X Trends Agent - Built with LangGraph & Agent Stack SDK

About

This Agent is an AI-powered conversational tool for accessing trends on X. It is built using LangGraph for orchestration and the Agent Stack SDK, specializing in identifying trending topics by country, researching the context behind each trend, and providing friendly, engaging summaries with emojis and source citations.

The agent is designed to run entirely locally using Ollama and the IBM Granite 4 model.

Agent Workflow (LangGraph)

Below is the visual representation of the agent's logic flow orchestrated by LangGraph:

graph TD
    Start((Start)) --> GetTrends[analyze_and_get_trends]
    GetTrends --> CountryDetect{Detect Country}
    CountryDetect -->|Country Found| BuildURL[trends24.in/country/]
    CountryDetect -->|No Country| MainURL[trends24.in/]
    BuildURL --> ExtractTrends[Extract Top 5]
    MainURL --> ExtractTrends[Extract Top 5]
    ExtractTrends --> ResearchContext[research_trends_context]
    ResearchContext --> Loop[Search Reason for each Trend]
    Loop --> Synthesize[synthesize_report]
    Synthesize --> End((End))

Loading

Requirements

Minimum Requirements

• Python: Version 3.11 or higher.
• Dependency Management: uv is recommended for managing Python packages.
• Ollama: Required for running the local LLM (granite4:tiny-h).

Python Dependencies

Main packages used:

• langchain & langgraph: For orchestrating the agent's reasoning flow.
• langchain-ollama: To interact with the local LLM.
• agentstack-sdk: For A2A protocol and server management.
• ddgs: Web search via DuckDuckGo.

Installation

1. Clone the repository:

   git clone <repository-url>
   cd x_trends_agent

2. Install dependencies:

   uv sync

3. Install Ollama and the model:

   # Pull the required model
   ollama pull granite4:tiny-h

Running the Agent

1. Start the agent:

   uv run server

   The agent will start and be ready on http://127.0.0.1:8002 (configurable via PORT env var).

How It Works

The agent uses a StateGraph (LangGraph) to process queries through a dynamic pipeline:

1. Country Detection: Uses the LLM to identify if the user is asking about a specific region.
2. Trend Retrieval: Scrapes trends24.in specifically for the target country (or global) to get the top 5 trending topics.
3. Context Research: For each of the top 5 trends, it performs targeted web searches to find the "why" behind the trend.
4. Friendly Synthesis: Generates a coherent report with emojis (💡, 📰, 🚀) and, most importantly, includes the source URLs of the news found.

Usage Examples

• "What's trending in Mexico?"
• "Show me global trends on X today"
• "What's happening in Spain?"

Project Structure

• src/langgraph_agents/agent.py: Main logic of the agent, including the LangGraph definition and A2A server.
• pyproject.toml: Dependency and script definitions.

Disclaimer

This agent is built for demonstration purposes. Ensure you have Ollama running locally for the LLM to function properly.