@codebolt/codeboltjs 2.0.16 → 2.0.17

package/dist/agentlib/promptbuilder.js

@@ -1,805 +0,0 @@
-"use strict";
-var __importDefault = (this && this.__importDefault) || function (mod) {
-    return (mod && mod.__esModule) ? mod : { "default": mod };
-};
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.InitialPromptBuilder = void 0;
-const systemprompt_1 = require("./systemprompt");
-const js_yaml_1 = __importDefault(require("js-yaml"));
-const fs_1 = __importDefault(require("fs"));
-const path_1 = __importDefault(require("path"));
-// All interfaces moved to appropriate type files
-/**
- * PromptBuilder class for constructing prompts using a fluent interface.
- * Allows chaining methods to build complex prompts from user messages.
- */
-class InitialPromptBuilder {
-    /**
-     * Creates a new PromptBuilder instance.
-     *
-     * @param userMessage - The initial user message containing text and mentioned entities
-     * @param codebolt - Optional codebolt API instance for automatic tool and environment loading
-     */
-    constructor(userMessage, codebolt) {
-        var _a;
-        /** System prompt text */
-        this.systemPromptText = "";
-        /** Task instruction text */
-        this.taskInstructionText = "";
-        /** Available tools */
-        this.tools = [];
-        /** Conversation history */
-        this.conversationHistory = [];
-        /** Environment details */
-        this.environmentDetails = "";
-        /** File contents cache */
-        this.fileContents = new Map();
-        // Flags for what needs to be loaded during build
-        this.shouldLoadMCPTools = false;
-        this.mcpAdditionalServers = ["codebolt"];
-        this.shouldLoadAgentTools = false;
-        this.shouldLoadEnvironmentDetails = false;
-        this.shouldLoadAll = false;
-        // Handle both InitialUserMessage and CLIUserMessage types
-        if ('message' in userMessage && typeof userMessage.message === 'object') {
-            // This is a CLIUserMessage with nested message structure
-            this.message = userMessage.message.userMessage || ((_a = userMessage.data) === null || _a === void 0 ? void 0 : _a.text.trim()) || "";
-            this.mentionedFiles = userMessage.message.mentionedFiles || [];
-            // Convert string array to MCPTool array for CLIUserMessage
-            this.mentionedMCPs = (userMessage.message.mentionedMCPs || []).map((name) => ({ name }));
-            this.mentionedAgents = userMessage.message.mentionedAgents || [];
-        }
-        else {
-            // This is an InitialUserMessage
-            this.message = userMessage.messageText || userMessage.userMessage || "";
-            this.mentionedFiles = userMessage.mentionedFiles || [];
-            this.mentionedMCPs = userMessage.mentionedMCPs || [];
-            this.mentionedAgents = userMessage.mentionedAgents || [];
-        }
-        this.promptParts = [this.message];
-        this.codebolt = codebolt;
-    }
-    /**
-     * Sets the codebolt API instance for automatic operations.
-     *
-     * @param codebolt - The codebolt API instance
-     * @returns The PromptBuilder instance for chaining
-     */
-    setCodeboltAPI(codebolt) {
-        this.codebolt = codebolt;
-        return this;
-    }
-    /**
-     * Marks MCP tools to be loaded during build from the mentioned MCPs and additional servers.
-     *
-     * @param additionalServers - Additional MCP servers to load tools from
-     * @returns The PromptBuilder instance for chaining
-     */
-    addMCPTools(additionalServers = ["codebolt"]) {
-        this.shouldLoadMCPTools = true;
-        this.mcpAdditionalServers = additionalServers;
-        return this;
-    }
-    /**
-     * Marks mentioned agents to be converted to OpenAI tool format during build.
-     *
-     * @returns The PromptBuilder instance for chaining
-     */
-    addAgentTools() {
-        this.shouldLoadAgentTools = true;
-        return this;
-    }
-    /**
-     * Marks environment details to be loaded during build.
