@yourgpt/llm-sdk 2.6.0-beta.1 → 2.6.0-beta.3

package/dist/providers/ollama/index.js

@@ -1,5 +1,90 @@
 'use strict';
 
+// src/core/tool-output.ts
+var imageTypes = /* @__PURE__ */ new Set([
+  "image/png",
+  "image/jpeg",
+  "image/webp",
+  "image/gif"
+]);
+function normalizeToolOutput(value) {
+  const output = value;
+  if (output?.type !== "content" || !Array.isArray(output.content) || !output.content.length) {
+    throw new Error("Tool model output must contain a nonempty content array");
+  }
+  return {
+    type: "content",
+    content: output.content.map((part) => {
+      if (part?.type === "text" && typeof part.text === "string")
+        return { type: "text", text: part.text };
+      if (part?.type !== "image")
+        throw new Error("Unsupported tool output content type");
+      const sources = ["data", "url", "fileId"].filter((key) => key in part);
+      if (sources.length !== 1)
+        throw new Error(
+          "Tool image requires exactly one of data, url, or fileId"
+        );
+      if ("fileId" in part) {
+        if (part.provider !== "openai" || typeof part.fileId !== "string" || !part.fileId.trim())
+          throw new Error(
+            "Tool image file references must name the openai provider and fileId"
+          );
+        return { type: "image", fileId: part.fileId, provider: "openai" };
+      }
+      if ("url" in part) {
+        let url;
+        try {
+          url = new URL(part.url);
+        } catch {
+          throw new Error("Invalid tool image URL");
+        }
+        if (!["https:", "http:"].includes(url.protocol) || url.username || url.password)
+          throw new Error(
+            "Tool image URL must use HTTP(S) without credentials"
+          );
+        return { type: "image", url: url.href };
+      }
+      if (!("data" in part) || !imageTypes.has(part.mediaType))
+        throw new Error("Unsupported tool image media type");
+      let data = part.data instanceof Uint8Array ? Buffer.from(part.data).toString("base64") : part.data;
+      if (typeof data !== "string")
+        throw new Error("Tool image data must be base64 or bytes");
+      if (data.startsWith("data:")) {
+        const match = /^data:([^;,]+);base64,(.+)$/.exec(data);
+        if (!match || match[1] !== part.mediaType)
+          throw new Error(
+            "Tool image data URL media type does not match mediaType"
+          );
+        data = match[2];
+      }
+      if (!data.length || data.length % 4 !== 0 || !/^[A-Za-z0-9+/]+={0,2}$/.test(data))
+        throw new Error("Invalid tool image base64 data");
+      if (Buffer.byteLength(data, "base64") > 20 * 1024 * 1024)
+        throw new Error("Tool image exceeds 20 MiB");
+      if (Buffer.from(data, "base64").toString("base64") !== data)
+        throw new Error("Invalid tool image base64 data");
+      return { type: "image", data, mediaType: part.mediaType };
+    })
+  };
+}
+function toolOutputForProvider(value, provider) {
+  const output = normalizeToolOutput(value);
+  return output.content.map((part) => {
+    if (part.type === "text")
+      return {
+        type: "text",
+        text: part.text
+      };
+    throw new Error(
+        "Image tool results require OpenAI Responses or Anthropic; this endpoint is not supported"
+      );
+  });
+}
+function textToolOutput(value, fallback) {
+  if (value === void 0) return fallback;
+  return toolOutputForProvider(value).map((part) => part.text).join("\n");
+}
+
 // src/core/utils.ts
 function generateId(prefix = "id") {
   return `${prefix}_${Date.now()}_${Math.random().toString(36).substring(2, 9)}`;
@@ -150,7 +235,7 @@ async function formatMessagesForOllama(messages, systemPrompt) {
   for (const msg of messages) {
     const baseMessage = {
       role: msg.role,
-      content: msg.content ?? ""
+      content: msg.role === "tool" ? textToolOutput(msg.modelOutput, msg.content ?? "") : msg.content ?? ""
     };
     if (msg.role === "user" && msg.metadata?.attachments) {
       const images = await extractImagesFromAttachments(
@@ -187,7 +272,7 @@ async function processRawMessagesForOllama(rawMessages, systemPrompt) {
   for (const msg of rawMessages) {
     const baseMessage = {
       role: msg.role,
-      content: msg.content ?? ""
+      content: msg.role === "tool" ? textToolOutput(msg.modelOutput, String(msg.content ?? "")) : msg.content ?? ""
     };
     if (msg.role === "user" && msg.attachments && Array.isArray(msg.attachments)) {
       const attachments = msg.attachments;

