Mipler - Local OSINT Investigation Wall

A fully local, offline-first, Ollama-powered OSINT investigation canvas.
No API keys. No cloud. No data leaves your machine.

---

What's New in v2

Change	Details
Ollama only	OpenAI and Custom API options have been completely removed. The app uses Ollama exclusively — free, local, private.
File Analysis Panel	Upload any file type (JSON, CSV, TXT, PDF, images, etc.), ask a natural-language question, and get an AI answer plus a downloadable mindmap JSON.
Mindmap export	After each file analysis, the AI generates a structured mindmap. Click ↓ JSON to download it for external use or import into other tools.
Two separate workspaces	"Workspace A" and "Workspace B" are completely independent canvases. Importing a JSON into one workspace does not affect the other.
Clear only active workspace	"Clear Workspace" now only clears whichever workspace is currently active — the other is untouched.
start.sh	One-command launcher: checks Node.js version, installs deps, builds if needed, checks Ollama, and starts the server bound to localhost only.
Security	Server is hardcoded to bind 127.0.0.1 — it cannot be reached from other machines on your network.

---

Requirements

• Node.js v18 or higher — nodejs.org
• Ollama — ollama.com
• A pulled Ollama model (e.g. llama3, mistral, gemma2)

No API keys. No internet connection required after initial model download.

---

Quick Start

# 1. Clone or unzip Mipler
cd mipler-main

# 2. Start Ollama in a separate terminal (with CORS enabled)
OLLAMA_ORIGINS=* ollama serve

# 3. Pull a model (first time only)
ollama pull llama3

# 4. Start Mipler
bash start.sh

Then open your browser at http://localhost:3000

---

What start.sh Does

The startup script handles everything automatically:

1. Checks Node.js version — exits with a clear message if below v18.
2. Installs dependencies — runs npm install only if node_modules is missing.
3. Builds the app — runs npm run build only if dist/ is missing or source files are newer.
4. Checks Ollama — warns (but does not block) if Ollama is not running. Lists available models.
5. Starts the server — binds to 127.0.0.1:3000 (localhost only).

To use a different port:

PORT=8080 bash start.sh

---

Features

Investigation Canvases (Two Separate Workspaces)

Mipler starts with two workspaces: Workspace A and Workspace B. They are completely independent:

• Each workspace has its own nodes, edges, and viewport.
• Importing a JSON file into Workspace A does not merge into Workspace B.
• "Clear Workspace" only clears the currently active one.
• You can add more workspaces with + New Investigation in the workspace picker.

Switching workspaces: Click the investigation name in the top-left toolbar to open the workspace picker.

Cards You Can Add

Button	Card Type	Description
N	Note	Free-text note card
Img	Image	Paste or embed an image
PDF	PDF	Embed a PDF document
WH	WHOIS	WHOIS domain lookup
DNS	DNS	DNS record lookup
RI	Reverse Image	Reverse image search
OS	OSINT Framework	OSINT methodology viewer
URL	Custom URL	Embed any web tool in an iframe

AI File Analysis Panel

Click AI in the toolbar to open the File Analysis Panel.

1. Upload files — drag-and-drop or click to browse. All file types accepted:
   • JSON, CSV, TXT, MD, LOG, XML, YAML — read as text
   • PDF, images — metadata used; content summarized
   • Any other file — best-effort text read
2. Ask a question — type what you want to know from the files.
3. Analyze — Mipler sends the files and question to your local Ollama model.
4. Get results:
   • Answer — AI's written response to your question.
   • Mindmap — Interactive collapsible mindmap tree based on the analysis.
   • ↓ JSON — Download the full mindmap + answer as a structured JSON file.

The downloaded JSON format:

{
  "answer": "AI's written answer...",
  "mindmap": {
    "root": "Your question",
    "nodes": [
      {
        "id": "1",
        "label": "Main finding",
        "children": [
          { "id": "1-1", "label": "Sub-point", "children": [] }
        ]
      }
    ]
  }
}

API Workspace (Ollama Chat)

Click API in the toolbar to open the Ollama chat workspace. This is a direct, conversation-style interface to your local Ollama model — useful for freeform OSINT queries, brainstorming, or research.