-     *
-     * @returns The PromptBuilder instance for chaining
-     */
-    addEnvironmentDetails() {
-        this.shouldLoadEnvironmentDetails = true;
-        return this;
-    }
-    /**
-     * Convenience method to mark all tools and environment details to be loaded during build.
-     * Equivalent to calling addMCPTools(), addAgentTools(), and addEnvironmentDetails().
-     *
-     * @param additionalServers - Additional MCP servers to load tools from
-     * @returns The PromptBuilder instance for chaining
-     */
-    addAllAutomatic(additionalServers = ["codebolt"]) {
-        this.shouldLoadAll = true;
-        this.mcpAdditionalServers = additionalServers;
-        return this;
-    }
-    /**
-     * Loads MCP tools from the codebolt API.
-     * Internal method called during build.
-     */
-    async loadMCPTools() {
-        if (!this.codebolt) {
-            console.warn("Codebolt API not available. Cannot load MCP tools automatically.");
-            return;
-        }
-        try {
-            // Get default tools from specified servers
-            const { data: defaultTools } = await this.codebolt.mcp.listMcpFromServers(this.mcpAdditionalServers);
-            this.tools = [...this.tools, ...defaultTools];
-            // Get tools from mentioned MCPs
-            if (this.mentionedMCPs && this.mentionedMCPs.length > 0) {
-                const { data: mcpTools } = await this.codebolt.mcp.getTools(this.mentionedMCPs);
-                this.tools = [...this.tools, ...mcpTools];
-            }
-        }
-        catch (error) {
-            console.error(`Error loading MCP tools: ${error}`);
-        }
-    }
-    /**
-     * Converts mentioned agents to OpenAI tool format.
-     * Internal method called during build.
-     */
-    loadAgentTools() {
-        if (this.mentionedAgents && this.mentionedAgents.length > 0) {
-            const agentTools = this.mentionedAgents.map(agent => ({
-                type: "function",
-                function: {
-                    name: `subagent--${agent.unique_id}`,
-                    description: agent.longDescription || agent.description || "Agent tool",
-                    parameters: {
-                        type: "object",
-                        properties: {
-                            task: {
-                                type: "string",
-                                description: "The task to be executed by the tool."
-                            }
-                        },
-                        required: ["task"]
-                    }
-                }
-            }));
-            this.tools = [...this.tools, ...agentTools];
-        }
-    }
-    /**
-     * Loads file contents and environment details from the codebolt API.
-     * Internal method called during build.
-     */
-    async loadEnvironmentDetails() {
-        if (!this.codebolt) {
-            console.warn("Codebolt API not available. Cannot load environment details automatically.");
-            return;
-        }
-        try {
-            // Load mentioned file contents
-            if (this.mentionedFiles && this.mentionedFiles.length > 0) {
-                for (const file of this.mentionedFiles) {
-                    try {
-                        const fileData = await this.codebolt.fs.readFile(file);
-                        this.fileContents.set(file, fileData);
-                    }
-                    catch (error) {
-                        console.error(`Error reading file ${file}: ${error}`);
-                        this.fileContents.set(file, "[File could not be read]");
-                    }
-                }
-            }
-            // Get project path and file listing
-            const { projectPath } = await this.codebolt.project.getProjectPath();
-            if (projectPath) {
-                const { success, result } = await this.codebolt.fs.listFile(projectPath, true);
-                if (success) {
-                    this.environmentDetails = `\n\n<environment_details>\n\n# Current Working Directory (${projectPath}) Files\n${result}\n</environment_details>`;
-                }
-            }
-        }
-        catch (error) {
-            console.error(`Error loading environment details: ${error}`);
-        }
-    }
-    /**
-     * Adds system prompt from a YAML file.
