textured 0.1.0

A Dart package to use LLMs to parse text into pre-defined data structures.

example/README.md

Textured CLI

A simple command-line interface for the textured package that demonstrates LLM-powered text parsing with JSON schema validation.

Features

• 📋 Schema Loading: Loads JSON Schema files to define output structure
• 🤖 LLM Integration: Uses Ollama for natural language processing
• ✅ Schema Validation: Validates LLM output against the provided schema
• 📊 Rich Output: Displays extracted data, validation results, and performance metrics
• 🎯 Error Handling: Comprehensive error messages and debugging support

Prerequisites

1. Ollama: Install and run Ollama locally

   # Install Ollama (visit https://ollama.ai for instructions)
      
   # Start Ollama server
   ollama serve
      
   # Pull a model (e.g., llama2)
   ollama pull llama2
   

2. Dart SDK: Make sure you have Dart installed (>=3.0.0)

Usage

Basic Usage

dart run example/textured_cli.dart <schema_file> [model]

Arguments

• schema_file: Path to a JSON Schema file (required)
• model: Ollama model name (optional, default: llama2)

Examples

Using the included tree schema:

dart run example/textured_cli.dart test/data/tree_metadata_schema.json

Using a different model:

dart run example/textured_cli.dart test/data/tree_metadata_schema.json mistral

With piped input:

echo "A beautiful cherry blossom tree in Tokyo, Japan" | dart run example/textured_cli.dart test/data/tree_metadata_schema.json

Interactive Mode

When run without piped input, the CLI enters interactive mode:

1. Enter your text to parse
2. Press Ctrl+D (Linux/Mac) or Ctrl+Z (Windows) when finished
3. The tool will process your input and display results

Output Format

The CLI displays:

• ✅ Success/Failure Status: Whether the LLM call succeeded
• 🔍 Validation Results: Schema validation status and any errors
• 📋 Extracted Data: The structured JSON output
• 📈 Performance Metrics: Response time, token counts, model info

Example Session

🌟 Textured CLI - LLM Text Parser
═══════════════════════════════════
Schema: test/data/tree_metadata_schema.json
Model: llama2

📋 Loading schema...
   ✓ Loaded: Tree Metadata Schema
   ✓ Properties: 32
   ✓ Required fields: 32

🤖 Configuring LLM...
   ⚡ Testing Ollama connection...
   ✓ Connection successful

📝 Enter your text to parse:
   (Type your text and press Ctrl+D when finished)

> This is a large oak tree in Central Park, New York.

📤 Processing text (47 characters)...
   🎯 Generating LLM prompt...
   ✓ Prompt generated (5344 characters)
   🚀 Calling LLM...
   ✓ Response received (8181ms)

📊 RESULTS
═══════════════════════════════════
✅ Success!

✅ Schema Validation: PASSED

📋 Extracted Data:
━━━━━━━━━━━━━━━━━━━
{
  "tree_id": null,
  "name": "Large Oak Tree",
  "city": "New York",
  "state": "NY",
  "is_alive": true,
  ...
}

📈 Metadata:
━━━━━━━━━━━━━━━━━━━
• Model: llama2
• Response time: 8175ms
• Content length: 775 characters

🎉 Processing complete!

Error Handling

The CLI provides helpful error messages for common issues:

• Missing Ollama: Instructions to start Ollama server
• Invalid Schema: JSON format validation errors
• Missing Files: File not found errors
• Model Issues: Model availability problems

Debugging

Add --debug flag for detailed error information:

dart run example/textured_cli.dart test/data/tree_metadata_schema.json --debug

Creating Custom Schemas

You can use any JSON Schema Draft 7 file. The schema should define:

• Required properties
• Property types and constraints
• Examples (optional but helpful for LLM accuracy)

See test/data/tree_metadata_schema.json for a comprehensive example.

Performance Tips

1. Use appropriate models: Larger models (e.g., mistral) may be more accurate but slower
2. Optimize temperature: Lower values (0.1-0.3) for more consistent extraction
3. Schema design: Well-defined schemas with examples improve accuracy
4. Hardware: Ensure sufficient RAM for your chosen model

Troubleshooting

"Connection failed"

• Check if Ollama is running: ollama serve
• Verify the model is available: ollama list

"Model not found"

• Pull the model: ollama pull <model_name>
• Check available models: ollama list

"Schema validation failed"

• Review the schema file for JSON syntax errors
• Ensure the schema follows JSON Schema Draft 7 format

Poor extraction accuracy

• Try a larger model (e.g., mistral, codellama)
• Add examples to your schema
• Lower the temperature parameter
• Provide clearer, more structured input text