Two-Stage AI Brain — smarter than RAG alone  

PLX‑AILM

Your files. Your AI. Zero hallucination.
Upload knowledge, get a personal API — query answers that come only from your data.

POST https://api.ailm.primeloopx.xyz/api/query
Content-Type: application/json

{
  "api_key": "plx_your_key_here",
  "query":   "WHO ARE YOU"
}

→ "We are PrimeLoopX, building AI for innovators..."
Powered by PrimeTriple4 🔍 FAISS Vector Search ⚡ FastAPI 🔐 SHA-256 Key Security 🐘 PHP 8.3 + Node 18 📄 Your .txt Files Only
AI reasoning stages per query
0%
Hallucination — answers from your data only
3B
Qwen 2.5 parameter model on-premise
SHA‑256
API key stored as cryptographic hash

Why PLX-AILM?

Not generic AI. Your data. Your answers. Your edge.

Real-Time Query Tracking

Monitor every query, usage count, and monthly reset date from your dashboard — always in sync with your subscription.

↑ Accuracy ↓ Noise

Zero Hallucination

Qwen only reads your indexed files. No internet. No guessing. Every answer is grounded in data you uploaded.

Private by Design

Your files stay isolated per user. API keys are SHA-256 hashed and shown exactly once. The AI engine is never public-facing.

💰 Cost Effective ⊕ Real-Time Insights ⚙ Per-User Isolation ✦ FAISS Vector Search 🚀 Instant API Access 🛡 No Hallucination

From File to Answer in Four Steps

Simple, powerful, secure — live in minutes

Register & Activate

Create your account. Contact admin to activate your subscription plan (Basic, Standard, or Advanced).

Upload Knowledge Files

Upload .txt files — product docs, FAQs, lore, policies. Each is indexed using FAISS vector search.

Generate API Key

Generate your personal API key — shown once, stored as SHA-256 hash. Use it in any app or script.

Query & Get Answers

POST your question with your API key. The Two-Stage Brain picks the right file and chunk, then answers.

Qwen Thinks Twice Before Answering

Most RAG systems retrieve then answer. PLX-AILM adds two reasoning stages before the final response — eliminating wrong-file and wrong-chunk errors.

S1

Stage 1 — File Priority Brain

Before reading anything, Qwen receives your query and just the list of your filenames. It ranks each file by semantic relevance (0.0–1.0) — so FAISS searches the right file first, not a random match.

Example: "Who are you?" → about_us.txt scores 0.95, pricing.txt scores 0.10

S2

Stage 2 — Chunk Disambiguation Brain

FAISS returns multiple similar chunks. Qwen reads each chunk and picks the one that matches the intent of your query — not just the similarity score. No more contextually wrong answers.

Example: Two chunks both mention "who we are" — Qwen picks the one describing identity, not product features.

Full Pipeline
Your Query Stage 1: File Ranking FAISS Vector Search Stage 2: Chunk Pick Final Answer

Simple Pricing, No Surprises

All plans activated by admin. No payment gateway — direct activation.

Save 30%
Basic
$12 /month

Perfect for individuals exploring private AI on their own knowledge base.

Get Started  ↗
  • 100 queries / month
  • FAISS + Two-Stage Brain
  • Real-Time Usage Dashboard
  • SHA-256 API Key Security
Advanced
$97 /month

For companies deploying AI at scale with high query volume and priority support.

Get Started  ↗
  • 500 queries / month
  • Priority AI Processing
  • Unlimited File Uploads
  • Advanced Reporting & Logs
  • Priority Admin Support
🤝 We donate 2% of every membership to pediatric wellbeing initiatives

Frequently Asked Questions

Only .txt files are accepted. This keeps indexing fast and deterministic. Each file is chunked and embedded using sentence-transformers all-MiniLM-L6-v2, then stored in your personal FAISS index.
Yes. Each user has a completely isolated knowledge folder on the server. The Python AI engine runs at localhost:8001 — it is never exposed to the internet. Your API key is stored only as a SHA-256 hash, never in plain text.
Standard RAG systems retrieve chunks and send them directly to the LLM. PLX-AILM adds two extra Qwen reasoning passes: Stage 1 ranks which of your files is most relevant (preventing cross-file noise), and Stage 2 selects the best chunk by semantic intent (preventing wrong-context answers). The result is dramatically more accurate responses.
Register your account, then contact admin at admin@primeloopx.xyz. There is no automated payment gateway — activation is manual, which means direct support and instant confirmation.
Queries return a 403 with a clear message. Your limit resets automatically 30 days from your activation date — no manual intervention needed. You can track remaining queries live in your dashboard.
Yes. It is a standard POST JSON endpoint — usable from Python, Node.js, PHP, cURL, Postman, or any HTTP client. As long as you pass your api_key and query fields, you'll get an answer back.

Ready to give your AI a brain?

Register now, upload your first knowledge file, and make your first query — in under 5 minutes.

Create Free Account  ↗

Questions? Email us at admin@primeloopx.xyz