multimind-sdk/dist/src/contextTransfer.js

This SDK gives JavaScript/TypeScript developers full access to advanced AI features like agent orchestration, RAG, and fine-tuning — without needing to manage backend code.

import { py } from './bridge/multimind-bridge.js';
export class ContextTransferManager {
    constructor(config) {
        this.supportedModels = {
            "chatgpt": "ChatGPT",
            "deepseek": "DeepSeek",
            "claude": "Claude",
            "gemini": "Gemini",
            "mistral": "Mistral",
            "llama": "Llama",
            "cohere": "Cohere",
            "anthropic_claude": "Anthropic Claude",
            "openai_gpt4": "OpenAI GPT-4",
            "gpt4": "GPT-4",
            "gpt-4": "GPT-4",
            "gpt3": "GPT-3",
            "gpt-3": "GPT-3",
            "claude-3": "Claude-3",
            "claude-2": "Claude-2",
            "claude-1": "Claude-1"
        };
        this.config = {
            maxContextLength: 32000,
            defaultSummaryLength: 1000,
            includeMetadata: true,
            preserveFormatting: true,
            smartTruncation: true,
            contextCompression: false,
            ...config
        };
    }
    async extractContext(messages, lastN = 5, smartExtraction = true) {
        if (!messages.length)
            return [];
        if (smartExtraction) {
            return this.smartExtractContext(messages, lastN);
        }
        else {
            const extractedCount = Math.min(lastN, messages.length);
            const extractedMessages = messages.slice(-extractedCount);
            console.log(`Extracted ${extractedMessages.length} messages from conversation`);
            return extractedMessages;
        }
    }
    smartExtractContext(messages, lastN) {
        if (messages.length <= lastN)
            return messages;
        const recentMessages = messages.slice(-lastN);
        const importantContext = [];
        for (const msg of messages.slice(0, -lastN)) {
            if (msg.role === 'system' || this.isImportantContext(msg.content)) {
                importantContext.push(msg);
            }
        }
        if (importantContext.length) {
            const contextToInclude = importantContext.slice(-2);
            const combined = [
                ...contextToInclude,
                ...recentMessages.slice(-(lastN - contextToInclude.length))
            ];
            console.log(`Smart extraction: ${contextToInclude.length} important + ${combined.length - contextToInclude.length} recent messages`);
            return combined;
        }
        return recentMessages;
    }
    isImportantContext(content) {
        const importantKeywords = [
            "system", "setup", "configuration", "requirements", "constraints",
            "important", "note", "warning", "error", "critical", "essential"
        ];
        const contentLower = content.toLowerCase();
        return importantKeywords.some(keyword => contentLower.includes(keyword));
    }
    async summarizeContext(messages, summaryType = "concise") {
        if (!messages.length)
            return "No conversation context available.";
        switch (summaryType) {
            case "structured":
                return this.createStructuredSummary(messages);
            case "detailed":
                return this.createDetailedSummary(messages);
            default:
                return this.createConciseSummary(messages);
        }
    }
    createConciseSummary(messages) {
        const summaryParts = [];
        for (const message of messages) {
            let content = message.content;
            if (content.length > 500) {
                content = content.substring(0, 500) + "...";
            }
            switch (message.role) {
                case 'user':
                    summaryParts.push(`User: ${content}`);
                    break;
                case 'assistant':
                    summaryParts.push(`Assistant: ${content}`);
                    break;
                case 'system':
                    summaryParts.push(`System: ${content}`);
                    break;
            }
        }
        const summary = summaryParts.join('\n');
        console.log(`Generated concise summary with ${summaryParts.length} parts`);
        return summary;
    }
    createDetailedSummary(messages) {
        const summaryParts = [];
        for (let i = 0; i < messages.length; i++) {
            const message = messages[i];
            switch (message.role) {
                case 'user':
                    summaryParts.push(`User (Message ${i + 1}): ${message.content}`);
                    break;
                case 'assistant':
                    summaryParts.push(`Assistant (Response ${i + 1}): ${message.content}`);
                    break;
                case 'system':
                    summaryParts.push(`System Configuration: ${message.content}`);