package/dist/providers/ollama/index.mjs

@@ -1,3 +1,88 @@
+// src/core/tool-output.ts
+var imageTypes = /* @__PURE__ */ new Set([
+  "image/png",
+  "image/jpeg",
+  "image/webp",
+  "image/gif"
+]);
+function normalizeToolOutput(value) {
+  const output = value;
+  if (output?.type !== "content" || !Array.isArray(output.content) || !output.content.length) {
+    throw new Error("Tool model output must contain a nonempty content array");
+  }
+  return {
+    type: "content",
+    content: output.content.map((part) => {
+      if (part?.type === "text" && typeof part.text === "string")
+        return { type: "text", text: part.text };
+      if (part?.type !== "image")
+        throw new Error("Unsupported tool output content type");
+      const sources = ["data", "url", "fileId"].filter((key) => key in part);
+      if (sources.length !== 1)
+        throw new Error(
+          "Tool image requires exactly one of data, url, or fileId"
+        );
+      if ("fileId" in part) {
+        if (part.provider !== "openai" || typeof part.fileId !== "string" || !part.fileId.trim())
+          throw new Error(
+            "Tool image file references must name the openai provider and fileId"
+          );
+        return { type: "image", fileId: part.fileId, provider: "openai" };
+      }
+      if ("url" in part) {
+        let url;
+        try {
+          url = new URL(part.url);
+        } catch {
+          throw new Error("Invalid tool image URL");
+        }
+        if (!["https:", "http:"].includes(url.protocol) || url.username || url.password)
+          throw new Error(
+            "Tool image URL must use HTTP(S) without credentials"
+          );
+        return { type: "image", url: url.href };
+      }
+      if (!("data" in part) || !imageTypes.has(part.mediaType))
+        throw new Error("Unsupported tool image media type");
+      let data = part.data instanceof Uint8Array ? Buffer.from(part.data).toString("base64") : part.data;
+      if (typeof data !== "string")
+        throw new Error("Tool image data must be base64 or bytes");
+      if (data.startsWith("data:")) {
+        const match = /^data:([^;,]+);base64,(.+)$/.exec(data);
+        if (!match || match[1] !== part.mediaType)
+          throw new Error(
+            "Tool image data URL media type does not match mediaType"
+          );
+        data = match[2];
+      }
+      if (!data.length || data.length % 4 !== 0 || !/^[A-Za-z0-9+/]+={0,2}$/.test(data))
+        throw new Error("Invalid tool image base64 data");
+      if (Buffer.byteLength(data, "base64") > 20 * 1024 * 1024)
+        throw new Error("Tool image exceeds 20 MiB");
+      if (Buffer.from(data, "base64").toString("base64") !== data)
+        throw new Error("Invalid tool image base64 data");
+      return { type: "image", data, mediaType: part.mediaType };
+    })
+  };
+}
+function toolOutputForProvider(value, provider) {
+  const output = normalizeToolOutput(value);
+  return output.content.map((part) => {
+    if (part.type === "text")
+      return {
+        type: "text",
+        text: part.text
+      };
+    throw new Error(
+        "Image tool results require OpenAI Responses or Anthropic; this endpoint is not supported"
+      );
+  });
+}
+function textToolOutput(value, fallback) {
+  if (value === void 0) return fallback;
+  return toolOutputForProvider(value).map((part) => part.text).join("\n");
+}
+
 // src/core/utils.ts
 function generateId(prefix = "id") {
   return `${prefix}_${Date.now()}_${Math.random().toString(36).substring(2, 9)}`;
@@ -148,7 +233,7 @@ async function formatMessagesForOllama(messages, systemPrompt) {
   for (const msg of messages) {
     const baseMessage = {
       role: msg.role,
-      content: msg.content ?? ""
+      content: msg.role === "tool" ? textToolOutput(msg.modelOutput, msg.content ?? "") : msg.content ?? ""
     };
     if (msg.role === "user" && msg.metadata?.attachments) {
       const images = await extractImagesFromAttachments(
@@ -185,7 +270,7 @@ async function processRawMessagesForOllama(rawMessages, systemPrompt) {
   for (const msg of rawMessages) {
     const baseMessage = {
       role: msg.role,
-      content: msg.content ?? ""
+      content: msg.role === "tool" ? textToolOutput(msg.modelOutput, String(msg.content ?? "")) : msg.content ?? ""
     };
     if (msg.role === "user" && msg.attachments && Array.isArray(msg.attachments)) {
       const attachments = msg.attachments;

