Advanced Social Media Agent

An AI agent for intelligent social media analysis, built on the Apify platform. It implements production-grade patterns like ReAct and Reflexion for reasoning and self-improvement.

Overview

This actor is a specialized AI agent designed to analyze social media content. Its architecture is based on modern research patterns, including ReAct for transparent reasoning and Reflexion for learning from past runs. It's structured as a 12-factor agent, emphasizing small, focused tasks and robust error handling.

Key Features

• Intelligent Reasoning: Uses the ReAct pattern to interleave chain-of-thought reasoning with tool-using actions (Thought -> Action -> Observation).
• Self-Improvement: Implements the Reflexion pattern for verbal reinforcement learning, allowing the agent to learn from each execution.
• Advanced Memory System: Maintains context with a multi-tier system (Working, Episodic, Semantic, and Procedural memory).
• Human-in-the-Loop: Includes approval gates for high-risk actions and supports interrupt/resume capabilities.
• Production Ready: Features full observability (tracing, metrics), error recovery with self-healing, and cost control with budget limits.
• Flexible Modes: Can operate in Autonomous, Supervised, Interactive, or Research (read-only) modes.

How to Use

Basic Python Usage

from src.core.agent import AgentBuilder, AgentMode

# Build the agent
agent = (AgentBuilder()
  .with_name("SocialMediaAnalyzer")
  .with_mode(AgentMode.SUPERVISED)
  .with_memory()
  .with_reflection()
  .with_budget(1.0) # $1 USD limit
  .with_tools(instagram_tools)
  .with_tools(analysis_tools)
  .build(llm_client))

# Run a task
result = await agent.run(
  task="Analyze the last 10 posts from @openai and summarize AI trends"
)
print(f"Success: {result.success}")
print(f"Response: {result.output}")
print(f"Cost: ${result.cost_usd:.4f}")

Project Structure

The actor is organized into clear modules:
* core/: Contains the main agent orchestrator, state management, and control flow engine.
* patterns/: Houses the implementations of ReAct and Reflexion.
* memory/: Manages the multi-tier memory system.
* tools/: Includes tools for platforms like Instagram and for analysis tasks.
* observability/: Handles distributed tracing and metrics.

Input / Output

Input Configuration

Output

The agent returns a result object containing:
* success: A boolean indicating if the task completed successfully.
* output: The final response or analysis from the agent.
* cost_usd: The total cost of the execution in USD.

The results are also stored according to the defined dataset_schema.json for structured access via the Apify platform.