llm-messages 0.3.0 → 0.4.1

package/dist/index.d.cts

@@ -10,7 +10,7 @@
 /** A supported provider. */
 type Provider = 'openai' | 'anthropic' | 'gemini';
 /** A stable, machine readable code describing a non fatal conversion event. */
-type WarningCode = 'generated-id' | 'unmapped-tool-result' | 'merged-role' | 'dropped-content' | 'invalid-json-arguments' | 'system-midstream' | 'gemini-url-image';
+type WarningCode = 'generated-id' | 'unmapped-tool-result' | 'merged-role' | 'dropped-content' | 'invalid-json-arguments' | 'system-midstream' | 'gemini-url-image' | 'gemini-url-media' | 'unsupported-modality';
 /** A non fatal event raised during conversion. */
 interface Warning {
     code: WarningCode;
@@ -33,8 +33,25 @@ interface OpenAIImagePart {
         detail?: 'auto' | 'low' | 'high' | 'original';
     };
 }
-/** A content part. Unknown part types (audio, files) are preserved verbatim. */
-type OpenAIContentPart = OpenAITextPart | OpenAIImagePart | {
+interface OpenAIAudioPart {
+    type: 'input_audio';
+    /** `data` is raw base64 (no data URL prefix); `format` is `wav` or `mp3`. */
+    input_audio: {
+        data: string;
+        format: string;
+    };
+}
+interface OpenAIFilePart {
+    type: 'file';
+    /** `file_data` is a `data:<mediaType>;base64,<data>` data URL, or use `file_id`. */
+    file: {
+        file_data?: string;
+        file_id?: string;
+        filename?: string;
+    };
+}
+/** A content part. Unknown part types are preserved verbatim. */
+type OpenAIContentPart = OpenAITextPart | OpenAIImagePart | OpenAIAudioPart | OpenAIFilePart | {
     type: string;
     [key: string]: unknown;
 };
@@ -99,7 +116,21 @@ interface AnthropicImageBlock {
         url: string;
     };
 }
-type AnthropicContentBlock = AnthropicTextBlock | AnthropicToolUseBlock | AnthropicToolResultBlock | AnthropicImageBlock | {
+interface AnthropicDocumentBlock {
+    type: 'document';
+    source: {
+        type: 'base64';
+        media_type: string;
+        data: string;
+    } | {
+        type: 'url';
+        url: string;
+    } | {
+        type: 'file';
+        file_id: string;
+    };
+}
+type AnthropicContentBlock = AnthropicTextBlock | AnthropicToolUseBlock | AnthropicToolResultBlock | AnthropicImageBlock | AnthropicDocumentBlock | {
     type: string;
     [key: string]: unknown;
 };
@@ -228,6 +259,28 @@ declare function parseDataUrl(url: string): {
 /** Reassembles a base64 data URL. The inverse of {@link parseDataUrl}. */
 declare function toDataUrl(mediaType: string, data: string): string;
 
+/**
+ * A provider-neutral non-image media part (audio or document). Images keep their
+ * own dedicated path in `image.ts`; everything else flows through here.
+ */
+type MediaModality = 'audio' | 'document';
+type MediaSource = {
+    kind: 'base64';
+    mediaType: string;
+    data: string;
+} | {
+    kind: 'url';
+    url: string;
+} | {
+    kind: 'file_id';
+    id: string;
+};
+interface MediaPart {
+    modality: MediaModality;
+    source: MediaSource;
+    filename?: string;
+}
+
 /** A provider-neutral finish reason. */
 type FinishReason = 'stop' | 'tool_calls' | 'length' | 'content_filter' | 'unknown';
 /** Provider-neutral token usage. */
@@ -252,4 +305,4 @@ declare function normalizeResponse(body: unknown, route: {
     from: Provider;
 }, options?: ConvertOptions): NormalizedResponse;
 