                    break;
            }
        }
        const summary = summaryParts.join('\n\n');
        console.log(`Generated detailed summary with ${summaryParts.length} parts`);
        return summary;
    }
    createStructuredSummary(messages) {
        const userMessages = [];
        const assistantMessages = [];
        const systemMessages = [];
        for (const message of messages) {
            switch (message.role) {
                case 'user':
                    userMessages.push(message.content);
                    break;
                case 'assistant':
                    assistantMessages.push(message.content);
                    break;
                case 'system':
                    systemMessages.push(message.content);
                    break;
            }
        }
        const summaryParts = [];
        if (systemMessages.length) {
            summaryParts.push("System Context:");
            summaryParts.push(...systemMessages.map(msg => `- ${msg}`));
            summaryParts.push("");
        }
        summaryParts.push("Conversation Flow:");
        for (let i = 0; i < Math.max(userMessages.length, assistantMessages.length); i++) {
            if (i < userMessages.length) {
                summaryParts.push(`User: ${userMessages[i]}`);
            }
            if (i < assistantMessages.length) {
                summaryParts.push(`Assistant: ${assistantMessages[i]}`);
            }
            summaryParts.push("");
        }
        const summary = summaryParts.join('\n').trim();
        console.log(`Generated structured summary with ${userMessages.length} user and ${assistantMessages.length} assistant messages`);
        return summary;
    }
    async loadConversationFromFile(filePath, formatType = "auto") {
        try {
            if (formatType === "auto") {
                formatType = this.detectFileFormat(filePath);
            }
            switch (formatType) {
                case "json":
                    return this.loadJsonConversation(filePath);
                case "txt":
                    return this.loadTextConversation(filePath);
                case "markdown":
                    return this.loadMarkdownConversation(filePath);
                default:
                    throw new Error(`Unsupported format: ${formatType}`);
            }
        }
        catch (error) {
            console.error(`Error loading conversation from ${filePath}:`, error);
            throw error;
        }
    }
    detectFileFormat(filePath) {
        const extension = filePath.split('.').pop()?.toLowerCase();
        if (extension === "json")
            return "json";
        if (extension === "md" || extension === "markdown")
            return "markdown";
        return "txt";
    }
    async loadJsonConversation(filePath) {
        const fs = await import('fs/promises');
        const data = JSON.parse(await fs.readFile(filePath, 'utf-8'));
        let messages;
        if (Array.isArray(data)) {
            messages = data;
        }
        else if (data.messages) {
            messages = data.messages;
        }
        else if (data.conversation) {
            messages = data.conversation;
        }
        else {
            throw new Error("Invalid JSON structure. Expected list of messages or dict with 'messages' key.");
        }
        console.log(`Loaded ${messages.length} messages from JSON file ${filePath}`);
        return messages;
    }
    async loadTextConversation(filePath) {
        const fs = await import('fs/promises');
        const content = await fs.readFile(filePath, 'utf-8');
        const lines = content.trim().split('\n');
        const messages = [];
        let currentRole = null;
        let currentContent = [];
        for (const line of lines) {
            if (line.startsWith('User:') || line.startsWith('Assistant:') || line.startsWith('System:')) {
                if (currentRole && currentContent.length) {
                    messages.push({
                        role: currentRole,
                        content: currentContent.join('\n').trim()
                    });
                }
                if (line.startsWith('User:')) {
                    currentRole = "user";
                }
                else if (line.startsWith('Assistant:')) {
                    currentRole = "assistant";
                }
                else if (line.startsWith('System:')) {
                    currentRole = "system";
                }
                currentContent = [line.split(':', 1)[1]?.trim() || ''];
            }
            else {
                currentContent.push(line);
            }
        }
        if (currentRole && currentContent.length) {
            messages.push({
                role: currentRole,
                content: currentContent.join('\n').trim()
            });
        }
        console.log(`Loaded ${messages.length} messages from text file ${filePath}`);
