Agentic tool for statistical process control

A Quality Management System for Statistical Process Control (SPC), Measurement System Analysis (MSA), and Process Capability — powered by agents.

---

Overview

Quality engineers spend too much time clicking through interfaces and manually interpreting charts. SPC Quality AI automates the tedious parts of quality analysis just upload your data and ask questions in plain English.

No more:

• Guessing which control chart to use
• Manually calculating control limits
• Repeating the same MSA interpretations

Instead, the AI agents handle:

• Automatic detection of the correct chart type
• Identification of out-of-control points
• Plain-language interpretation of results
• Professional report generation

They’re not perfect, but they’re remarkably efficient assistants for modern quality engineers.

---

Features

Statistical Process Control (SPC)

• Control Charts: I-MR, Xbar-R, Xbar-S, P, NP, C, and U charts
• Automatic Chart Selection based on dataset characteristics
• Out-of-Control Detection with statistical rules
• Data Quality Validation before analysis
• Interactive Visualizations via Plotly

Measurement System Analysis (MSA)

• Gage R&R Studies (ANOVA method)
• Bias Studies with significance testing
• Linearity Studies for accuracy verification
• Stability Studies for long-term consistency
• Comprehensive MSA Reports with interpretation

Process Capability Analysis

• Capability Indices: Cp, Cpk, Pp, Ppk, Cpm
• Normality Testing: Anderson–Darling, Shapiro–Wilk, Kolmogorov–Smirnov
• Process Centering and Variation Analysis

---

Quick Start

1. Installation

# Clone the repository
git clone https://github.com/M1ndSmith/ASPC.git
cd ASPC

# Create and activate virtual environment
python -m venv aspcvenv
source aspcvenv/bin/activate  # On Windows: aspcvenv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Set up your API key
cp env.example .env
# Edit .env and add your GROQ_API_KEY (or key for your chosen LLM)

2. Configure LLM (Optional)

The system uses Groq by default (free tier available). To switch providers, edit agent_config/config.yaml:

llm:
  provider: "groq"        # or "openai", "anthropic"
  model: "llama-3.1-8b-instant"

3. Start the API Server

uvicorn api.main:app --host 0.0.0.0 --port 8000

Server available at: http://localhost:8000

---

Using the CLI

The CLI (./spc) provides the easiest way to interact with the agents. It automatically activates the virtual environment.

Basic Commands

# Check server health
./spc health

# Run control chart analysis
./spc chat control-charts "Analyze this data" -f your_data.csv

# Run MSA study
./spc chat msa "Run a Gage R&R study" -f measurement_data.csv

# Run capability analysis
./spc chat capability "Calculate Cp and Cpk with USL=10.5, LSL=9.5" -f process_data.csv

Available Agents

Agent	Purpose
control-charts	Control chart and SPC analysis
msa	Measurement System Analysis
capability	Process Capability Analysis

CLI Options

Option	Short	Description	Default
--file	-f	CSV file to upload	None
--thread-id	-t	Conversation thread ID	cli_session
--user-id	-u	User identifier	cli_user
--timeout		Request timeout (seconds)	120
--verbose	-v	Verbose output with metadata	False
--output	-o	Save response to JSON file	None
--url		API base URL	http://localhost:8000

---

CLI Examples

Control Charts

# Basic analysis
./spc chat control-charts "Analyze this data" -f data.csv

# Verbose output
./spc chat control-charts "Check for out-of-control points" -f data.csv -v

# Save results
./spc chat control-charts "Generate full report" -f data.csv -o report.json

MSA Example

./spc chat msa "Conduct a Gage R&R study" -f gage_data.csv

Capability Example

./spc chat capability "Assess capability with USL=10.5, LSL=9.5, target=10.0" -f process_data.csv

Conversational Threads

# Initial message
./spc chat control-charts "Analyze this data" -f data.csv -t analysis_001

# Follow-up queries using same thread
./spc chat control-charts "What are the main issues?" -t analysis_001
./spc chat control-charts "Give me recommendations" -t analysis_001

Advanced Usage

Custom API URL:

./spc --url http://remote-server:8000 chat msa "Run study" -f data.csv

Extended Timeout:

./spc chat control-charts "Analyze large dataset" -f big.csv --timeout 300

Batch Automation:

#!/bin/bash
for file in data/*.csv; do
  echo "Processing $file..."
  ./spc chat control-charts "Analyze and report" -f "$file" -t batch_$(date +%Y%m%d)
done

---

API Usage

You can also call the API directly.

