Spaces:
Runtime error
Runtime error
| title: Blocks AI Assistant | |
| emoji: 🏢 | |
| colorFrom: red | |
| colorTo: red | |
| sdk: gradio | |
| sdk_version: 5.49.1 | |
| app_file: app.py | |
| pinned: false | |
| # Blocks Assistant – Hybrid AI + Analytics | |
| This Space hosts a fully self-contained AI Operations Assistant built for backlog mitigation, | |
| block analysis, and fiscal-quarter trend detection. | |
| ### Key Capabilities | |
| - 📊 Block summary (Pricing, Credit, Delivery, Ship Debit) | |
| - 🧮 Customer exposure analytics | |
| - 🔁 Dynamic FAQ learning based on user queries | |
| - 🧱 Multiple block order detection | |
| - 📅 Fiscal-year quarter-wise trend insights (Apr–Mar FY) | |
| - 🧠 Phi-2 LLM rewriting layer for polished operational insights | |
| - 🔒 Zero external APIs – all analytics & AI happen inside the Space | |
| ### How it works | |
| 1. Upload your backlog file: | |
| **Backlog_updated.xlsx** (required) | |
| 2. The assistant computes analytics: | |
| - Block exposure | |
| - Order-level details | |
| - Highest exposure customers | |
| - Multi-block orders | |
| - FY quarter trends | |
| 3. The LLM refines the insights: | |
| - Business-professional tone | |
| - Zero hallucinations (numbers are not changed) | |
| - Concise operations-style recommendations | |
| ### Technology Stack | |
| - Flask backend | |
| - Pandas for analytics | |
| - Transformers + Torch (Phi-2 model) | |
| - HTML + JS + CSS frontend (custom UI) | |
| - Runs fully on CPU | |
| ### No external API calls | |
| - No OpenAI | |
| - No remote LLMs | |
| - No data leaves Hugging Face | |
| Designed for operational reliability. | |
| Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference | |