Fix gemini message conversion

src/utils/gemini-format.ts

@@ -1,5 +1,15 @@
 import { Anthropic } from "@anthropic-ai/sdk"
-import { Content, EnhancedGenerateContentResponse, FunctionDeclaration, Part, SchemaType } from "@google/generative-ai"
+import {
+	Content,
+	EnhancedGenerateContentResponse,
+	FunctionCallPart,
+	FunctionDeclaration,
+	FunctionResponsePart,
+	InlineDataPart,
+	Part,
+	SchemaType,
+	TextPart,
+} from "@google/generative-ai"
 
 export function convertAnthropicContentToGemini(
 	content:
@@ -12,12 +22,12 @@ export function convertAnthropicContentToGemini(
 		  >
 ): Part[] {
 	if (typeof content === "string") {
-		return [{ text: content }]
+		return [{ text: content } as TextPart]
 	}
-	return content.map((block) => {
+	return content.flatMap((block) => {
 		switch (block.type) {
 			case "text":
-				return { text: block.text }
+				return { text: block.text } as TextPart
 			case "image":
 				if (block.source.type !== "base64") {
 					throw new Error("Unsupported image source type")
@@ -27,22 +37,55 @@ export function convertAnthropicContentToGemini(
 						data: block.source.data,
 						mimeType: block.source.media_type,
 					},
-				}
+				} as InlineDataPart
 			case "tool_use":
 				return {
 					functionCall: {
 						name: block.name,
 						args: block.input,
 					},
-				} as Part
+				} as FunctionCallPart
 			case "tool_result":
-				return {
-					functionResponse: {
-						name: block.tool_use_id,
-						response: {
-							content: block.content,
+				const name = block.tool_use_id.split("-")[0]
+				if (!block.content) {
+					return []
+				}
+				if (typeof block.content === "string") {
+					return {
+						functionResponse: {
+							name,
+							response: {
+								name,
+								content: block.content,
+							},
 						},
-					},
+					} as FunctionResponsePart
+				} else {
+					// The only case when tool_result could be array is when the tool failed and we're providing ie user feedback potentially with images
+					const textParts = block.content.filter((part) => part.type === "text")
+					const imageParts = block.content.filter((part) => part.type === "image")
+					const text = textParts.length > 0 ? textParts.map((part) => part.text).join("\n\n") : ""
+					const imageText = imageParts.length > 0 ? "\n\n(See next part for image)" : ""
+					return [
+						{
+							functionResponse: {
+								name,
+								response: {
+									name,
+									content: text + imageText,
+								},
+							},
+						} as FunctionResponsePart,
+						...imageParts.map(
+							(part) =>
+								({
+									inlineData: {
+										data: part.source.data,
+										mimeType: part.source.media_type,
+									},
+								} as InlineDataPart)
+						),
+					]
 				}
 			default:
 				throw new Error(`Unsupported content block type: ${(block as any).type}`)
@@ -52,7 +95,7 @@ export function convertAnthropicContentToGemini(
 
 export function convertAnthropicMessageToGemini(message: Anthropic.Messages.MessageParam): Content {
 	return {
-		role: message.role === "assistant" ? "model" : message.role,
+		role: message.role === "assistant" ? "model" : "user",
 		parts: convertAnthropicContentToGemini(message.content),
 	}
 }
@@ -77,6 +120,13 @@ export function convertAnthropicToolToGemini(tool: Anthropic.Messages.Tool): Fun
 	}
 }
 
+/*
+It looks like gemini likes to double escape certain characters when writing file contents: https://discuss.ai.google.dev/t/function-call-string-property-is-double-escaped/37867
+*/
+export function unescapeGeminiContent(content: string) {
+	return content.replace(/\\n/g, "\n").replace(/\\'/g, "'").replace(/\\"/g, '"')
+}
+
 export function convertGeminiResponseToAnthropic(
 	response: EnhancedGenerateContentResponse
 ): Anthropic.Messages.Message {
@@ -92,9 +142,12 @@ export function convertGeminiResponseToAnthropic(
 	const functionCalls = response.functionCalls()
 	if (functionCalls) {
 		functionCalls.forEach((call, index) => {
+			if ("content" in call.args && typeof call.args.content === "string") {
+				call.args.content = unescapeGeminiContent(call.args.content)
+			}
 			content.push({
 				type: "tool_use",
-				id: `tool_${index}`,
+				id: `${call.name}-${index}-${Date.now()}`,
 				name: call.name,
 				input: call.args,
 			})