• Chat tab — send messages and get responses from Ollama.
• Settings tab — configure the Ollama URL and select from available models (auto-detected).

Ollama Settings

Click the ⋮ menu → Ollama Settings to configure:

• Base URL — default http://localhost:11434
• Model — default llama3; type any model name or choose from detected ones
• Test Connection — verifies Ollama is reachable

Export & Import

• Export — saves the current workspace state as a .json file (all investigations, nodes, edges, AI history).
• Import — loads a workspace JSON. It loads into the current active workspace only — it does not overwrite other workspaces.
• AI Mindmap export — separate JSON download from the File Analysis Panel.

---

Security

• Localhost only — The server is bound to 127.0.0.1. It cannot be accessed from other devices.
• No outbound network calls — All AI requests go to your local Ollama instance. No data is sent to any cloud.
• No API keys stored — The v2 app does not ask for or store any API keys.
• No persistent storage — Workspace data lives in browser memory only. It is cleared when you close or refresh the tab. Use Export to save your work.
• No telemetry — Mipler does not collect, log, or transmit any usage data.

---

Ollama Setup Guide

Install Ollama

# macOS / Linux
curl -fsSL https://ollama.com/install.sh | sh

# Windows
# Download installer from https://ollama.com/download

Start with CORS enabled (required for browser access)

OLLAMA_ORIGINS=* ollama serve

Pull a model

ollama pull llama3      # Good general model (~4GB)
ollama pull mistral     # Fast, good quality (~4GB)
ollama pull gemma2      # Google's model (~5GB)
ollama pull phi3        # Lightweight (~2GB)

Verify it's working

curl http://localhost:11434/api/tags

---

Development

# Install dependencies
npm install

# Start dev server (hot-reload)
npm run dev

# Build for production
npm run build

# Serve production build
npm start
# or
bash start.sh

---

Example images

---

Project Structure

mipler-main/
├── start.sh                        # One-command launcher
├── server.js                       # Express server (localhost only)
├── src/
│   ├── App.tsx                     # Root component
│   ├── components/
│   │   ├── FileAnalysisPanel.tsx   # NEW: file upload + AI analysis + mindmap
│   │   ├── ApiWorkspace.tsx        # Ollama chat workspace
│   │   ├── Toolbar.tsx             # Top toolbar
│   │   ├── WallCanvas.tsx          # ReactFlow canvas
│   │   ├── MiplerCardNode.tsx      # Card node renderer
│   │   ├── cards/                  # Individual card components
│   │   ├── edges/                  # Edge components
│   │   ├── modals/
│   │   │   ├── ApiSettingsModal.tsx  # Ollama settings (renamed)
│   │   │   ├── ExportModal.tsx
│   │   │   ├── ImportModal.tsx
│   │   │   └── ...
│   │   └── ui/                     # Button, Input, Modal
│   ├── store/
│   │   └── useWorkspaceStore.ts    # Zustand store (Ollama-only, 2 workspaces)
│   ├── types/
│   │   └── index.ts                # TypeScript types
│   └── utils/
│       ├── aiApi.ts                # Ollama API calls + file analysis
│       ├── fileSystem.ts
│       └── osintApi.ts
└── README.md

---

Troubleshooting

"Ollama not detected"

• Run OLLAMA_ORIGINS=* ollama serve in a terminal.
• Make sure the URL in Ollama Settings matches where Ollama is running.

"Ollama error 404"

• The model you specified may not be pulled yet. Run ollama pull llama3.

"AI analysis returns no mindmap"

• Try a smaller, faster model like phi3 or mistral.
• Make sure your question is specific enough for the model to structure a response.

Workspace data lost after refresh

• This is by design — all data is in-memory. Use Export before closing the tab.

Port already in use

• Run PORT=8080 bash start.sh