package/dist/providers/openai/index.d.mts

@@ -1,10 +1,11 @@
-import { a7 as OpenAIParameterOptions, L as LanguageModel } from '../../types-BW3csYo5.mjs';
-export { O as OpenAIReasoningEffort } from '../../types-BW3csYo5.mjs';
-import { O as OpenAIProviderConfig, A as AIProvider } from '../../types-NOswlhm9.mjs';
+import { ae as OpenAIParameterOptions, O as OpenAITransportOptions, L as LanguageModel } from '../../types-Dg1Bcg1k.mjs';
+export { h as OpenAIReasoningEffort, ab as OpenAIResponseItem, ac as OpenAIResponseMetadata, aa as OpenAIResponsesOptions, P as ProviderMetadata } from '../../types-Dg1Bcg1k.mjs';
+import { O as OpenAIProviderConfig, A as AIProvider } from '../../types-DHQEMFHA.mjs';
+import { S as SpeechModel, T as TranscriptionModel } from '../../audio-BBSwnjo8.mjs';
 import 'zod';
-import '../../base-CkeT7u-O.mjs';
+import '../../base-BICaQaRS.mjs';
 
-interface OpenAIProviderOptions extends OpenAIParameterOptions {
+interface OpenAIProviderOptions extends OpenAIParameterOptions, OpenAITransportOptions {
     /** API key (defaults to OPENAI_API_KEY env var) */
     apiKey?: string;
     /** Base URL for API (defaults to https://api.openai.com/v1) */
@@ -35,6 +36,24 @@ interface OpenAIProviderOptions extends OpenAIParameterOptions {
  */
 declare function openai(modelId: string, options?: OpenAIProviderOptions): LanguageModel;
 
+interface OpenAIAudioProviderOptions {
+    apiKey?: string;
+    baseUrl?: string;
+    organization?: string;
+    project?: string;
+    /** Custom fetch for proxies and tests. */
+    fetch?: typeof globalThis.fetch;
+}
+interface OpenAITranscriptionOptions {
+    timestampGranularities?: Array<"word" | "segment">;
+    includeLogprobs?: boolean;
+    chunkingStrategy?: "auto";
+}
+/** Server-side, buffered text-to-speech. No microphone or realtime session. */
+declare function openaiSpeech(modelId?: string, config?: OpenAIAudioProviderOptions): SpeechModel;
+/** Recorded-file transcription. Timestamp and diarization options are model-gated. */
+declare function openaiTranscription(modelId?: string, config?: OpenAIAudioProviderOptions): TranscriptionModel;
+
 /**
  * OpenAI Provider
  *
@@ -67,4 +86,4 @@ declare function openai(modelId: string, options?: OpenAIProviderOptions): Langu
 declare function createOpenAI(config?: OpenAIProviderConfig): AIProvider;
 declare const createOpenAIProvider: typeof createOpenAI;
 