-export { type AnthropicContentBlock, type AnthropicConversation, type AnthropicImageBlock, type AnthropicMessage, type AnthropicTextBlock, type AnthropicToolResultBlock, type AnthropicToolUseBlock, type ConversationOf, type ConvertOptions, type FinishReason, type GeminiContent, type GeminiConversation, type GeminiFileDataPart, type GeminiFunctionCallPart, type GeminiFunctionResponsePart, type GeminiInlineDataPart, type GeminiPart, type GeminiTextPart, type NormalizedImage, type NormalizedResponse, type OpenAIAssistantMessage, type OpenAIContentPart, type OpenAIImagePart, type OpenAIMessage, type OpenAISystemMessage, type OpenAITextPart, type OpenAIToolCall, type OpenAIToolMessage, type OpenAIUserMessage, type Provider, type Usage, type Warning, type WarningCode, convert, fromAnthropic, fromGemini, normalizeResponse, parseDataUrl, responseFromAnthropic, responseFromGemini, responseFromOpenAI, toAnthropic, toDataUrl, toGemini };
+export { type AnthropicContentBlock, type AnthropicConversation, type AnthropicDocumentBlock, type AnthropicImageBlock, type AnthropicMessage, type AnthropicTextBlock, type AnthropicToolResultBlock, type AnthropicToolUseBlock, type ConversationOf, type ConvertOptions, type FinishReason, type GeminiContent, type GeminiConversation, type GeminiFileDataPart, type GeminiFunctionCallPart, type GeminiFunctionResponsePart, type GeminiInlineDataPart, type GeminiPart, type GeminiTextPart, type MediaModality, type MediaPart, type MediaSource, type NormalizedImage, type NormalizedResponse, type OpenAIAssistantMessage, type OpenAIAudioPart, type OpenAIContentPart, type OpenAIFilePart, type OpenAIImagePart, type OpenAIMessage, type OpenAISystemMessage, type OpenAITextPart, type OpenAIToolCall, type OpenAIToolMessage, type OpenAIUserMessage, type Provider, type Usage, type Warning, type WarningCode, convert, fromAnthropic, fromGemini, normalizeResponse, parseDataUrl, responseFromAnthropic, responseFromGemini, responseFromOpenAI, toAnthropic, toDataUrl, toGemini };

package/dist/index.d.ts

@@ -10,7 +10,7 @@
 /** A supported provider. */
 type Provider = 'openai' | 'anthropic' | 'gemini';
 /** A stable, machine readable code describing a non fatal conversion event. */
-type WarningCode = 'generated-id' | 'unmapped-tool-result' | 'merged-role' | 'dropped-content' | 'invalid-json-arguments' | 'system-midstream' | 'gemini-url-image';
+type WarningCode = 'generated-id' | 'unmapped-tool-result' | 'merged-role' | 'dropped-content' | 'invalid-json-arguments' | 'system-midstream' | 'gemini-url-image' | 'gemini-url-media' | 'unsupported-modality';
 /** A non fatal event raised during conversion. */
 interface Warning {
     code: WarningCode;
@@ -33,8 +33,25 @@ interface OpenAIImagePart {
         detail?: 'auto' | 'low' | 'high' | 'original';
     };
 }
-/** A content part. Unknown part types (audio, files) are preserved verbatim. */
-type OpenAIContentPart = OpenAITextPart | OpenAIImagePart | {
+interface OpenAIAudioPart {
+    type: 'input_audio';
+    /** `data` is raw base64 (no data URL prefix); `format` is `wav` or `mp3`. */
+    input_audio: {
+        data: string;
+        format: string;
+    };
+}
+interface OpenAIFilePart {
+    type: 'file';
+    /** `file_data` is a `data:<mediaType>;base64,<data>` data URL, or use `file_id`. */
+    file: {
+        file_data?: string;
+        file_id?: string;
+        filename?: string;
+    };
+}
+/** A content part. Unknown part types are preserved verbatim. */
+type OpenAIContentPart = OpenAITextPart | OpenAIImagePart | OpenAIAudioPart | OpenAIFilePart | {
     type: string;
     [key: string]: unknown;
 };
@@ -99,7 +116,21 @@ interface AnthropicImageBlock {
         url: string;
     };
 }
-type AnthropicContentBlock = AnthropicTextBlock | AnthropicToolUseBlock | AnthropicToolResultBlock | AnthropicImageBlock | {
+interface AnthropicDocumentBlock {
+    type: 'document';
+    source: {
+        type: 'base64';
+        media_type: string;
+        data: string;
+    } | {
+        type: 'url';
+        url: string;
+    } | {
+        type: 'file';
+        file_id: string;
+    };
+}
+type AnthropicContentBlock = AnthropicTextBlock | AnthropicToolUseBlock | AnthropicToolResultBlock | AnthropicImageBlock | AnthropicDocumentBlock | {
     type: string;
     [key: string]: unknown;
 };
@@ -228,6 +259,28 @@ declare function parseDataUrl(url: string): {
 /** Reassembles a base64 data URL. The inverse of {@link parseDataUrl}. */
 declare function toDataUrl(mediaType: string, data: string): string;
 
