Rebuild Aurora CRM intelligence stack

.env.example

@@ -0,0 +1,7 @@
+FLASK_APP=app
+FLASK_ENV=development
+OLLAMA_HOST=http://localhost:11434
+LLM_MODEL=llama3.2:latest
+LLM_CONTEXT=4096
+LLM_MAX_TOKENS=768
+LLM_TEMPERATURE=0.2

.gitignore

@@ -0,0 +1,12 @@
+.venv/
+__pycache__/
+*.pyc
+instance/
+.env
+uploads/*
+!uploads/.gitkeep
+*.log
+.DS_Store
+node_modules/
+client/node_modules/
+server/node_modules/

README.md

@@ -1 +1,128 @@
-# beta
+# Aurora CRM Intelligence Suite
+
+Aurora CRM Intelligence Suite is a local-first command center for predictive strategy, risk analytics, AI advisory chat, and deep document comparison. Every insight runs on your workstation through the [Ollama](https://ollama.com) runtime, so nothing ever leaves your RTX 4090 (or compatible) GPU.
+
+> **Security reminder**
+> No cloud API keys are required. Keep your model checkpoints private and stored locally.
+
+## Project structure
+
+```
+.
+├── app/                  # Flask application package
+│   ├── services/         # Local inference + file helper utilities
+│   ├── static/           # Custom CSS and interaction scripts
+│   └── templates/        # Multi-page Jinja UI
+├── uploads/              # Local document catalog (created automatically)
+├── app.py                # Flask entry point
+├── requirements.txt      # Python dependencies (Flask + requests)
+├── install.ps1           # Windows installer shim (PowerShell)
+├── run.ps1               # Windows auto runner
+├── install.bat           # Convenience launcher that calls install.ps1
+├── run.bat               # Convenience launcher that calls run.ps1
+└── README.md             # You are here
+```
+
+## Prerequisites
+
+- Python **3.10+** (tested with CPython 3.12)
+- PowerShell 5.1+ (bundled with Windows 11) for the automation scripts
+- NVIDIA GPU with recent drivers (4090 recommended) or enough CPU cores if you prefer CPU inference
+- [Ollama for Windows](https://ollama.com/download/windows) installed and running (`ollama serve` starts automatically once installed)
+- A local **GGUF** checkpoint available to Ollama (use `ollama pull llama3.2` or `ollama pull qwen2.5-coder` etc.)
+
+## Quick start (Windows 11)
+
+1. Install [Ollama](https://ollama.com/download/windows) if you have not already, then open **PowerShell** and pull a chat-capable model (for example Meta's Llama 3.2):
+   ```powershell
+   ollama pull llama3.2
+   ```
+   Ollama automatically stores the GGUF checkpoint in its managed directory and serves it to local clients.
+2. Open **PowerShell** and `cd` into the project directory.
+3. If this is your first time running local scripts, allow the session to run them:
+   ```powershell
+   Set-ExecutionPolicy -Scope Process RemoteSigned
+   ```
+4. Run the installer:
+   ```powershell
+   ./install.ps1
+   ```
+   The script will:
+   - Detect Python 3
+   - Create (or reuse) a `.venv` virtual environment
+   - Install dependencies from `requirements.txt`
+5. Create your `.env` file and set the Ollama options:
+   ```powershell
+   Copy-Item .env.example .env
+   notepad .env
+   ```
+   Update `LLM_MODEL` with the model name you pulled (for example `llama3.2:latest`). Leave `OLLAMA_HOST` at the default unless you run Ollama on a different machine.
+6. Launch the stack:
+   ```powershell
+   ./run.ps1
+   ```
+   The runner activates the virtual environment, exports variables from `.env`, verifies the model path, and starts Flask in debug mode at <http://127.0.0.1:5000/>.
+
+You can double-click `install.bat` or `run.bat` from Explorer—they simply relay to the PowerShell scripts with execution policy bypassed.
+
+## Quick start (macOS/Linux)
+
+```bash
+python3 -m venv .venv
+source .venv/bin/activate
+pip install -r requirements.txt
+cp .env.example .env  # edit with your Ollama host and model
+export OLLAMA_HOST="http://localhost:11434"
+export LLM_MODEL="llama3.2:latest"
+flask --app app --debug run
+```
+
+Visit <http://127.0.0.1:5000/> once the server reports it is running.
+
+## Feature tour
+