        return messages;
    }
    async loadMarkdownConversation(filePath) {
        const fs = await import('fs/promises');
        const content = await fs.readFile(filePath, 'utf-8');
        const lines = content.split('\n');
        const messages = [];
        let currentRole = null;
        let currentContent = [];
        for (const line of lines) {
            if (line.startsWith('### User:') || line.startsWith('### Assistant:') || line.startsWith('### System:')) {
                if (currentRole && currentContent.length) {
                    messages.push({
                        role: currentRole,
                        content: currentContent.join('\n').trim()
                    });
                }
                if (line.startsWith('### User:')) {
                    currentRole = "user";
                }
                else if (line.startsWith('### Assistant:')) {
                    currentRole = "assistant";
                }
                else if (line.startsWith('### System:')) {
                    currentRole = "system";
                }
                currentContent = [line.split(':', 1)[1]?.trim() || ''];
            }
            else {
                currentContent.push(line);
            }
        }
        if (currentRole && currentContent.length) {
            messages.push({
                role: currentRole,
                content: currentContent.join('\n').trim()
            });
        }
        console.log(`Loaded ${messages.length} messages from markdown file ${filePath}`);
        return messages;
    }
    async saveFormattedPrompt(content, outputFile, formatType = "txt") {
        try {
            switch (formatType) {
                case "json":
                    await this.saveJsonPrompt(content, outputFile);
                    break;
                case "markdown":
                    await this.saveMarkdownPrompt(content, outputFile);
                    break;
                default:
                    await this.saveTextPrompt(content, outputFile);
            }
            console.log(`Formatted prompt saved to ${outputFile} in ${formatType} format`);
        }
        catch (error) {
            console.error(`Error saving formatted prompt to ${outputFile}:`, error);
            throw error;
        }
    }
    async saveTextPrompt(content, outputFile) {
        const fs = await import('fs/promises');
        await fs.writeFile(outputFile, content, 'utf-8');
    }
    async saveJsonPrompt(content, outputFile) {
        const fs = await import('fs/promises');
        const promptData = {
            prompt: content,
            metadata: {
                created_at: new Date().toISOString(),
                format: "json",
                length: content.length
            }
        };
        await fs.writeFile(outputFile, JSON.stringify(promptData, null, 2), 'utf-8');
    }
    async saveMarkdownPrompt(content, outputFile) {
        const fs = await import('fs/promises');
        const markdownContent = `# Formatted Prompt

## Content

${content}

---
*Generated by MultiMind Context Transfer*
`;
        await fs.writeFile(outputFile, markdownContent, 'utf-8');
    }
    async transferContext(fromModel, toModel, inputFile, outputFile, lastN = 5, includeSummary = true, summaryType = "concise", smartExtraction = true, outputFormat = "txt", options = {}) {
        const messages = await this.loadConversationFromFile(inputFile);
        const extractedMessages = await this.extractContext(messages, lastN, smartExtraction);
        let summary;
        if (includeSummary) {
            summary = await this.summarizeContext(extractedMessages, summaryType);
        }
        else {
            summary = await this.summarizeContext(extractedMessages.slice(-1), "concise");
        }
        const formattedPrompt = await this.formatForTargetModel(toModel, summary, fromModel, options);
        await this.saveFormattedPrompt(formattedPrompt, outputFile, outputFormat);
        return formattedPrompt;
    }
    async formatForTargetModel(targetModel, summary, sourceModel, options = {}) {
        const targetModelLower = targetModel.toLowerCase().replace(/[ -]/g, "_");
        try {
            const adapter = await this.getAdapter(targetModelLower);
            return adapter.formatContext(summary, sourceModel, options);
        }
        catch (error) {
            return this.formatGeneric(summary, sourceModel, targetModel, options);
        }
    }
    async getAdapter(modelName) {
        try {
            return await py `AdapterFactory.get_adapter(${modelName})`;
        }
        catch (error) {
            throw new Error(`Model '${modelName}' not supported`);
        }
    }
    formatGeneric(summary, sourceModel, targetModel, options = {}) {
        const includeMetadata = options.include_metadata ?? this.config.includeMetadata;
        let prompt = `You are ${targetModel}, an AI assistant.