-export { OpenAIParameterOptions, type OpenAIProviderOptions, createOpenAI, openai as createOpenAIModel, createOpenAIProvider, openai };
+export { type OpenAIAudioProviderOptions, OpenAIParameterOptions, type OpenAIProviderOptions, type OpenAITranscriptionOptions, OpenAITransportOptions, createOpenAI, openai as createOpenAIModel, createOpenAIProvider, openai, openaiSpeech, openaiTranscription };

package/dist/providers/openai/index.d.ts

@@ -1,10 +1,11 @@
-import { a7 as OpenAIParameterOptions, L as LanguageModel } from '../../types-BW3csYo5.js';
-export { O as OpenAIReasoningEffort } from '../../types-BW3csYo5.js';
-import { O as OpenAIProviderConfig, A as AIProvider } from '../../types-DA54bSmX.js';
+import { ae as OpenAIParameterOptions, O as OpenAITransportOptions, L as LanguageModel } from '../../types-Dg1Bcg1k.js';
+export { h as OpenAIReasoningEffort, ab as OpenAIResponseItem, ac as OpenAIResponseMetadata, aa as OpenAIResponsesOptions, P as ProviderMetadata } from '../../types-Dg1Bcg1k.js';
+import { O as OpenAIProviderConfig, A as AIProvider } from '../../types-Dikj3En4.js';
+import { S as SpeechModel, T as TranscriptionModel } from '../../audio-BBSwnjo8.js';
 import 'zod';
-import '../../base-BdTpUAWI.js';
+import '../../base-CpRPB8u8.js';
 
-interface OpenAIProviderOptions extends OpenAIParameterOptions {
+interface OpenAIProviderOptions extends OpenAIParameterOptions, OpenAITransportOptions {
     /** API key (defaults to OPENAI_API_KEY env var) */
     apiKey?: string;
     /** Base URL for API (defaults to https://api.openai.com/v1) */
@@ -35,6 +36,24 @@ interface OpenAIProviderOptions extends OpenAIParameterOptions {
  */
 declare function openai(modelId: string, options?: OpenAIProviderOptions): LanguageModel;
 
+interface OpenAIAudioProviderOptions {
+    apiKey?: string;
+    baseUrl?: string;
+    organization?: string;
+    project?: string;
+    /** Custom fetch for proxies and tests. */
+    fetch?: typeof globalThis.fetch;
+}
+interface OpenAITranscriptionOptions {
+    timestampGranularities?: Array<"word" | "segment">;
+    includeLogprobs?: boolean;
+    chunkingStrategy?: "auto";
+}
+/** Server-side, buffered text-to-speech. No microphone or realtime session. */
+declare function openaiSpeech(modelId?: string, config?: OpenAIAudioProviderOptions): SpeechModel;
+/** Recorded-file transcription. Timestamp and diarization options are model-gated. */
+declare function openaiTranscription(modelId?: string, config?: OpenAIAudioProviderOptions): TranscriptionModel;
+
 /**
  * OpenAI Provider
  *
@@ -67,4 +86,4 @@ declare function openai(modelId: string, options?: OpenAIProviderOptions): Langu
 declare function createOpenAI(config?: OpenAIProviderConfig): AIProvider;
 declare const createOpenAIProvider: typeof createOpenAI;
 
-export { OpenAIParameterOptions, type OpenAIProviderOptions, createOpenAI, openai as createOpenAIModel, createOpenAIProvider, openai };
+export { type OpenAIAudioProviderOptions, OpenAIParameterOptions, type OpenAIProviderOptions, type OpenAITranscriptionOptions, OpenAITransportOptions, createOpenAI, openai as createOpenAIModel, createOpenAIProvider, openai, openaiSpeech, openaiTranscription };