Python Example

import requests

response = requests.post(
    "http://localhost:8000/chat/control-charts",
    data={"message": "Analyze this data", "thread_id": "my_session", "user_id": "engineer1"},
    files={"file": open("data.csv", "rb")}
)
print(response.json()["response"])

curl Example

curl -X POST http://localhost:8000/chat/control-charts \
  -F "message=Analyze this data" \
  -F "thread_id=session1" \
  -F "user_id=user1" \
  -F "file=@data.csv"

Endpoints

Method	Endpoint	Description
GET	/	API overview
GET	/health	Health check
POST	/chat/{agent_id}	Chat with an agent (control-charts, msa, capability)

---

Data Format

Each CSV should include the relevant measurement columns. The system will auto-detect column roles but supports explicit naming.

Control Charts

date,measurement
2024-01-01,10.2
2024-01-02,10.1
2024-01-03,9.9

Subgroup Data

subgroup,measurement
1,10.2
1,10.1
1,9.9
2,10.0
2,10.3

MSA (Gage R&R)

Part,Operator,Measurement,Trial
1,A,10.2,1
1,A,10.1,2
1,B,10.3,1
1,B,10.2,2

Process Capability

measurement
10.2
10.1
9.9
10.0

The system is tolerant of alternative column names and will infer structure automatically.

---

Sample Data

Control Charts

• spc_individual_in_control.csv
• spc_individual_out_of_control.csv
• spc_subgroup_data.csv
• spc_np_chart_data.csv
• spc_c_chart_data.csv

MSA Studies

• msa_gage_rr_excellent.csv
• msa_gage_rr_poor.csv
• msa_bias_study.csv
• msa_linearity_study.csv
• msa_stability_study.csv

Process Capability

• capability_excellent.csv
• capability_off_center.csv
• capability_high_variation.csv
• capability_skewed_data.csv

---

Output

Each analysis produces:

• Detailed statistical summaries
• HTML report with charts
• Plain-language recommendations
• Out-of-control detection (SPC)
• Acceptance criteria (MSA)
• Capability indices (Cp, Cpk, Pp, Ppk)

Reports are stored in temp_uploads/.

---

Project Structure

├── agent_config/               # LLM and agent settings
│   ├── config.yaml             # Model provider config
│   └── agent_prompts/          # Agent prompt templates
├── api/                        # FastAPI server
├── control_chart_system/       # SPC logic
├── msa_system/                 # MSA logic
├── process_capability_system/  # Capability analysis logic
├── data_samples/               # Example datasets
├── spc_cli.py                  # Command-line tool

---

Example Workflow

# 1. Start the server
uvicorn api.main:app --host 0.0.0.0 --port 8000 &

# 2. Validate measurement system
./spc chat msa "Conduct Gage R&R study" -f gage_data.csv

# 3. If MSA passes, monitor process
./spc chat control-charts "Check process stability" -f process_data.csv

# 4. If process is stable, check capability
./spc chat capability "Calculate Cp and Cpk with USL=10.5, LSL=9.5" -f process_data.csv

---

Use Cases

Manufacturing

• Monitor process stability using control charts
• Validate measurement systems for precision
• Assess process capability for continuous improvement

Research & Development

• Verify measurement reliability before experiments
• Ensure data integrity in prototypes or trials
• Automate statistical exploration of test results

---

Conversation Memory

Agents retain context within a session (--thread-id), allowing natural multi-step discussions:

./spc chat control-charts "Analyze this" -f data.csv -t project_001
./spc chat control-charts "Explain those out-of-control points" -t project_001
./spc chat control-charts "Give improvement suggestions" -t project_001

---

Reports example

agent chat

---

Contributing

This is a work in progress project. Contributions are welcome code, documentation, or any ideas.

1. Open an issue describing your suggestion or bug
2. Fork and submit a pull request

---

Credits

Built with:

• Langchain — Prebuilt agents
• FastAPI — REST API
• Plotly — Visualization
• scipy, numpy, pandas — Statistical backbone

---

Disclaimer: This tool uses AI (LLMs) for statistical interpretation. Always verify results with standard software and domain expertise before making decisions. The AI assists quality engineers — it doesn’t replace them.