LLMFlow Search is an agent that finds accurate answers to complex questions using a smart search strategy.

It automatically refines queries: if the initial results are incomplete or inaccurate, the agent generates additional queries to fill in the gaps.

The agent explores information from various sources — Wikipedia, DuckDuckGo, and websites directly. It:

• Identifies which parts need verification
• Expands or narrows the search as needed
• Detects and resolves contradictions
• Chooses more precise wording

The result is a coherent, verified answer based on real data. It works in multiple languages and can bypass site restrictions.

• Python 3.8+
• OpenAI API key
• Chrome/Chromium (for Selenium-based web searches)

1. Clone the repository:

   git clone https://github.com/KazKozDev/LLMFlow-Search.git
   cd LLMFlow-Search

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables:

   cp .env.example .env

   Edit the .env file, adding your OpenAI API key.

Start the application:

python __main__.py

The system uses a config.json file for configuration. Example configuration:

{
    "llm": {
        "model": "gpt-4o-mini",
        "temperature": 0.2,
        "max_tokens": 2048
    },
    "search": {
        "max_results": 5,
        "safe_search": true,
        "parse_top_results": 3,
        "use_selenium": true,
        "use_cache": true
    },
    "memory": {
        "path": "./memory",
        "max_items": 100
    },
    "report": {
        "output_file": "report.md",
        "include_sources": true,
        "max_source_length": 1500
    }
}

LLMFlow Search Agent consists of the following main modules:

1. Planning Module: Analyzes queries and creates a search plan
2. Tools Module: Provides tools for searching DuckDuckGo, Wikipedia, and web pages
3. Memory Module: Stores and retrieves information for context-aware processing
4. Report Generator: Synthesizes information into comprehensive reports
5. LLM Service: Provides interaction with OpenAI language models

If you like this project, please give it a star ⭐

For questions, feedback, or support, reach out to:

Artem KK | MIT LICENSE