-     *
-     * @param filepath - Path to the YAML file containing system prompts
-     * @param key - Key identifier for the specific prompt
-     * @param exampleFilePath - Optional path to example file to append
-     * @returns The PromptBuilder instance for chaining
-     */
-    addSystemPrompt(filepath, key, exampleFilePath) {
-        try {
-            const systemPrompt = new systemprompt_1.SystemPrompt(filepath, key);
-            this.systemPromptText = systemPrompt.toPromptText();
-            // Add example file if provided
-            if (exampleFilePath) {
-                try {
-                    const example = fs_1.default.readFileSync(path_1.default.resolve(exampleFilePath), 'utf8');
-                    this.systemPromptText += `\n\n<example_agent>\n\`\`\`\n${example}\n\`\`\`\n</example_agent>`;
-                }
-                catch (error) {
-                    console.error(`Error loading example file: ${error}`);
-                }
-            }
-        }
-        catch (error) {
-            console.error(`Error loading system prompt: ${error}`);
-        }
-        return this;
-    }
-    /**
-     * Adds task instruction from a YAML file.
-     *
-     * @param filepath - Path to the YAML file containing task instructions
-     * @param refsection - Section name within the YAML file
-     * @returns The PromptBuilder instance for chaining
-     */
-    addTaskInstruction(filepath, refsection) {
-        try {
-            const fileContents = fs_1.default.readFileSync(path_1.default.resolve(filepath), 'utf8');
-            const data = js_yaml_1.default.load(fileContents);
-            const task = data[refsection];
-            if (task) {
-                this.taskInstructionText = `Task Description: ${task.description}\nExpected Output: ${task.expected_output}`;
-            }
-        }
-        catch (error) {
-            console.error(`Error loading task instruction: ${error}`);
-        }
-        return this;
-    }
-    /**
-     * Manually adds environment details with file listing.
-     * Use addEnvironmentDetails() for automatic loading instead.
-     *
-     * @param projectPath - The project path
-     * @param fileListResult - The file listing result
-     * @returns The PromptBuilder instance for chaining
-     */
-    setEnvironmentDetails(projectPath, fileListResult) {
-        if (projectPath && fileListResult) {
-            this.environmentDetails = `\n\n<environment_details>\n\n# Current Working Directory (${projectPath}) Files\n${fileListResult}\n</environment_details>`;
-        }
-        return this;
-    }
-    /**
-     * Adds MCP tools information to the prompt.
-     * Only adds content if there are mentioned MCPs.
-     *
-     * @returns The PromptBuilder instance for chaining
-     */
-    addMCPToolsToPrompt() {
-        if (this.mentionedMCPs.length > 0) {
-            const mcpTools = this.mentionedMCPs.map(mcp => {
-                const name = mcp.name || mcp.toolbox || "Unknown";
-                const toolName = mcp.toolName || "";
-                const tools = mcp.tools ? JSON.stringify(mcp.tools) : "[]";
-                let mcpInfo = `MCP: ${name}`;
-                if (toolName) {
-                    mcpInfo += `, Tool: ${toolName}`;
-                }
-                mcpInfo += `, Tools: ${tools}`;
-                return mcpInfo;
-            }).join("\n");
-            this.promptParts.push(`[MCP Tools]\n${mcpTools}`);
-        }
-        return this;
-    }
-    /**
-     * Adds agent information to the prompt.
-     * Only adds content if there are mentioned agents.
-     *
-     * @returns The PromptBuilder instance for chaining
-     */
-    addAgentsToPrompt() {
-        if (this.mentionedAgents.length > 0) {
-            const agentList = this.mentionedAgents.map(agent => {
-                const identifier = agent.name || agent.title || agent.id || agent.agent_id || "Unknown";
-                let agentInfo = `Agent: ${identifier}`;
-                if (agent.description) {
-                    agentInfo += ` - ${agent.description}`;
-                }
-                return agentInfo;
-            }).join("\n");
-            this.promptParts.push(`[Agents]\n${agentList}`);
-        }
-        return this;
-    }
-    /**
-     * Manually sets the available tools for the conversation.
