Knowledge OS — Turning Any File into Instant, Cited Answers with Redis 8

This is a submission for the Redis AI Challenge: Real-Time AI Innovators.

Knowledge OS — From Data to Decisions in Seconds
Submission for the Redis AI Challenge

🚀 Turn messy PDFs, Word files, spreadsheets, and URLs into instant, cited answers — powered by Redis 8 + AI agents.
No endless searching. No manual note-taking. Just drop, ask, and decide.

What I Built:
Knowledge OS is an AI-powered command center for documents.
Drop in PDFs, DOCX, XLSX, images, or even URLs—our AI agent swarm ingests, cleans, summarizes, and indexes everything so you can ask natural questions and get cited answers in seconds.

Typical asks

“Summarize all invoices over $5,000.”

“What’s the refund policy in this contract?”

“Key points and conclusion from this article URL.”

Every answer links back to the exact source passage. ⚡️

Demo
🔴 Live: (Netlify): https://knowledgeosdemo.netlify.app/

🎥 Video: https://youtu.be/Lys6WacZxTc

📷 Screenshots: N/A

Dashboard & file drop

Agents in action (Ingest → OCR → Summarize → Index)

Redis Cloud: Streams + Vector Search

Smart chat with citations

Why This Matters
Teams waste hours hunting through PDFs and tabs. Knowledge OS turns that into seconds with reliable, cited answers—great for audits, ops, research, and finance workflows.

How I Used Redis 8
Redis Cloud v8 is the real-time backbone:

Streams – Orchestrates AI agents
ingest → ocr → embed → index → answer

Vector Search – Embedding-based retrieval across all pages and URLs

RedisJSON – Rich metadata (title, dates, vendor, totals, tags)

Semantic/summary caching – Sub-10ms repeat answers and table rollups

This combo gives me low-latency answers with source citations at interactive speeds.

Architecture (High-Level)
pgsql
Copy
Edit
Upload/URL
│
▼
Ingest Agent ──► OCR/Parser ──► Chunk & Embed ──► Indexer
│ │ │ │
└──► Redis Streams (task handoffs) │ │
▼ ▼
Redis Vector RedisJSON (metadata)
Index
│
User Chat ──► Retriever ──► LLM (with citations) ──► Answer + Source links
▲ └─► Cache (Redis) for repeats
└─────────────► Metrics / Logs
Tech Stack
Frontend: React / Next.js (demo UI), CapCut for demo video

Agents/Backend: Node/Python, queues via Redis Streams

Search: Redis Vector Search (OpenAI/all-MiniLM embeddings)

Storage/Metadata: RedisJSON

Hosting: Netlify (demo), Redis Cloud (data layer)

How to Try It (Local)
1) Environment
bash
Copy
Edit
export REDIS_URL="rediss://:@:"
export OPENAI_API_KEY= # or your LLM provider
2) Install + run
bash
Copy
Edit
npm install
npm run dev
3) In the app
Drop a PDF/Doc/URL

Ask a question

Click citations to jump to source
What’s Next
Role-based redaction (PII hiding) before indexing

Multi-doc table extraction → CSV export

Org spaces & SSO

Fine-tuned domain prompts

Q&A are most welcomed! Ask away!!

Team / Credits
Solo build by: Johnathan Jake @jjake486@gmail.com
Thanks, Redis team & judges!