1/images/generations (sync response, returns URLs)
+ * TTS:    POST /v1/audio/speech      (sync response, returns raw audio bytes)
+ * Auth: Authorization: Bearer $OPENAI_API_KEY
+ * Docs: https://platform.openai.com/docs/api-reference/images/create
+ *       https://platform.openai.com/docs/api-reference/audio/createSpeech
+ */
+import { writeFileSync, mkdirSync } from "node:fs";
+import { join } from "node:path";
+import { randomUUID } from "node:crypto";
+import { getUserCacheDir } from "../paths.js";
+const DEFAULT_TIMEOUT_MS = 180_000;
+export class OpenAIMediaTransport {
+    supportedTypes = ["image", "tts"];
+    baseUrl;
+    timeoutMs;
+    constructor(config) {
+        this.baseUrl = config.baseUrl.replace(/\/+$/, "");
+        this.timeoutMs = config.timeoutMs ?? DEFAULT_TIMEOUT_MS;
+    }
+    async generate(request, apiKey, signal) {
+        switch (request.mediaType) {
+            case "image":
+                return this.generateImage(request, apiKey, signal);
+            case "tts":
+                return this.generateTTS(request, apiKey, signal);
+            default:
+                throw new Error(`OpenAIMediaTransport: unsupported mediaType "${request.mediaType}"`);
+        }
+    }
+    // 鈹€鈹€ Image (gpt-image-2) 鈥?sync 鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€
+    async generateImage(request, apiKey, signal) {
+        const start = Date.now();
+        const url = `${this.baseUrl}/v1/images/generations`;
+        const body = {
+            model: request.model,
+            prompt: request.prompt,
+            n: request.n ?? 1,
+            size: request.size ?? "1024x1024",
+            response_format: "url",
+        };
+        if (request.quality)
+            body.quality = request.quality;
+        if (request.style)
+            body.style = request.style;
+        if (request.background)
+            body.background = request.background;
+        const res = await fetch(url, {
+            method: "POST",
+            headers: {
+                "Content-Type": "application/json",
+                Authorization: `Bearer ${apiKey}`,
+            },
+            body: JSON.stringify(body),
+            signal: signal ?? AbortSignal.timeout(this.timeoutMs),
+        });
+        if (!res.ok) {
+            const text = await res.text().catch(() => "");
+            throw new Error(`OpenAI images API error ${res.status}: ${text}`);
+        }
+        const data = await res.json();
+        const mediaUrls = (data.data ?? [])
+            .map(d => d.url)
+            .filter((u) => !!u);
+        return {
+            mediaUrls,
+            model: request.model,
+            size: request.size ?? "1024x1024",
+            durationMs: Date.now() - start,
+        };
+    }
+    // 鈹€鈹€ TTS (tts-1 / tts-1-hd) 鈥?sync 鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€鈹€
+    async generateTTS(request, apiKey, signal) {
+        const start = Date.now();
+        const text = request.text || request.prompt;
+        if (!text) {
+            throw new Error("OpenAIMediaTransport: text or prompt is required for TTS");
+        }
+        const url = `${this.baseUrl}/v1/audio/speech`;
+        const body = {
+            model: request.model || "tts-1",
+            input: text,
+            voice: request.voice ?? "alloy",
+            response_format: request.audioFormat ?? "mp3",
+        };
+        if (request.speed !== undefined)
+            body.speed = request.speed;
+        const res = await fetch(url, {
+            method: "POST",
+            headers: {
+                "Content-Type": "application/json",
+                Authorization: `Bearer ${apiKey}`,
+            },
+            body: JSON.stringify(body),
+            signal: signal ?? AbortSignal.timeout(this.timeoutMs),
+        });
+        if (!res.ok) {
+            const errText = await res.text().catch(() => "");
+            throw new Error(`OpenAI TTS API error ${res.status}: ${errText}`);
+        }
+        // Response is raw audio bytes 鈥?save to temp file
+        const audioBuffer = Buffer.from(await res.arrayBuffer());
+        const cacheDir = join(getUserCacheDir(), "tts");
+        mkdirSync(cacheDir, { recursive: true });
+        const filename = `openai-tts-${randomUUID()}.mp3`;
+        const filePath = join(cacheDir, filename);
+        writeFileSync(filePath, audioBuffer);
+        return {
+            mediaUrls: [`file://${filePath}`],
+            model: request.model || "tts-1",
+            durationMs: Date.now() - start,
+            billingUnit: "per_character",
+            billingQuantity: text.length,
+        };
+    }
+}

package/dist/transports/openai-responses.d.ts

@@ -0,0 +1,63 @@
+/**
+ * OpenAI Responses API TransportSSE streaming implementation.
+ *
+ * Implements the OpenAI Responses API (`POST /v1/responses`),
+ * the officially recommended path for GPT-5.x text generation.
+ *
+ * Key differences from OpenAI Chat Completions:
+ *   - Endpoint: POST {baseUrl}/v1/responses
+ *   - Request body uses `input` (not `messages`), `instructions`, `reasoning`
+ *   - SSE events: response.output_text.delta, response.function_call_arguments.delta,
+ *     response.completed, etc.
+ *   - Tool defs: { type: "function", name, parameters } (not nested under `function:`)
+ *   - Tool results: { type: "function_call_output", call_id, output }
+ *   - Context persistence: previous_response_id for server-side session continuation
+ *   - Structured output: `text: { format: { type: "json_schema", ... } }`
+ *   - Reasoning: `reasoning: { effort, summary }` for GPT-5.x models
+ *
+ * Wire format reference:
+ *   https://developers.openai.com/api/docs/api-reference/responses/create
+ *   https://developers.openai.com/api/docs/api-reference/responses/streaming-events
+ *
+ * Design: Closely mirrors volcengine-responses.ts patterns while adapting to
+ * OpenAI-specific wire format. Shared LLMChunk output makes upper layers
+ * transport-agnostic.
+ */
+import type { LLMChunk, LLMRequest, LLMTransport } from "../transport.js";
+import type { ProviderQuirks } from "../provider-def.js";
+import type { FileUploadAdapter } from "../file-upload-service.js";
+export interface OpenAIResponsesTransportConfig {
+    baseUrl: string;
+    extraHeaders?: Record<string, string>;
+    timeoutMs?: number;
+    quirks?: ProviderQuirks;
+    fileUploadAdapter?: FileUploadAdapter;
+}
+export declare class OpenAIResponsesTransport implements LLMTransport {
+    private baseUrl;
+    private extraHeaders;
+    private timeoutMs;
+    private quirks;
+    private fileUploadAdapter?;
+    constructor(config: OpenAIResponsesTransportConfig);
+    stream(request: LLMRequest, apiKey: string, signal?: AbortSignal): AsyncGenerator<LLMChunk>;
+    private buildRequestBody;
+    private fetchAndStream;
+    private handleNonStreamingResponse;
+    /**
+     * Parse OpenAI Responses API SSE stream.
+     *
+     * Event format: "event: <type>\ndata: <json>\n\n"
+     * Key events:
+     *   - response.output_text.delta           閳?text content delta
+     *   - response.reasoning_summary_text.delta 閳?reasoning summary text
+     *   - response.function_call_arguments.delta 閳?tool call arguments streaming
+     *   - response.output_item.added           閳?new output item started
+     *   - response.output_item.done            閳?output item completed
+     *   - response.content_part.done           閳?content part completed (annotations)
+     *   - response.completed                   閳?full response complete with usage
+     *   - response.failed                      閳?error
+     */
+    private parseSSEStream;
+    private processEvent;
+}