-     * Use addMCPTools() and addAgentTools() for automatic loading instead.
-     *
-     * @param tools - Array of OpenAI tools
-     * @returns The PromptBuilder instance for chaining
-     */
-    setTools(tools) {
-        this.tools = [...tools];
-        return this;
-    }
-    /**
-     * Adds additional tools to the existing tool list.
-     *
-     * @param tools - Array of OpenAI tools to add
-     * @returns The PromptBuilder instance for chaining
-     */
-    addTools(tools) {
-        this.tools = [...this.tools, ...tools];
-        return this;
-    }
-    /**
-     * Builds the user message content with files (without environment details).
-     *
-     * @returns Array of content blocks
-     */
-    buildUserMessageContent() {
-        let finalPrompt = `
-                    The user has sent the following query:
-                   <user_query> ${this.message} </user_query>.
-                `;
-        // Attach files if mentioned
-        if (this.mentionedFiles && this.mentionedFiles.length > 0) {
-            finalPrompt += `The Attached files are:`;
-            for (const file of this.mentionedFiles) {
-                const fileData = this.fileContents.get(file) || "[File content not available]";
-                finalPrompt += `File Name: ${file}, File Path: ${file}, Filedata: ${fileData}`;
-            }
-        }
-        const content = [
-            { type: "text", text: finalPrompt }
-        ];
-        // Add environment details
-        if (this.environmentDetails) {
-            content.push({ type: "text", text: this.environmentDetails });
-        }
-        // Add task instruction if available
-        if (this.taskInstructionText) {
-            content.push({
-                type: "text",
-                text: this.taskInstructionText
-            });
-        }
-        return content;
-    }
-    /**
-     * Builds the OpenAI conversation format.
-     *
-     * @returns Array of OpenAI messages
-     */
-    buildOpenAIMessages() {
-        const messages = [];
-        // Add system message if available
-        if (this.systemPromptText) {
-            messages.push({
-                role: "system",
-                content: this.systemPromptText
-            });
-        }
-        // Add user message
-        const userContent = this.buildUserMessageContent();
-        messages.push({
-            role: "user",
-            content: userContent
-        });
-        // Add any existing conversation history
-        messages.push(...this.conversationHistory);
-        return messages;
-    }
-    /**
-     * Gets the available tools in OpenAI format.
-     *
-     * @returns Array of OpenAI tools
-     */
-    getTools() {
-        return this.tools;
-    }
-    /**
-     * Gets the loaded file contents.
-     *
-     * @returns Map of file paths to their contents
-     */
-    getFileContents() {
-        return new Map(this.fileContents);
-    }
-    /**
-     * Adds a message to the conversation history.
-     *
-     * @param role - The role of the message sender
-     * @param content - The content of the message
-     * @param toolCallId - Optional tool call ID for tool messages
-     * @param toolCalls - Optional tool calls for assistant messages
-     * @returns The PromptBuilder instance for chaining
-     */
-    addToConversationHistory(role, content, toolCallId, toolCalls) {
-        const message = { role, content };
-        if (toolCallId) {
-            message.tool_call_id = toolCallId;
-        }
-        if (toolCalls) {
-            message.tool_calls = toolCalls;
-        }
-        this.conversationHistory.push(message);
-        return this;
-    }
-    /**
-     * Gets the current conversation history.
-     *
-     * @returns Array of conversation entries
-     */
-    getConversationHistory() {
-        return [...this.conversationHistory];
-    }
-    /**
-     * Adds an LLM response to the conversation history.
-     * This method processes the LLM response and adds the assistant's message to the conversation.
-     * It handles both resolved responses and promises that resolve to responses.