+/**
+ * A provider-neutral non-image media part (audio or document). Images keep their
+ * own dedicated path in `image.ts`; everything else flows through here.
+ */
+type MediaModality = 'audio' | 'document';
+type MediaSource = {
+    kind: 'base64';
+    mediaType: string;
+    data: string;
+} | {
+    kind: 'url';
+    url: string;
+} | {
+    kind: 'file_id';
+    id: string;
+};
+interface MediaPart {
+    modality: MediaModality;
+    source: MediaSource;
+    filename?: string;
+}
+
 /** A provider-neutral finish reason. */
 type FinishReason = 'stop' | 'tool_calls' | 'length' | 'content_filter' | 'unknown';
 /** Provider-neutral token usage. */
@@ -252,4 +305,4 @@ declare function normalizeResponse(body: unknown, route: {
     from: Provider;
 }, options?: ConvertOptions): NormalizedResponse;
 
-export { type AnthropicContentBlock, type AnthropicConversation, type AnthropicImageBlock, type AnthropicMessage, type AnthropicTextBlock, type AnthropicToolResultBlock, type AnthropicToolUseBlock, type ConversationOf, type ConvertOptions, type FinishReason, type GeminiContent, type GeminiConversation, type GeminiFileDataPart, type GeminiFunctionCallPart, type GeminiFunctionResponsePart, type GeminiInlineDataPart, type GeminiPart, type GeminiTextPart, type NormalizedImage, type NormalizedResponse, type OpenAIAssistantMessage, type OpenAIContentPart, type OpenAIImagePart, type OpenAIMessage, type OpenAISystemMessage, type OpenAITextPart, type OpenAIToolCall, type OpenAIToolMessage, type OpenAIUserMessage, type Provider, type Usage, type Warning, type WarningCode, convert, fromAnthropic, fromGemini, normalizeResponse, parseDataUrl, responseFromAnthropic, responseFromGemini, responseFromOpenAI, toAnthropic, toDataUrl, toGemini };
+export { type AnthropicContentBlock, type AnthropicConversation, type AnthropicDocumentBlock, type AnthropicImageBlock, type AnthropicMessage, type AnthropicTextBlock, type AnthropicToolResultBlock, type AnthropicToolUseBlock, type ConversationOf, type ConvertOptions, type FinishReason, type GeminiContent, type GeminiConversation, type GeminiFileDataPart, type GeminiFunctionCallPart, type GeminiFunctionResponsePart, type GeminiInlineDataPart, type GeminiPart, type GeminiTextPart, type MediaModality, type MediaPart, type MediaSource, type NormalizedImage, type NormalizedResponse, type OpenAIAssistantMessage, type OpenAIAudioPart, type OpenAIContentPart, type OpenAIFilePart, type OpenAIImagePart, type OpenAIMessage, type OpenAISystemMessage, type OpenAITextPart, type OpenAIToolCall, type OpenAIToolMessage, type OpenAIUserMessage, type Provider, type Usage, type Warning, type WarningCode, convert, fromAnthropic, fromGemini, normalizeResponse, parseDataUrl, responseFromAnthropic, responseFromGemini, responseFromOpenAI, toAnthropic, toDataUrl, toGemini };

package/dist/index.js

@@ -92,13 +92,16 @@ function imageToAnthropic(image) {
 function imageFromGemini(part) {
   if (isRecord(part) && isRecord(part.inlineData)) {
     const data = part.inlineData;
-    if (typeof data.mimeType === "string" && typeof data.data === "string") {
+    if (typeof data.mimeType === "string" && typeof data.data === "string" && data.mimeType.startsWith("image/")) {
       return { kind: "base64", mediaType: data.mimeType, data: data.data };
     }
   }
   if (isRecord(part) && isRecord(part.fileData)) {
     const data = part.fileData;
-    if (typeof data.fileUri === "string") return { kind: "url", url: data.fileUri };
+    const mime = typeof data.mimeType === "string" ? data.mimeType : "";
+    if (typeof data.fileUri === "string" && (mime === "" || mime.startsWith("image/"))) {
+      return { kind: "url", url: data.fileUri };
+    }
   }
   return null;
 }
@@ -113,6 +116,124 @@ function imageToGemini(image, reporter) {
   return { fileData: { fileUri: image.url } };
 }
 
