result-parser-agent 0.2.1

A deep agent for extracting metrics from raw result files using LangGraph and intelligent parsing

🎯 Results Parser Agent

A powerful, intelligent agent for extracting metrics from raw result files using LangGraph and AI-powered parsing. The agent automatically analyzes unstructured result files and extracts specific metrics into structured JSON output with high accuracy.

🚀 Features

• 🤖 AI-Powered Parsing: Uses advanced LLMs (GROQ, OpenAI, Anthropic, Google Gemini, Ollama) for intelligent metric extraction
• 📁 Flexible Input: Process single files or entire directories of result files
• 🎯 Pattern Recognition: Automatically detects and adapts to different file formats and structures
• ⚙️ Rich Configuration: YAML/JSON configuration with environment variable support
• 📊 Structured Output: Direct output in Pydantic schemas for easy integration
• 🛠️ Professional CLI: Feature-rich command-line interface with comprehensive options
• 🔧 Python API: Easy integration into existing Python applications
• 🔄 Error Recovery: Robust error handling and retry mechanisms

📦 Installation

Quick Install (Recommended)

pip install result-parser-agent

Development Install

# Clone the repository
git clone https://github.com/yourusername/result-parser-agent.git
cd result-parser-agent

# Install with uv (recommended)
curl -LsSf https://astral.sh/uv/install.sh | sh
uv sync
uv pip install -e .

# Or install with pip
pip install -e .

🎯 Quick Start

1. Set up your API key

# For GROQ (default - recommended for speed and reliability)
export GROQ_API_KEY="your-groq-api-key-here"

# For OpenAI
export OPENAI_API_KEY="your-openai-api-key-here"

# For Anthropic
export ANTHROPIC_API_KEY="your-anthropic-api-key-here"

# For Google Gemini
export GOOGLE_API_KEY="your-google-api-key-here"

2. Use the CLI

# Parse a directory of result files
result-parser --dir ./benchmark_results --metrics "RPS,latency,throughput" --output results.json

# Parse a single file with custom LLM
result-parser --file ./specific_result.txt --metrics "accuracy,precision" --provider openai --model gpt-4

# Use YAML configuration file
result-parser --config ./my_config.yaml --file ./results.txt --metrics "RPS,throughput"

# Override specific settings
result-parser --dir ./results --metrics "RPS" --provider groq --model llama3.1-70b-versatile --temperature 0.2

# Verbose output with debug info
result-parser --dir ./results --metrics "RPS" --verbose

# Custom output file
result-parser --file ./results.txt --metrics "throughput,latency" --output my_results.json

3. Use the Python API

from result_parser_agent import (
    ResultsParserAgent, 
    get_groq_config, 
    get_openai_config,
    load_config_from_file,
    modify_config
)
import os

# Method 1: Use pre-configured provider configs
config = get_groq_config(
    model="llama3.1-8b-instant",
    metrics=["RPS", "latency", "throughput"]
)

# Method 2: Load from YAML file
config = load_config_from_file("./my_config.yaml")

# Method 3: Modify default config
config = modify_config(
    provider="openai",
    model="gpt-4",
    temperature=0.2,
    metrics=["accuracy", "precision", "recall"]
)

# Initialize agent
agent = ResultsParserAgent(config)

# Parse results (file or directory)
result_update = await agent.parse_results(
    input_path="./benchmark_results",  # or "./results.txt"
    metrics=["RPS", "latency", "throughput"]
)

# Output structured data
print(result_update.json(indent=2))

📋 Configuration

Configuration File Example

# config.yaml
agent:
  # LLM configuration
  llm:
    provider: "groq"                    # groq, openai, anthropic, google, ollama
    model: "llama3.1-8b-instant"        # Fast and efficient for parsing tasks
    api_key: null                       # Set to null to use environment variable
    temperature: 0.1                    # Temperature for LLM responses
    max_tokens: 4000                    # Maximum tokens for responses
  
  # Agent behavior
  max_retries: 3
  chunk_size: 2000
  timeout: 300

parsing:
  # Metrics to extract from result files
  metrics:
    - "RPS"
    - "latency"
    - "throughput"
    - "accuracy"
    - "precision"
    - "recall"
    - "f1_score"
  
  # Parsing options
  case_sensitive: false
  fuzzy_match: true
  min_confidence: 0.7

output:
  format: "json"
  pretty_print: true
  include_metadata: true

logging:
  level: "INFO"
  format: "{time} | {level} | {message}"
  file: null

Environment Variables

You can also configure the agent using environment variables:

# API Keys
export GOOGLE_API_KEY="your-google-api-key-here"
export OPENAI_API_KEY="your-openai-api-key-here"
export ANTHROPIC_API_KEY="your-anthropic-api-key-here"
export GROQ_API_KEY="your-groq-api-key-here"

# Configuration
export PARSER_AGENT__LLM__PROVIDER="google"
export PARSER_AGENT__LLM__MODEL="gemini-2.0-flash"
export PARSER_PARSING__METRICS='["RPS", "latency", "throughput"]'
export PARSER_OUTPUT__FORMAT="json"

🛠️ CLI Reference

Command Options

result-parser [OPTIONS]

Options:
  -d, --dir TEXT              Directory containing result files to parse (use --dir OR --file)
  -f, --file TEXT             Single result file to parse (use --dir OR --file)
  -m, --metrics TEXT          Comma-separated list of metrics to extract (required, e.g., 'RPS,latency,throughput')
  -o, --output PATH           Output JSON file path (default: results.json)
  -v, --verbose               Enable verbose logging
  --log-level TEXT            Logging level [default: INFO]
  --pretty-print              Pretty print JSON output [default: True]
  --no-pretty-print           Disable pretty printing
  --help                      Show this message and exit

Usage Examples

# Parse all files in a directory
result-parser --dir ./benchmark_results --metrics "RPS,latency" --output results.json

# Parse a single file
result-parser --file ./specific_result.txt --metrics "accuracy,precision"

# Verbose output for debugging
result-parser --dir ./results --metrics "RPS" --verbose

# Custom output file
result-parser --file ./results.txt --metrics "throughput,latency" --output my_results.json

# Compact JSON output
result-parser --dir ./results --metrics "accuracy" --no-pretty-print