@noanswer/context-compose/dist/src/core/tools/start-context.js

Orchestrate complex AI interactions with Context Compose. A powerful CLI and server for building, validating, and managing context for large language models using the Model Context Protocol (MCP).

import { join } from 'node:path';
import * as z from 'zod';
import { StartContextToolSchema } from '../../schemas/start-context.js';
import { InvalidContextError } from '../errors.js';
import { combinePrompts, processContextSections, } from '../utils/prompt-combiner.js';
import { SecuritySchemas, defaultPathValidator } from '../utils/security.js';
import { readYamlFile } from '../utils/yaml.js';
/**
 * Reads the main context file (e.g., feature-context.yaml) asynchronously.
 */
async function readContextFile(projectRoot, contextName) {
    // Validate inputs for security
    SecuritySchemas.projectRoot.parse(projectRoot);
    SecuritySchemas.contextName.parse(contextName);
    const contextFilePath = join(projectRoot, '.contextcompose', `${contextName}-context.yaml`);
    // Validate path is within project root
    const validatedPath = defaultPathValidator.validateContextPath(contextFilePath, projectRoot);
    const context = await readYamlFile(validatedPath);
    if (context.kind !== 'context') {
        throw new InvalidContextError(`Invalid context file kind: ${context.kind}. Expected 'context'.`, { contextName, kind: context.kind });
    }
    return context;
}
/**
 * Main tool logic for starting a context.
 * It reads the specified context file and constructs the initial prompt.
 */
export async function executeStartContext(input) {
    try {
        // 1. Read the main context file
        const contextYaml = await readContextFile(input.projectRoot, input.contextName);
        // 2. Process all referenced files (personas, rules, etc.)
        const processedSections = await processContextSections(input.projectRoot, contextYaml.context, input.enhancedPrompt);
        // 3. Combine prompts into a single string
        const combinedPrompt = combinePrompts(contextYaml, processedSections, input.enhancedPrompt);
        return {
            success: true,
            message: `Context '${input.contextName}' started successfully.`,
            contextName: input.contextName,
            combinedPrompt,
            files: contextYaml.context,
        };
    }
    catch (error) {
        return {
            success: false,
            message: `Failed to start context: ${error instanceof Error ? error.message : String(error)}`,
            contextName: input.contextName,
        };
    }
}
/**
 * Validates the input arguments for the Start Context tool and executes the core logic.
 * This function serves as the main entry point for the MCP tool, ensuring
 * that the provided arguments are valid before proceeding.
 * It reads the specified context file and constructs the initial prompt.
 */
export async function executeStartContextTool(args) {
    try {
        const validatedArgs = StartContextToolSchema.parse(args);
        return await executeStartContext(validatedArgs);
    }
    catch (error) {
        if (error instanceof z.ZodError) {
            const errorMessages = error.errors.map((e) => e.message).join(', ');
            return {
                success: false,
                message: `Invalid input: ${errorMessages}`,
                contextName: args?.contextName || '',
            };
        }
        // Handle other errors (e.g., from executeStartContext)
        const errorMessage = error instanceof Error ? error.message : String(error);
        return {
            success: false,
            message: `Failed to start context: ${errorMessage}`,
            contextName: args?.contextName || '',
        };
    }
}
//# sourceMappingURL=start-context.js.map