A user was previously working with ${sourceModel} on the following conversation:

${summary}

Please continue helping the user from where they left off. Maintain the context and provide helpful responses.`;
        if (includeMetadata) {
            prompt += `\n\n---\nContext transferred from ${sourceModel} to ${targetModel} using MultiMind SDK`;
        }
        return prompt;
    }
    getSupportedModels() {
        return Object.keys(this.supportedModels);
    }
    async getModelInfo(modelName) {
        try {
            const capabilities = await py `AdapterFactory.get_model_capabilities(${modelName})`;
            return capabilities;
        }
        catch (error) {
            return {
                name: modelName,
                supportedFormats: ["text"],
                maxContextLength: 8000,
                supportsCode: true,
                supportsImages: false,
                supportsTools: false
            };
        }
    }
    async listAllModels() {
        try {
            return await py `AdapterFactory.list_all_capabilities()`;
        }
        catch (error) {
            console.error('Error getting model capabilities:', error);
            return {};
        }
    }
}
export class ContextTransferAPI {
    constructor() {
        this.manager = new ContextTransferManager();
        this.supportedFormats = ["json", "txt", "markdown"];
        this.supportedModels = this.manager.getSupportedModels();
    }
    async transferContextAPI(sourceModel, targetModel, conversationData, options = {}) {
        try {
            const defaultOptions = {
                lastN: 5,
                includeSummary: true,
                summaryType: "concise",
                smartExtraction: true,
                outputFormat: "txt",
                includeMetadata: true,
                includeCodeContext: false,
                includeReasoning: false,
                includeSafety: false,
                includeCreativity: false,
                includeExamples: false,
                includeStepByStep: false,
                includeMultimodal: false,
                includeWebSearch: false
            };
            const finalOptions = { ...defaultOptions, ...options };
            const messages = this.processConversationData(conversationData);
            const extractedMessages = await this.manager.extractContext(messages, finalOptions.lastN, finalOptions.smartExtraction);
            let summary;
            if (finalOptions.includeSummary) {
                summary = await this.manager.summarizeContext(extractedMessages, finalOptions.summaryType);
            }
            else {
                summary = await this.manager.summarizeContext(extractedMessages.slice(-1), "concise");
            }
            const formattingOptions = Object.fromEntries(Object.entries(finalOptions).filter(([key]) => key.startsWith('include_') && key !== 'includeMetadata'));
            const formattedPrompt = await this.manager.formatForTargetModel(targetModel, summary, sourceModel, formattingOptions);
            const response = {
                success: true,
                formattedPrompt,
                metadata: {
                    sourceModel,
                    targetModel,
                    summaryType: finalOptions.summaryType,
                    smartExtraction: finalOptions.smartExtraction,
                    messagesProcessed: messages.length,
                    messagesExtracted: extractedMessages.length,
                    promptLength: formattedPrompt.length,
                    createdAt: new Date().toISOString(),
                    outputFormat: finalOptions.outputFormat
                }
            };
            if (finalOptions.includeMetadata) {
                response.metadata.modelCapabilities = {
                    source: await this.manager.getModelInfo(sourceModel),
                    target: await this.manager.getModelInfo(targetModel)
                };
            }
            console.log(`Context transfer completed: ${sourceModel} -> ${targetModel}`);
            return response;
        }
        catch (error) {
            console.error('Context transfer failed:', error);
            return {
                success: false,
                error: error instanceof Error ? error.message : String(error),
                errorType: error instanceof Error ? error.constructor.name : 'Unknown'
            };
        }
    }
    processConversationData(data) {
        if (typeof data === 'string') {
            // This would need to be handled asynchronously in a real implementation
            throw new Error('File path processing not implemented in this demo');
        }
        else if (Array.isArray(data)) {
            return data;
        }
        else if (typeof data === 'object') {
            if (data.messages) {
                return data.messages;
            }
            else if (data.conversation) {
                return data.conversation;
            }
            else {
                return [data];
            }
        }
        else {
            throw new Error(`Unsupported data type: ${typeof data}`);
        }
    }
    async getSupportedModels() {
        try {
            const capabilities = await this.manager.listAllModels();
            return {
                success: true,
                models: capabilities,
                totalModels: Object.keys(capabilities).length,
                supportedFormats: this.supportedFormats,
                metadata: {
                    generatedAt: new Date().toISOString(),
                    apiVersion: "2.0"
                }
            };
        }
        catch (error) {
            console.error('Failed to get supported models:', error);
            return {
                success: false,
                error: error instanceof Error ? error.message : String(error),
                errorType: error instanceof Error ? error.constructor.name : 'Unknown'
            };
        }
    }
    async getModelCapabilities(modelName) {
        try {
            const capabilities = await this.manager.getModelInfo(modelName);
            return {
                success: true,
                model: modelName,
                capabilities,
                metadata: {
                    generatedAt: new Date().toISOString()
                }
            };
        }
        catch (error) {
            return {
                success: false,
                error: error instanceof Error ? error.message : String(error),
