File size: 1,528 Bytes
d3c90c1
 
 
 
 
 
 
 
 
 
388ac2e
d3c90c1
388ac2e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d3c90c1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
---
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