-     *
-     * @param llmResponse - The LLM response (can be a promise or resolved response)
-     * @returns The PromptBuilder instance for chaining
-     *
-     * @example
-     * ```typescript
-     * // Example of using addLLMResponse in an agent workflow loop
-     * import { PromptBuilder } from './promptbuilder';
-     * import llm from '../modules/llm';
-     *
-     * async function agentWorkflowWithLLMResponse(userMessage: any, codebolt: CodeboltAPI) {
-     *     // Step 1: Build initial prompt
-     *     const promptBuilder = new PromptBuilder(userMessage, codebolt);
-     *     let currentPrompt = await promptBuilder
-     *         .addMCPTools()
-     *         .addAgentTools()
-     *         .addEnvironmentDetails()
-     *         .buildInferenceParams();
-     *
-     *     let completed = false;
-     *     let maxTurns = 20;
-     *     let turn = 0;
-     *
-     *     // Step 2: Main conversation loop (similar to agent.ts while loop)
-     *     while (!completed && turn < maxTurns) {
-     *         // Get LLM response
-     *         const llmResponse = llm.inference(currentPrompt);
-     *
-     *         // Add LLM response to conversation history
-     *         await promptBuilder.addLLMResponse(llmResponse);
-     *
-     *         // Process the response using LLMOutputHandler
-     *         const outputHandler = new LLMOutputHandler(llmResponse, codebolt);
-     *         await outputHandler.sendMessageToUser();
-     *
-     *         // Check if task is completed
-     *         if (outputHandler.isCompleted()) {
-     *             completed = true;
-     *             break;
-     *         }
-     *
-     *         // Execute tools and get results
-     *         const toolResults = await outputHandler.runTools();
-     *
-     *         // Add tool results to conversation
-     *         if (toolResults.length > 0) {
-     *             promptBuilder.addToolResults(toolResults);
-     *         } else {
-     *             // Add default continuation message when no tools executed
-     *             promptBuilder.addDefaultContinuationMessage();
-     *         }
-     *
-     *         // Build next prompt for the loop
-     *         currentPrompt = await promptBuilder.buildInferenceParams();
-     *         turn++;
-     *     }
-     * }
-     * ```
-     */
-    async addLLMResponse(llmResponse) {
-        try {
-            // Resolve the response if it's a promise
-            const resolvedResponse = await Promise.resolve(llmResponse);
-            if (!resolvedResponse || !resolvedResponse.completion) {
-                console.warn("Invalid LLM response provided");
-                return this;
-            }
-            const completion = resolvedResponse.completion;
-            let assistantMessage = null;
-            // Handle different response formats
-            if (completion.choices && completion.choices.length > 0) {
-                // OpenAI-style response with choices
-                const choice = completion.choices[0];
-                if (choice.message) {
-                    assistantMessage = {
-                        role: "assistant",
-                        content: choice.message.content || "",
-                        tool_calls: choice.message.tool_calls || undefined
-                    };
-                }
-            }
-            else if (completion.content) {
-                // Direct content response
-                assistantMessage = {
-                    role: "assistant",
-                    content: completion.content
-                };
-            }
-            else if (completion.message) {
-                // Message format response
-                assistantMessage = {
-                    role: "assistant",
-                    content: completion.message.content || "",
-                    tool_calls: completion.message.tool_calls || undefined
-                };
-            }
-            // Add the assistant message to conversation history
-            if (assistantMessage) {
-                this.conversationHistory.push(assistantMessage);
-            }
-            else {
-                // Fallback for cases where no valid message is found
-                this.conversationHistory.push({
-                    role: "assistant",
-                    content: "I apologize, but I was unable to provide a proper response."
-                });
-            }
-        }
-        catch (error) {
-            console.error("Error adding LLM response to conversation:", error);
-            // Add error message to conversation history
-            this.conversationHistory.push({
-                role: "assistant",
-                content: "An error occurred while processing my response."
-            });
-        }
-        return this;
-    }
-    /**
-     * Adds tool results to the conversation history.
-     * This method is typically used after executing tools from an LLM response.