+- **Overview dashboard** – animated chrome with catalog health, workflow readiness, and immersive storytelling.
+- **Data steward** – upload TXT, MD, CSV, or JSON files and request local AI summaries with quality scoring, tags, and opportunity prompts.
+- **Predictive strategy** – craft revenue scenarios, score growth outlooks, and review initiative guidance with progress dials and impact distributions.
+- **Risk command** – quantify exposure, inspect mitigation cards, and monitor early warnings with visual coverage bars.
+- **Advisor lounge** – chat with the embedded strategist, auto-load document summaries, and keep every conversation on device.
+- **Comparison lab** – run deep comparisons between two documents to expose alignment, divergence, and actionable recommendations.
+
+Each workflow uses structured prompts defined in `app/services/insights.py`. Responses are parsed as JSON and rendered dynamically with a custom glassmorphism UI—no external CDNs required.
+
+## Configuration notes
+
+- `OLLAMA_HOST` – Base URL where Ollama listens (default `http://localhost:11434`).
+- `LLM_MODEL` – **required**. Model name as listed by `ollama list` (for example `llama3.2:latest`).
+- `LLM_CONTEXT` – Context window supplied to Ollama (default `4096`). Increase if your model supports a larger context.
+- `LLM_MAX_TOKENS` – Maximum tokens to generate per response (default `768`).
+- `LLM_TEMPERATURE` – Sampling temperature (default `0.2`).
+- Uploaded documents remain on disk inside the local `uploads/` folder. Delete files there to clear the catalog.
+
+## Automation scripts
+
+The Windows automation lives under `scripts/windows/` and is intentionally PowerShell-only—no external tooling required.
+
+- `install.ps1` – ensures Python is present, provisions `.venv`, and installs dependencies.
+- `run.ps1` – verifies the environment, loads `.env`, reminds you if Ollama variables are unset, and starts Flask using the virtual environment's interpreter.
+
+Both scripts print friendly diagnostics (missing Python, missing model, etc.) so issues are easy to resolve.
+
+## Validation checklist
+
+- Flask routes use built-in templating; no frontend framework is required.
+- All inference stays local by calling the Ollama HTTP API on `OLLAMA_HOST`; there are **no** outbound internet calls from the Flask app.
+- File uploads are limited to text-based formats and capped at 10 MB per file to keep processing light.
+- UI assets rely entirely on bundled CSS/JS with layered gradients, animated nav, and custom progress visuals.
+
+## Troubleshooting
+
+| Symptom | Resolution |
+| --- | --- |
+| "Local inference is not configured" error | Launch Ollama (it starts automatically after installation), run `ollama pull <model>`, and set `LLM_MODEL` in `.env`. |
+| App loads but responses are empty | Confirm the pulled Ollama model supports chat instruction tuning and that `LLM_CONTEXT` is within the model's limit. |
+| Slow responses on Windows | Verify Ollama is using your GPU (run `ollama run llama3.2 "test"` in a console and watch GPU usage). |
+| PowerShell blocks the scripts | Run PowerShell as Administrator once and execute `Set-ExecutionPolicy -Scope Process RemoteSigned`. |
+| Installer reports Python not found | Install Python 3.10+ from <https://www.python.org/downloads/> and tick "Add to PATH" during setup. |
+
+Enjoy orchestrating predictive CRM intelligence locally!

app.py

@@ -0,0 +1,6 @@
+from app import create_app
+
+app = create_app()
+
+if __name__ == "__main__":
+    app.run(debug=True)

app/__init__.py

@@ -0,0 +1,18 @@
+from flask import Flask
+from pathlib import Path
+
+from dotenv import load_dotenv
+
+
+def create_app() -> Flask:
+    load_dotenv()
+    app = Flask(__name__)
+    app.config["UPLOAD_FOLDER"] = Path(app.root_path).parent / "uploads"
+    app.config["UPLOAD_FOLDER"].mkdir(parents=True, exist_ok=True)
+    app.config["MAX_CONTENT_LENGTH"] = 10 * 1024 * 1024  # 10 MB per file
+    app.config["SECRET_KEY"] = app.config.get("SECRET_KEY") or "dev-secret-key"
+
+    from . import routes  # noqa: WPS433
+
+    routes.register(app)
+    return app