ai.js

Go to the documentation of this file.

/**
 * @file ai.js
 * @module AIRoutes
 * @description AI/LLM-powered text processing routes. Provides text rewriting/rewording capabilities using llama.cpp HTTP API with job queuing, rate limiting, and SSE streaming support. Used for AI-assisted ticket descriptions, professional email composition, and content enhancement.
 * @see {@link ../services/llmQueue} for LLM job queue management
 * @see {@link ../services/llmSSE} for Server-Sent Events streaming
 * @see {@link ../middleware/llmCacheRateLimit} for rate limiting and caching
 * @apiDefine AIGroup AI
 * @apiGroup AI
 * @apiHeader {string} [Authorization] Bearer token (optional for some endpoints).
 * @apiError (Error 400) BadRequest No text provided.
 * @apiError (Error 429) TooManyRequests Rate limit exceeded.
 * @apiError (Error 500) ServerError LLM processing failed.
 */
 
const express = require('express');
const router = express.Router();
const { llmQueue } = require('../services/llmQueue');
const llmSSE = require('../services/llmSSE');
 
// Use llama.cpp HTTP API for rewording
const fetch = require('node-fetch');
const llmCacheRateLimit = require('../middleware/llmCacheRateLimit');
 
/**
 * @api {post} /ai/reword Reword/rewrite text using AI
 * @apiName RewordText
 * @apiGroup AI
 * @apiDescription Process text through LLM (llama.cpp) to rewrite, improve, or adapt content based on optional prompt. Jobs are queued and rate-limited per tenant. Result is cached for identical requests. Used for professional email composition, ticket description enhancement, and content improvement.
 * @apiParam {string} text Original text to reword.
 * @apiParam {string} [prompt] Optional rewriting instruction (e.g., "make it more professional", "simplify for non-technical audience").
 * @apiParam {string} [tenantId] Tenant ID for rate limiting and caching.
 * @apiSuccess {string} reworded AI-rewritten text.
 * @apiError (Error 400) BadRequest No text provided.
 * @apiError (Error 429) TooManyRequests Rate limit exceeded (10 requests per minute per tenant).
 * @apiExample {curl} Example usage:
 *     curl -X POST -d '{"text":"server not work","prompt":"make it professional"}' https://api.example.com/ai/reword
 * @apiExample {json} Response example:
 *     {
 *       "reworded": "The server is currently experiencing connectivity issues and requires immediate attention."
 *     }
 */
router.post('/reword', llmCacheRateLimit, async (req, res) => {
  const { text, prompt, tenantId } = req.body;
  console.log('[AI REWORD] Incoming request:', { text, prompt, tenantId });
  if (!text) {
    console.warn('[AI REWORD] No text provided');
    return res.status(400).json({ error: 'No text provided' });
  }
 
  // Enqueue LLM job
  const job = await llmQueue.add('reword', { text, prompt, tenantId });
  const result = await job.waitUntilFinished();
  res.json({ reworded: result.result });
});
 
router.use(llmSSE);
 
module.exports = router;