-     *
-     * @param toolResults - Array of tool results to add to the conversation
-     * @returns The PromptBuilder instance for chaining
-     */
-    addToolResults(toolResults) {
-        if (toolResults && toolResults.length > 0) {
-            // Add each tool result as a separate message
-            toolResults.forEach(toolResult => {
-                this.conversationHistory.push({
-                    role: "tool",
-                    content: toolResult.content,
-                    tool_call_id: toolResult.tool_call_id
-                });
-            });
-        }
-        return this;
-    }
-    /**
-     * Adds a user message to the conversation history.
-     * Useful for adding follow-up questions or additional context during the conversation.
-     *
-     * @param message - The user message to add
-     * @returns The PromptBuilder instance for chaining
-     */
-    addUserMessage(message) {
-        const userMessage = {
-            role: "user",
-            content: message
-        };
-        this.conversationHistory.push(userMessage);
-        return this;
-    }
-    /**
-     * Adds a default continuation message when no tools were executed.
-     * This is used in the agent workflow to prompt the LLM to either complete the task
-     * or ask for more information.
-     *
-     * @returns The PromptBuilder instance for chaining
-     */
-    addDefaultContinuationMessage() {
-        const defaultMessage = "If you have completed the user's task, use the attempt_completion tool. If you require additional information from the user, use the ask_followup_question tool. Otherwise, if you have not completed the task and do not need additional information, then proceed with the next step of the task. (This is an automated message, so do not respond to it conversationally.)";
-        this.addUserMessage([{
-                type: "text",
-                text: defaultMessage
-            }]);
-        return this;
-    }
-    /**
-     * Adds a custom text section to the prompt.
-     *
-     * @param title - Optional title for the section
-     * @param content - The content to add
-     * @returns The PromptBuilder instance for chaining
-     */
-    addCustomSection(title, content) {
-        if (content.trim()) {
-            const section = title ? `[${title}]\n${content}` : content;
-            this.promptParts.push(section);
-        }
-        return this;
-    }
-    /**
-     * Adds a system instruction to the prompt.
-     *
-     * @param instruction - The system instruction to add
-     * @returns The PromptBuilder instance for chaining
-     */
-    addSystemInstruction(instruction) {
-        if (instruction.trim()) {
-            this.promptParts.push(`[System Instruction]\n${instruction}`);
-        }
-        return this;
-    }
-    /**
-     * Adds context information to the prompt.
-     *
-     * @param context - The context information to add
-     * @returns The PromptBuilder instance for chaining
-     */
-    addContext(context) {
-        const userMessage = {
-            role: "user",
-            content: context
-        };
-        this.conversationHistory.push(userMessage);
-        return this;
-    }
-    /**
-     * Builds and returns the OpenAI message format with tools.
-     * This method performs all async operations that were marked during setup.
-     *
-     * @returns Object with messages, tools, and other LLM parameters
-     */
-    async build() {
-        // Perform all async operations based on what was requested
-        if (this.shouldLoadAll) {
-            await this.loadMCPTools();
-            this.loadAgentTools();
-            await this.loadEnvironmentDetails();
-        }
-        else {
-            if (this.shouldLoadMCPTools) {
-                await this.loadMCPTools();
-            }
-            if (this.shouldLoadAgentTools) {
-                this.loadAgentTools();
-            }
-            if (this.shouldLoadEnvironmentDetails) {
-                await this.loadEnvironmentDetails();
-            }
-        }
-        return {
-            messages: this.buildOpenAIMessages(),
-            tools: this.getTools(),
-            full: true,
-            max_tokens: 8192,
-            tool_choice: "auto",
-        };
-    }
-    /**
-     * Gets the current prompt parts without building the final string.
-     * Useful for debugging or inspection.
-     *
-     * @returns Array of current prompt parts
-     */
-    getPromptParts() {
-        return [...this.promptParts];
-    }
-    /**
-     * Clears all prompt parts except the initial message.
-     * Useful for starting over with the same message.
