PROMETHEUS
v1.0 · LOCAL-FIRST · MIT
Personal Intelligence Platform

XANA OS

Prometheus · 3D Globe · AI Memory · OSINT · Threat Analysis
Self-hosted intelligence dashboard. Real aircraft, ships, satellites, cyber feeds, and an AI that actually remembers your conversations. All local. All free. Zero cloud.
Real Aircraft Live Vessels Satellite Tracking Memory-Linked AI OSINT Center Autonomous Recon Zero API Keys Fully Offline ChromaDB
0
Active Modules
0
Memory Vectors
0
Live Data Sources
ZERO
API Keys Needed

YOUR OWN OPS ROOM

XANA OS is a self-hosted, fully offline-capable intelligence dashboard. Think of it as your personal ops room — real aircraft overhead, ships at sea, satellites in orbit, global news, cyber threat feeds, and an AI that actually remembers your conversations.

The AI memory system is the weird part. It doesn't fine-tune any model on your data. Instead, when you ask something, it pulls the most relevant chunks from your conversation history and drops them right into the prompt. The model reasons over your actual past.

No GPU. No API keys. No data leaves your machine. Swap models in one line.

XANA-ORACLE · ACTIVE

HOW THE
MEMORY WORKS

No training. No fine-tuning. No black box. Just retrieval — fast, transparent, and yours to inspect.

01
Your Question

You ask something. Could be vague, technical, investigative — doesn't matter.

02
Vector Search · ChromaDB

Your question is embedded and compared against tens of thousands of semantic vectors from your conversation history.

03
Prompt Assembly

System prompt + memory chunks + recent history + your question — all assembled into one rich context block.

04
Local LLM · Ollama

The model reasons over your data. Entirely on your machine. Never uploaded anywhere.

GETTING
IT RUNNING

Two dependencies. Three commands. Done.

01

Prerequisites

Python 3.10+ and Ollama — the local LLM runtime. Grab it at ollama.com and pull a model.

BASH
ollama pull llama3.2    # default · ~4 GB · runs on any modern CPU
ollama pull mistral    # alternative
ollama pull qwen2.5     # lighter option
02

Clone the Repo

BASH
git clone https://github.com/sriharideveloper/xanaops.git
cd xana-os
03

One-Line Launch

BASH
./setup.sh    # creates venv, installs deps, checks Ollama
./start.sh    # launches at localhost:8501
04

Build Memory DB (Optional · Highly Recommended)

Export your ChatGPT conversations and run the pipeline to unlock ORACLE, CHRONOS, NEURAL MAP, and DOSSIER.

PIPELINE
# 1. ChatGPT → Settings → Data Controls → Export Data
# 2. Drop conversations-*.json in project root, then:

python legacy/parse_chats.py    # → master_chats.csv
python legacy/1_prep_data.py    # → paired_chats.csv
python legacy/build-brain.py   # → xana_memory_db/

./start.sh                      # restart — all memory modules online

WHAT'S
INSIDE

Eight modules. Half run immediately. The rest unlock once you've built the memory database.

01
MOD-01
GLOBE

3D live globe — real aircraft, vessel positions, satellite orbits, threat overlays. pydeck-powered.

No DB Required
02
MOD-02
ORACLE

Memory-linked AI chat grounded in your actual conversation history. Agentic commands for recon, search, weather, ops.

Requires DB
03
MOD-03
PHANTOM

Autonomous multi-source investigation. Feed a target, watch it sweep OSINT sources, cross-reference memory, build entity graphs.

Requires DB
04
MOD-04
OSINT

Live intelligence center — world news, CISA exploit feeds, IP/domain recon, crypto markets, weather. All free.

No DB Required
05
MOD-05
ARCHIVE

Raw semantic vector search across your entire conversation history. Find anything in milliseconds.

Requires DB
06
MOD-06
NEURAL MAP

3D semantic topology visualization. Your conversation history as clusters and constellations in 3D space.

Requires DB
07
MOD-07
DOSSIER

AI-generated intelligence profiles synthesized from memory. Entity analysis, pattern extraction, structured reports.

Requires DB
08
MOD-08
CHRONOS

Temporal pattern analysis. Discover how your interests and thinking have evolved over time.

Requires DB

WHAT IT'S
BUILT ON

Every component is open-source. Nothing proprietary.

Web Framework
Streamlit
Local LLM
Ollama
Vector Memory
ChromaDB
Embeddings
MiniLM-L6
3D Globe
pydeck
Charts
Plotly
Graph Analysis
NetworkX
Orbital Math
CelesTrak

TUNE IT FOR
YOUR HARDWARE

Runs on anything from a dusty laptop to a GPU rig.

🥔
POTATO
Old laptop · CPU-only · <8GB RAM
model llama3.2:1b
memory n=3 chunks
history 2 turns
context 1500 chars
MID-RANGE
Modern CPU · 8-16GB · Default
model llama3.2
memory n=5 chunks
history 4 turns
context 4000 chars
🔥
BEAST
GPU · 16GB+ VRAM
model llama3.1:8b+
memory n=10 chunks
history 8 turns
context 8000 chars

WHAT YOU
CAN SAY

Type directly in ORACLE. No button clicking.

play [song] on youtube

Opens YouTube search

google [query]

Direct Google search

open [app]

Launch local app

weather [city]

Live from Open-Meteo

ip [address]

IP geolocation & risk

lookup [domain]

Domain recon sweep

phantom [target]

Full investigation protocol

recon [target]

Quick multi-source recon

exec [command]

Safe shell (allowlisted)

status

Full system diagnostics

WHERE THE DATA
COMES FROM

All free. No sign-ups. No keys.

Source What It Provides
OpenSky Network Live aircraft transponder data — positions, altitude, speed
Finnish Digitraffic AIS Real vessel positions — Baltic Sea & global
CelesTrak Satellite TLE orbital elements — Starlink, GPS, military
CARTO Dark map tiles for the 3D globe
Google News RSS World news headlines — real-time
CISA KEV CISA Known Exploited Vulnerabilities catalog
ip-api.com IP geolocation & ASN data
CoinGecko Cryptocurrency market data
Open-Meteo Weather forecasts — any city
GDELT Geopolitical event database

YOUR DATA NEVER LEAVES YOUR MACHINE

XANA OS runs entirely local. The LLM runs via Ollama — no API calls, no telemetry, no model vendor seeing your queries. Your conversation history and vector database live in your filesystem. The only outbound traffic is read-only requests to the public data sources listed above.

The only exception is the optional Windy webcams API — even then, all you're sending is a location coordinate.

THE CREW
BEHIND THE
CONSOLE

SH
SRIHARI
@sriharideveloper
Builder · Architect · Operator

Built XANA OS from scratch. Obsessed with local-first AI, intelligence tooling, and the idea that you shouldn't need a cloud subscription to have a personal ops room. If it runs on your machine without phoning home — it's probably something Srihari would build.

Local AI OSINT Vector DBs Python Streamlit Open Source Privacy Tech LLMs ChromaDB Autonomous Agents
CREATOR OF XANA OS

SHOUTOUT TO
THE FUTURE

Two different visions. One shared belief — AI should serve humans, not the other way around.

🧠
EMPATHETIC AI · NEXT GENERATION
AVINA
// The Future of Empathetic AI

While XANA OS is your ops room for intelligence and memory, Avina is building something fundamentally different — AI that doesn't just respond, but genuinely understands. Context-aware, human-centered, and built with empathy at its core. This is where AI goes next: beyond answers, into understanding.

If XANA OS is the mind — Avina is the heart. Check it out.

Explore avina.dev