Freelancer| indie developer

medium

github

akshay Apsingi  • Peerlist

🧠 How It Works: The Agent Team Approach At its core, the system uses a multi-agent architecture orchestrated by a central controller. Planner Agent (The Strategist) Analyzes the incoming query. Decides if it requires fresh web search or logical reasoning. Creates a dynamic, step-by-step plan for other agents to follow. Specialist Agents (Executing the Plan) Search Agent: Fetches relevant information from the web if required. Reasoning Agent: Directly answers queries using internal knowledge when no web search is needed. Filtering Agent: Cleans, deduplicates, and filters the search results to retain only the most relevant context. Synthesizer/Analysis Agent: Combines the results into a clear, structured final report with sources linked at the bottom. Orchestration and Communication A central Orchestrator assigns tasks and manages workflow based on the Planner's output. A Task Manager tracks progress for each task. WebSockets are used to communicate real-time updates and deliver the final report to the user interface. 📈 Current Status (MVP) The system dynamically plans and routes tasks through multiple agents. Able to handle different types of queries (research, reasoning) and produce structured reports. Backend is built with Python (FastAPI). Currently integrates with Google Gemini models.