+// src/media.ts
+function modalityFromMime(mediaType) {
+  return mediaType.startsWith("audio/") ? "audio" : "document";
+}
+function mediaFromOpenAI(part) {
+  if (!isRecord(part)) return null;
+  if (part.type === "input_audio" && isRecord(part.input_audio)) {
+    const audio = part.input_audio;
+    if (typeof audio.data === "string") {
+      const format = typeof audio.format === "string" ? audio.format : "wav";
+      return { modality: "audio", source: { kind: "base64", mediaType: `audio/${format}`, data: audio.data } };
+    }
+  }
+  if (part.type === "file" && isRecord(part.file)) {
+    const file = part.file;
+    const filename = typeof file.filename === "string" ? file.filename : void 0;
+    if (typeof file.file_data === "string") {
+      const parsed = parseDataUrl(file.file_data);
+      if (parsed) return { modality: "document", source: { kind: "base64", ...parsed }, filename };
+    }
+    if (typeof file.file_id === "string") {
+      return { modality: "document", source: { kind: "file_id", id: file.file_id }, filename };
+    }
+  }
+  return null;
+}
+function mediaToOpenAI(media) {
+  const { modality, source } = media;
+  if (modality === "audio") {
+    if (source.kind !== "base64") return null;
+    const audio = {
+      type: "input_audio",
+      input_audio: { data: source.data, format: source.mediaType.replace(/^audio\//, "") }
+    };
+    return audio;
+  }
+  if (source.kind === "base64") {
+    const file = {
+      type: "file",
+      file: {
+        file_data: toDataUrl(source.mediaType, source.data),
+        ...media.filename ? { filename: media.filename } : {}
+      }
+    };
+    return file;
+  }
+  if (source.kind === "file_id") {
+    const file = {
+      type: "file",
+      file: { file_id: source.id, ...media.filename ? { filename: media.filename } : {} }
+    };
+    return file;
+  }
+  return null;
+}
+function mediaFromAnthropic(block) {
+  if (!isRecord(block) || block.type !== "document" || !isRecord(block.source)) return null;
+  const source = block.source;
+  if (source.type === "base64" && typeof source.media_type === "string" && typeof source.data === "string") {
+    return { modality: "document", source: { kind: "base64", mediaType: source.media_type, data: source.data } };
+  }
+  if (source.type === "url" && typeof source.url === "string") {
+    return { modality: "document", source: { kind: "url", url: source.url } };
+  }
+  if (source.type === "file" && typeof source.file_id === "string") {
+    return { modality: "document", source: { kind: "file_id", id: source.file_id } };
+  }
+  return null;
+}
+function mediaToAnthropic(media, reporter) {
+  if (media.modality === "audio") {
+    reporter.warn("unsupported-modality", "Anthropic has no audio input; dropped an audio part.");
+    return null;
+  }
+  const { source } = media;
+  if (source.kind === "base64") {
+    return { type: "document", source: { type: "base64", media_type: source.mediaType, data: source.data } };
+  }
+  if (source.kind === "url") {
+    return { type: "document", source: { type: "url", url: source.url } };
+  }
+  return { type: "document", source: { type: "file", file_id: source.id } };
+}
+function mediaFromGemini(part) {
+  if (isRecord(part) && isRecord(part.inlineData)) {
+    const data = part.inlineData;
+    if (typeof data.mimeType === "string" && typeof data.data === "string" && !data.mimeType.startsWith("image/")) {
+      return {
+        modality: modalityFromMime(data.mimeType),
+        source: { kind: "base64", mediaType: data.mimeType, data: data.data }
+      };
+    }
+  }
+  if (isRecord(part) && isRecord(part.fileData)) {
+    const data = part.fileData;
+    const mime = typeof data.mimeType === "string" ? data.mimeType : "";
+    if (typeof data.fileUri === "string" && mime !== "" && !mime.startsWith("image/")) {
+      return { modality: modalityFromMime(mime), source: { kind: "url", url: data.fileUri } };
+    }
+  }
+  return null;
+}
+function mediaToGemini(media, reporter) {
+  const { source } = media;
+  if (source.kind === "base64") {
+    return { inlineData: { mimeType: source.mediaType, data: source.data } };
+  }
+  if (source.kind === "url") {
+    reporter.warn(
+      "gemini-url-media",
