Skip to content

GhostExodus OSINT Platform v1.1.0

Latest
Latest

Choose a tag to compare

View all tags
@Dezirae-Stark Dezirae-Stark released this 16 Apr 04:06
v1.1.0
ec9e172

GhostExodus OSINT Platform v1.1.0

Counter-extremism intelligence monitoring suite for UK law enforcement support.

All data stays on device. No cloud dependencies. Fully local AI.

What's New in v1.1.0

Custom AI Analyst Model — ghostexodus-analyst

GhostExodus now ships its own custom Ollama model purpose-built for OSINT threat analysis, replacing the generic llama3.1:8b base model.

What changed:

  • ghostexodus.Modelfile defines the custom model built on llama3.1:8b
  • Inference parameters baked in at model level — no runtime tuning required:
Parameter Value Effect
temperature 0.1 Deterministic analytical output, minimal hallucination
num_ctx 4096 Full context window for long message threads
top_k 20 Tighter token sampling for structured outputs
repeat_penalty 1.1 Suppresses repetitive phrasing in reports

  • Specialised OSINT system prompt baked into the model covering:

    • UK CONTEST/Prevent and Five Eyes analytical frameworks
    • JSON-only output discipline (no preamble, no markdown fences)
    • Formal intelligence report prose standards
    • Anti-hallucination and uncertainty handling directives

  • 5 few-shot examples baked into conversation history — the model learns correct output format before your first message:

    • CRITICAL/OPERATIONAL_PLANNING classification (pre-attack coordination)
    • NONE severity (grievance expression vs operational threat)
    • HIGH/RECRUITMENT classification (cell recruitment with OPSEC)
    • Entity extraction → valid JSON array
    • Stylometric authorship comparison → structured similarity scoring

Result: Consistently valid JSON output on first attempt across all 5 task types. No format-prompting required.

Model Setup Script — setup_models.bat

New Windows script that automates the full model installation:

  1. Checks Ollama is installed and starts the service if not running
  2. Detects if ghostexodus-analyst already exists (offers rebuild if so)
  3. Pulls llama3.1:8b base model (~4.7 GB)
  4. Runs ollama create ghostexodus-analyst -f ghostexodus.Modelfile

Run from the GhostExodus install directory or Start Menu:

Start Menu → GhostExodus → Setup AI Model

Installer Integration

The Windows installer (GhostExodus_Setup_v1.1.0.exe) now includes:

  • ghostexodus.Modelfile deployed alongside the application
  • setup_models.bat deployed alongside the application
  • Optional post-install checkbox: "Download and configure AI analyst model (~4.7 GB)" — unchecked by default, opt-in
  • Start Menu shortcut: Setup AI Model
  • Updated prerequisite warning with model setup instructions

Dependency Version Pinning

All 23 Python dependencies in requirements.txt are now pinned to exact versions matching the verified production environment, per security audit recommendation:

fastapi==0.135.3          uvicorn==0.44.0
sqlmodel==0.0.38          chromadb==1.5.7
llama-index==0.14.20      telethon==1.43.0
weasyprint==68.1          websockets==16.0
apscheduler==3.11.2       httpx==0.28.1
... (23 packages total)

This locks every fresh install to the exact build state that passed the security audit.

Automated Windows Installer Build (CI/CD)

A GitHub Actions workflow now builds and publishes GhostExodus_Setup_v*.exe automatically on every version tag push. No manual Windows machine required.

Pipeline: windows-latest runner → React frontend → PyInstaller bundle → Inno Setup → attached to release

Trigger manually from the Actions tab, or automatically by pushing a v* tag.

System Requirements

Component Requirement
OS Windows 10 build 17763+ or Windows 11 (64-bit only)
RAM 16 GB minimum
GPU NVIDIA GPU with 8 GB+ VRAM recommended (Ollama)
Storage 25 GB free (application + ghostexodus-analyst model)

Install before running:

Installation

Installer (Recommended)

  1. Install Ollama
  2. Run GhostExodus_Setup_v1.1.0.exe as Administrator
  3. At end of install, optionally tick "Download and configure AI analyst model"
    — or run Start Menu → GhostExodus → Setup AI Model at any time
  4. Edit %AppData%\GhostExodus\.env with your Telegram credentials
  5. Launch GhostExodus

From Source

git clone https://github.com/Dezirae-Stark/CT-OSINT-AI-Tools.git
cd CT-OSINT-AI-Tools
python -m venv venv && venv\Scripts\activate
pip install -r requirements.txt
cd frontend && npm install && npm run build && cd ..
xcopy /E /Y /I frontend\dist backend\static
copy .env.example .env
setup_models.bat
ollama pull nomic-embed-text
cd backend && uvicorn main:app --host 127.0.0.1 --port 8000

Upgrade from v1.0.0

  1. Pull latest: git pull origin main
  2. Run model setup: setup_models.bat (one-time, ~4.7 GB download)
  3. Restart GhostExodus — the platform will automatically use ghostexodus-analyst

No database migrations required. Existing data is fully compatible.

Full Changelog

Commit Change
ec9e172 fix: correct Inno Setup OutputDir path
c687822 fix: correct .env.example source path in ghostexodus.iss
119aeca ci: add GitHub Actions Windows installer build pipeline
305e554 chore: bump installer AppVersion to 1.1.0
82a7228 docs: update README for ghostexodus-analyst custom model
2fb0650 Add custom ghostexodus-analyst Ollama model + installer integration
5754255 Pin all dependency versions to known-good installed state

Full diff v1.0.0...v1.1.0

Built with Claude Code

Assets 3
Loading