AI Sales Intelligence Dashboard

A multi-page web dashboard that turns Sales.xlsx and Prices.xlsx into plain-English business recommendations. Built by the Society for AI in Enterprise Systems (SAIES) at California State University, Los Angeles.

Live demo: https://saies-sales-ai-dashboard.onrender.com/

Features

• Overview — KPI summary, recommended product focus, and sales trend chart
• Product Insights — Per-product status (Healthy, Growth Opportunity, Needs Attention) with filters and search
• Forecast — Next-month revenue, profit, and unit predictions based on historical sales
• Price Impact — Slider-based simulator for testing price change scenarios
• Ask Sales AI — Plain-English Q&A grounded in the uploaded data, powered by Gemini with a Groq fallback
• Carbon Impact — CO2e emissions breakdown by type, scope, carrier, and time period
• Data Upload — Upload your own Sales, Prices, and Carbon Emissions files; preview the cleaned and matched data. The app auto-detects the data format and currency.

Data warehouse

The live dashboard reads from saies_warehouse.db, a SQLite warehouse rebuilt by:

python etl_star_schema.py

The ETL builds a star schema with explicit primary keys, foreign keys, and indexes on every fact-table foreign key. dim_date includes BI-friendly calendar attributes such as year, quarter, month name, week number, day of week, weekend flags, and period-end flags. Carbon emissions are loaded from Carbon Emissions.xlsx into fact_carbon during the same build, with emission type and scope normalized into dimensions.

SQLite is used for the local and Render demo because it ships as a single file with the app. MySQL design artifacts remain in mysql/ for the advisor-facing ERD/schema package; use mysql/saies_warehouse_schema.sql if the deployment needs to be moved to a MySQL server.

Advisor-facing warehouse documentation:

• docs/warehouse_erd.md
• docs/data_dictionary.md
• docs/validation_report.md
• mysql/saies_warehouse_schema.sql
• mysql/saies_warehouse_model.mwb
• mysql/saies_warehouse_erd_all_diagrams.pdf

Optional MySQL load path:

pip install -r requirements-mysql.txt
python etl_star_schema.py
python scripts/load_mysql_warehouse.py --user root --password your_mysql_password

The MySQL loader recreates the saies_warehouse schema from mysql/saies_warehouse_schema.sql, copies the rows from saies_warehouse.db, and prints validation totals. The website still uses SQLite unless you intentionally change the runtime configuration.

Prerequisites

• Python 3.9 or later
• The following packages:

pip install pandas openpyxl requests

Setup

1. Clone the repository:

git clone https://github.com/bta5-csula/saies-control-tower.git
cd saies-control-tower

2. Create a .env file in the project root with your API keys:

GEMINI_API_KEY=your_gemini_key_here
GROQ_API_KEY=your_groq_key_here

The app works without API keys — keyword-based answers will still run. The AI chat falls back to Groq if Gemini is unavailable, and to keyword answers if neither key is set.

3. Run the server:

python server.py

Then open http://127.0.0.1:8000 in your browser.

Data format

The app auto-detects the uploaded file format. Two schemas are supported out of the box:

File	ERPSim format	LAX / general format
Sales	MATERIAL_NUMBER, SIM_CALENDAR_DATE, QUANTITY, NET_VALUE, COST	Date, Material/Product, Quantity, Revenue, Cost
Prices	MATERIAL_NUMBER, SIM_CALENDAR_DATE, PRICE	Date, Material/Product, Price_USD_per_kg, Distribution_Channel
Carbon (optional)	TYPE, SUBTYPE, SCOPE, TOTAL_CO2E_EMISSIONS	Scope_1/2/3_CO2e_kg columns per row

The app ships with default sample data. Use the Data Upload page to replace it with your own files for the current session. Additional sample datasets are in sample_data/ for testing.

Deployment

The app is designed to run on Render as a Python web service:

• Build command: pip install pandas openpyxl requests
• Start command: python server.py
• Environment variables: Set GEMINI_API_KEY and GROQ_API_KEY in the Render dashboard

Render injects a PORT environment variable automatically; the server reads it on startup.