                errorType: error instanceof Error ? error.constructor.name : 'Unknown'
            };
        }
    }
    async validateConversationFormat(data) {
        try {
            const messages = this.processConversationData(data);
            const analysis = {
                totalMessages: messages.length,
                userMessages: 0,
                assistantMessages: 0,
                systemMessages: 0,
                unknownMessages: 0,
                averageMessageLength: 0,
                hasSystemContext: false
            };
            let totalLength = 0;
            for (const msg of messages) {
                const role = msg.role;
                const content = msg.content;
                switch (role) {
                    case 'user':
                        analysis.userMessages++;
                        break;
                    case 'assistant':
                        analysis.assistantMessages++;
                        break;
                    case 'system':
                        analysis.systemMessages++;
                        analysis.hasSystemContext = true;
                        break;
                    default:
                        analysis.unknownMessages++;
                }
                totalLength += content.length;
            }
            if (analysis.totalMessages > 0) {
                analysis.averageMessageLength = totalLength / analysis.totalMessages;
            }
            return {
                success: true,
                valid: true,
                analysis,
                recommendations: this.generateRecommendations(analysis)
            };
        }
        catch (error) {
            return {
                success: false,
                valid: false,
                error: error instanceof Error ? error.message : String(error),
                errorType: error instanceof Error ? error.constructor.name : 'Unknown'
            };
        }
    }
    generateRecommendations(analysis) {
        const recommendations = [];
        if (analysis.totalMessages === 0) {
            recommendations.push("No messages found in conversation data");
        }
        if (analysis.userMessages === 0) {
            recommendations.push("No user messages found - ensure conversation has user input");
        }
        if (analysis.assistantMessages === 0) {
            recommendations.push("No assistant messages found - ensure conversation has AI responses");
        }
        if (analysis.unknownMessages > 0) {
            recommendations.push(`Found ${analysis.unknownMessages} messages with unknown roles`);
        }
        if (analysis.averageMessageLength > 1000) {
            recommendations.push("Long messages detected - consider using smart extraction");
        }
        if (analysis.totalMessages > 20) {
            recommendations.push("Large conversation detected - consider using smart extraction and detailed summary");
        }
        if (!analysis.hasSystemContext) {
            recommendations.push("No system context found - consider adding system messages for better context");
        }
        return recommendations;
    }
    async batchTransfer(transfers) {
        const results = [];
        let successful = 0;
        let failed = 0;
        for (let i = 0; i < transfers.length; i++) {
            const transferConfig = transfers[i];
            try {
                const result = await this.transferContextAPI(transferConfig.sourceModel, transferConfig.targetModel, transferConfig.conversationData, transferConfig.options);
                // Only add transferIndex if metadata exists
                if (result.metadata) {
                    result.metadata.transferIndex = i;
                }
                results.push(result);
                if (result.success) {
                    successful++;
                }
                else {
                    failed++;
                }
            }
            catch (error) {
                results.push({
                    success: false,
                    error: error instanceof Error ? error.message : String(error),
                    errorType: error instanceof Error ? error.constructor.name : 'Unknown',
                    // Do not set metadata for error case
                });
                failed++;
            }
        }
        return {
            success: failed === 0,
            totalTransfers: transfers.length,
            successfulTransfers: successful,
            failedTransfers: failed,
            results,
            metadata: {
                completedAt: new Date().toISOString()
            }
        };
    }
    createChromeExtensionConfig() {
        return {
            apiVersion: "2.0",
            supportedModels: this.supportedModels,
            supportedFormats: this.supportedFormats,
            defaultOptions: {
                lastN: 5,
                includeSummary: true,
                summaryType: "concise",
                smartExtraction: true,
                outputFormat: "txt"
            },
            chromeExtension: {
                manifestVersion: 3,
                permissions: ["activeTab", "storage"],
                contentScripts: ["content.js"],
                backgroundScripts: ["background.js"],
                popup: "popup.html"
            },
            endpoints: {
                transfer: "/api/transfer",
                models: "/api/models",
                validate: "/api/validate",
                batch: "/api/batch"
            },
            metadata: {
                generatedAt: new Date().toISOString(),
                sdkVersion: "2.0.0"
            }
        };
    }
}
// Convenience functions
export async function quickTransfer(sourceModel, targetModel, conversationData, options = {}) {
    const api = new ContextTransferAPI();
    const result = await api.transferContextAPI(sourceModel, targetModel, conversationData, options);
    if (result.success && result.formattedPrompt) {
        return result.formattedPrompt;
    }
    else {
        throw new Error(`Transfer failed: ${result.error || 'Unknown error'}`);
    }
}
export async function getAllModels() {
    const api = new ContextTransferAPI();
    return api.getSupportedModels();
}
export async function validateConversation(data) {
    const api = new ContextTransferAPI();
    return api.validateConversationFormat(data);
}
//# sourceMappingURL=contextTransfer.js.map