-     *
-     * @returns The PromptBuilder instance for chaining
-     */
-    reset() {
-        this.promptParts = [this.message];
-        this.conversationHistory = [];
-        this.systemPromptText = "";
-        this.taskInstructionText = "";
-        this.tools = [];
-        this.environmentDetails = "";
-        this.fileContents.clear();
-        // Reset async loading flags
-        this.shouldLoadMCPTools = false;
-        this.shouldLoadAgentTools = false;
-        this.shouldLoadEnvironmentDetails = false;
-        this.shouldLoadAll = false;
-        this.mcpAdditionalServers = ["codebolt"];
-        return this;
-    }
-    /**
-     * Creates an LLM inference parameters object in the format expected by the sample code.
-     * This method performs all async operations that were marked during setup.
-     *
-     * @returns Object with messages, tools, and other LLM parameters
-     */
-    async buildInferenceParams() {
-        // Perform all async operations based on what was requested
-        if (this.shouldLoadAll) {
-            await this.loadMCPTools();
-            this.loadAgentTools();
-            await this.loadEnvironmentDetails();
-        }
-        else {
-            if (this.shouldLoadMCPTools) {
-                await this.loadMCPTools();
-            }
-            if (this.shouldLoadAgentTools) {
-                this.loadAgentTools();
-            }
-            if (this.shouldLoadEnvironmentDetails) {
-                await this.loadEnvironmentDetails();
-            }
-        }
-        return {
-            full: true,
-            messages: this.buildOpenAIMessages(),
-            tools: this.getTools(),
-            tool_choice: "auto",
-        };
-    }
-    /**
-     * Creates a minimal LLM inference parameters object with only the user message.
-     * This is useful for simple completions or when no tools/context are needed.
-     *
-     * @returns Object with messages (user only), no tools, and default tool_choice
-     */
-    //TODO: implement this
-    buildDefaultInferenceParams() {
-        return {
-            full: false,
-            messages: [this.message],
-            tools: [],
-            tool_choice: "auto",
-        };
-    }
-    /**
-     * Checks if the last LLM response indicates task completion.
-     * This method examines the conversation history for completion signals.
-     *
-     * @returns Boolean indicating if the task appears to be completed
-     */
-    isTaskCompleted() {
-        if (this.conversationHistory.length === 0) {
-            return false;
-        }
-        // Get the last assistant message
-        const lastAssistantMessage = this.conversationHistory
-            .slice()
-            .reverse()
-            .find(msg => msg.role === 'assistant');
-        if (!lastAssistantMessage) {
-            return false;
-        }
-        // Check if the message has tool calls with completion tools
-        if (lastAssistantMessage.tool_calls) {
-            const hasCompletionTool = lastAssistantMessage.tool_calls.some(toolCall => toolCall.function && toolCall.function.name.includes('attempt_completion'));
-            if (hasCompletionTool) {
-                return true;
-            }
-        }
-        // Check for completion keywords in the message content
-        const content = typeof lastAssistantMessage.content === 'string'
-            ? lastAssistantMessage.content
-            : '';
-        const completionKeywords = [
-            'task completed',
-            'task is completed',
-            'successfully completed',
-            'finished the task',
-            'task has been completed'
-        ];
-        return completionKeywords.some(keyword => content.toLowerCase().includes(keyword.toLowerCase()));
-    }
-    /**
-     * Gets the current conversation length (number of messages).
-     * Useful for determining when to summarize the conversation.
-     *
-     * @returns Number of messages in the conversation history
-     */
-    getConversationLength() {
-        return this.conversationHistory.length;
-    }
-    /**
-     * Checks if the conversation should be summarized based on length.
-     *
-     * @param maxLength - Maximum conversation length before summarization (default: 50)
-     * @returns Boolean indicating if summarization is needed
-     */
-    shouldSummarizeConversation(maxLength = 50) {
-        return this.getConversationLength() > maxLength;
-    }
-}
-exports.InitialPromptBuilder = InitialPromptBuilder;