+      "A media URL was emitted as Gemini fileData.fileUri; Gemini may require the Files API for non-Google URIs."
+    );
+    return { fileData: { fileUri: source.url } };
+  }
+  reporter.warn("unsupported-modality", "Gemini has no file-id media reference; dropped a file_id part.");
+  return null;
+}
+
 // src/providers/openai.ts
 function isSystem(message) {
   return message.role === "system" || message.role === "developer";
@@ -181,6 +302,12 @@ function userContent(content, reporter) {
       blocks.push(imageToAnthropic(image));
       continue;
     }
+    const media = mediaFromOpenAI(part);
+    if (media) {
+      const block = mediaToAnthropic(media, reporter);
+      if (block) blocks.push(block);
+      continue;
+    }
     reporter.warn("dropped-content", "Dropped an unsupported user content part.");
   }
   return blocks;
@@ -222,7 +349,7 @@ function asBlocks(content) {
   return content ? [{ type: "text", text: content }] : [];
 }
 function fromAnthropic(conversation, options = {}) {
-  void options;
+  const reporter = new Reporter(options);
   const out = [];
   if (conversation.system) {
     out.push({ role: "system", content: textOf(conversation.system) });
@@ -240,7 +367,7 @@ function fromAnthropic(conversation, options = {}) {
         });
       }
       if (contentBlocks.length > 0) {
-        out.push({ role: "user", content: userContentToOpenAI(contentBlocks) });
+        out.push({ role: "user", content: userContentToOpenAI(contentBlocks, reporter) });
       }
       continue;
     }
@@ -257,15 +384,24 @@ function fromAnthropic(conversation, options = {}) {
   }
   return out;
 }
-function userContentToOpenAI(blocks) {
-  const hasImage = blocks.some((block) => imageFromAnthropic(block) !== null);
-  if (!hasImage) return textOf(blocks);
+function userContentToOpenAI(blocks, reporter) {
+  const hasMedia = blocks.some((block) => imageFromAnthropic(block) !== null || mediaFromAnthropic(block) !== null);
+  if (!hasMedia) return textOf(blocks);
   const parts = [];
   for (const block of blocks) {
     const image = imageFromAnthropic(block);
     if (image) {
       parts.push(imageToOpenAI(image));
-    } else if (isRecord(block) && block.type === "text" && typeof block.text === "string") {
+      continue;
+    }
+    const media = mediaFromAnthropic(block);
+    if (media) {
+      const part = mediaToOpenAI(media);
+      if (part) parts.push(part);
+      else reporter.warn("dropped-content", "A document URL has no OpenAI Chat Completions equivalent; dropped.");
+      continue;
+    }
+    if (isRecord(block) && block.type === "text" && typeof block.text === "string") {
       parts.push({ type: "text", text: block.text });
     }
   }
@@ -332,6 +468,12 @@ function userParts(content, reporter) {
       parts.push(imageToGemini(image, reporter));
       continue;
     }
+    const media = mediaFromOpenAI(part);
+    if (media) {
+      const geminiPart = mediaToGemini(media, reporter);
+      if (geminiPart) parts.push(geminiPart);
+      continue;
+    }
     reporter.warn("dropped-content", "Dropped an unsupported user content part.");
   }
   return parts.length > 0 ? parts : [{ text: "" }];
@@ -401,7 +543,7 @@ function fromGemini(conversation, options = {}) {
       continue;
     }
     const contentParts = [];
-    let hasImage = false;
+    let hasMedia = false;
     for (const part of parts) {
       if (isRecord(part) && isRecord(part.functionResponse)) {
         const fr = part.functionResponse;
@@ -412,7 +554,16 @@ function fromGemini(conversation, options = {}) {
       const image = imageFromGemini(part);
       if (image) {
         contentParts.push(imageToOpenAI(image));
-        hasImage = true;
+        hasMedia = true;
+        continue;
+      }
+      const media = mediaFromGemini(part);
+      if (media) {
+        const openaiPart = mediaToOpenAI(media);
+        if (openaiPart) {
+          contentParts.push(openaiPart);
+          hasMedia = true;
+        }
         continue;
       }
       if (isRecord(part) && typeof part.text === "string") {
@@ -420,7 +571,7 @@ function fromGemini(conversation, options = {}) {
       }
     }
     if (contentParts.length > 0) {
-      if (hasImage) {
+      if (hasMedia) {
         out.push({ role: "user", content: contentParts });
       } else {
         const text = textOf(contentParts);