import { z } from "zod";
import { Tool } from "../types.js";
import { generateLLMInstructions } from "../../llm-instructions.js";
export const getInstructionsTool: Tool = {
name: "get_instructions",
description: "Get instructions for optimal MCP usage, including capabilities, best practices, and example workflows",
inputSchema: {
type: "object",
properties: {
topic: {
type: "string",
description: "Specific topic for instructions (optional)",
enum: ["search", "synthesis", "retrieval", "general"],
},
includeExamples: {
type: "boolean",
description: "Include workflow examples",
default: true,
},
},
},
modifies: false,
handler: async (client, args) => {
const schema = z.object({
topic: z.enum(["search", "synthesis", "retrieval", "general"]).optional(),
includeExamples: z.boolean().optional().default(true),
});
const params = schema.parse(args);
const instructions = generateLLMInstructions();
let responseText = `# Noverload MCP Usage Instructions\n\n`;
responseText += `## Version: ${instructions.version}\n\n`;
if (!params.topic || params.topic === "general") {
// Show all capabilities
responseText += `## Capabilities Overview\n\n`;
for (const [key, capability] of Object.entries(instructions.capabilities)) {
responseText += `### ${key.charAt(0).toUpperCase() + key.slice(1)}\n`;
responseText += `${capability.description}\n\n`;
responseText += `**Strengths:**\n`;
capability.strengths.forEach((s: string) => responseText += `- ${s}\n`);
responseText += `\n**When to use:**\n`;
capability.whenToUse.forEach((w: string) => responseText += `- ${w}\n`);
responseText += `\n`;
}
responseText += `## Best Practices\n\n`;
for (const [scenario, practice] of Object.entries(instructions.bestPractices)) {
responseText += `**${scenario.replace(/([A-Z])/g, ' $1').toLowerCase()}:** ${practice}\n`;
}
responseText += `\n`;
} else {
// Show specific capability
const capability = instructions.capabilities[params.topic];
if (capability) {
responseText += `## ${params.topic.charAt(0).toUpperCase() + params.topic.slice(1)} Capability\n\n`;
responseText += `${capability.description}\n\n`;
responseText += `### Strengths\n`;
capability.strengths.forEach(s => responseText += `- ${s}\n`);
responseText += `\n### Limitations\n`;
capability.limitations.forEach(l => responseText += `- ${l}\n`);
responseText += `\n### When to Use\n`;
capability.whenToUse.forEach(w => responseText += `- ${w}\n`);
}
}
if (params.includeExamples && instructions.exampleWorkflows.length > 0) {
responseText += `\n## Example Workflows\n\n`;
instructions.exampleWorkflows.forEach((example, idx) => {
responseText += `### ${idx + 1}. ${example.scenario}\n`;
responseText += `**Steps:**\n`;
example.steps.forEach((step, i) => responseText += `${i + 1}. ${step}\n`);
responseText += `**Expected Outcome:** ${example.expectedOutcome}\n\n`;
});
}
responseText += `## Token Management\n\n`;
responseText += `- Search results limit: ${instructions.tokenManagement.searchResultsLimit}\n`;
responseText += `- Use chunking: ${instructions.tokenManagement.useChunking ? 'Yes' : 'No'}\n`;
responseText += `- Get summaries first: ${instructions.tokenManagement.summaryFirst ? 'Yes' : 'No'}\n`;
responseText += `- Max content per query: ${instructions.tokenManagement.maxContentPerQuery.toLocaleString()} tokens\n`;
return {
content: [
{
type: "text",
text: responseText,
},
],
data: instructions,
};
},
};
