Autonomous AI Agent
Multi-Step Reasoning with Tool Use & Cited Reports
Autonomous agent that searches, synthesizes, and generates cited research reports with 5–10 tool calls per query.
Context
A research agent that does the legwork: it breaks a question down, reads across multiple sources, weighs what it finds, and writes it up with citations you can actually check. I built it to see how far the reason-and-act loop could go before it needed a human — and to make the parts you usually can’t see, the reasoning and the tool calls, visible while it works.
The problem
Research tasks require multi-step reasoning: decomposing questions, searching across sources, evaluating relevance, synthesizing findings, and producing coherent reports. Single-shot LLM calls can't handle this complexity reliably. The challenge was building an agent that reasons through research tasks autonomously while maintaining accuracy and providing transparent, cited outputs.
How I built it
Implemented ReAct (Reason + Act) pattern with GPT-4 function calling for structured tool invocation
Designed tool suite including web search, content extraction, summarization, and citation management
Built dual memory system: short-term conversation buffer for session context plus persistent ChromaDB vector memory for long-term knowledge
Added self-correction loops that detect incomplete or contradictory findings and trigger additional research steps
Created streaming Vue.js frontend showing real-time agent reasoning and tool invocations
Why these choices
ReAct over simple chain-of-thought
ReAct interleaves reasoning with action, allowing the agent to adapt its research strategy based on intermediate findings rather than committing to a fixed plan upfront.
Dual memory architecture
Conversation buffer maintains session coherence while vector memory enables the agent to build cumulative knowledge across sessions — critical for ongoing research topics.
Self-correction loops
Without self-correction, agents accept first-pass results regardless of quality. Adding reflection and retry steps significantly improved report completeness and accuracy.