LangGraph Text Analysis Workflow

A comprehensive LangGraph project demonstrating multi-node workflows, state management, and Ollama integration.

Quick Start

1. Install Ollama (if not already installed):

   • Visit https://ollama.ai
   • Download and install for your OS

2. Start Ollama and pull a model:

   ollama serve
   ollama pull llama3.2

3. Install Python dependencies:

   cd backend
   pip install -r requirements.txt

4. Run the demo:

   python main.py

Project Features

Architecture

• 2 Processing Nodes:

  • input_processor: Analyzes text and counts words
  • summarizer: Generates summary and sentiment analysis

• State Management: TypedDict-based state with 4 fields:

  • input_text: Original user input
  • word_count: Calculated by first node
  • summary: Generated by second node
  • sentiment: Generated by second node

Key Features

• ✅ Configurable Ollama model selection
• ✅ Memory persistence with checkpointing
• ✅ Streaming support for real-time updates
• ✅ Multiple output formats (text, JSON, markdown)
• ✅ Comprehensive logging
• ✅ Error handling
• ✅ CLI interface with arguments
• ✅ Example scripts

Usage Examples

Basic Usage

from src.graph.workflow import run_workflow

result = run_workflow("Your text here...")
print(result["summary"])
print(result["sentiment"])

Custom Model

result = run_workflow(
    input_text="Your text...",
    model_name="llama3.2"
)

Streaming

from src.graph.workflow import stream_workflow

for update in stream_workflow("Your text..."):
    print(update)

CLI Usage

# Use sample text
python main.py --sample 0

# Custom text
python main.py --text "Your custom text here"

# Different model
python main.py --sample 0 --model llama3.2

# JSON output
python main.py --sample 1 --format json

# Streaming mode
python main.py --sample 0 --stream

Project Structure

backend/
├── src/
│   ├── config/
│   │   └── models.py          # Model configuration & initialization
│   ├── graph/
│   │   ├── state.py            # State schema (TypedDict)
│   │   ├── nodes.py            # Node implementations
│   │   └── workflow.py         # Graph construction & compilation
│   └── utils/
│       └── helpers.py          # Utility functions
├── examples/
│   └── examples.py             # Usage examples
├── main.py                     # CLI entry point
├── requirements.txt
└── README.md

Advanced Features

Model Presets

from src.config.models import get_model_from_preset

# Use predefined configurations
model = get_model_from_preset("creative")    # High temperature
model = get_model_from_preset("precise")     # Low temperature
model = get_model_from_preset("balanced")    # Default

Persistent Conversations

workflow = create_workflow()

# First conversation
result1 = workflow.invoke(
    {"input_text": "First text..."},
    {"configurable": {"thread_id": "user-123"}}
)

# Continue same conversation
result2 = workflow.invoke(
    {"input_text": "Second text..."},
    {"configurable": {"thread_id": "user-123"}}
)

Direct Node Usage

from src.graph.nodes import input_processor, summarizer

state = {"input_text": "Your text..."}
state.update(input_processor(state))
state.update(summarizer(state))

Use Case

This workflow implements a text analysis pipeline:

1. Input Stage: User provides text
2. Processing Stage: Calculate word count and metadata
3. Analysis Stage: Generate summary and sentiment
4. Output Stage: Return comprehensive results

Perfect for:

• Content analysis
• Document summarization
• Sentiment monitoring
• Text preprocessing pipelines

Configuration

Environment Variables

Create a .env file:

OLLAMA_BASE_URL=http://localhost:11434
DEFAULT_MODEL=llama3.2

Model Configuration

Edit src/config/models.py to customize:

• Default model name
• Temperature settings
• Context window size
• Sampling parameters

Requirements

• Python 3.11+
• Ollama running locally
• At least one Ollama model installed

Troubleshooting

Model not found:

ollama pull llama3.2

Connection refused:

ollama serve

Import errors:

pip install -r requirements.txt

Examples

Run the examples script to see all features:

cd examples
python examples.py

Future Extensions

This project is designed to be extensible. Consider adding:

• Additional analysis nodes (keyword extraction, entity recognition)
• Database persistence
• API endpoints
• Frontend interface
• Batch processing
• Advanced error recovery
• Custom reducers for state management

License

MIT License - feel free to use and modify for your needs.