Uploaded June 2025 | Updated September 2026, 2 weeks ago
🔍 Building Agentic Search Systems with LLMs | Jina AI Talk
In this talk, we break down how to construct agentic search systems powered by Large Language Models (LLMs). Learn how to guide an LLM through iterative reasoning, combining user input, question-answer history, and search results to eventually produce reliable, final answers.
🧠Key Concepts Covered:
The prompt structure driving the agent’s reasoning process
How the LLM chooses between generating search queries vs. providing a final answer
Iterative search and response loops with Q&A accumulation
Managing growing context and dealing with LLM limitations (e.g., via summarization)
Using internal and external search engines to power your AI agent
Real-world use cases like competitive product analysis
Guardrails to ensure safe, controlled agent behavior
Future outlook: collaborative human-AI research agents
đź’ˇ Takeaway: With just a small prompt and minimal code, you can build powerful agentic systems. While production systems are more complex, this lightweight framework is often enough to solve meaningful enterprise problems.
🔍 Building Agentic Search Systems with LLMs | Jina AI Talk
In this talk, we break down how to construct agentic search systems powered by Large Language Models (LLMs). Learn how to guide an LLM through iterative reasoning, combining user input, question-answer history, and search results to eventually produce reliable, final answers.
🧠Key Concepts Covered:
The prompt structure driving the agent’s reasoning process
How the LLM chooses between generating search queries vs. providing a final answer
Iterative search and response loops with Q&A accumulation
Managing growing context and dealing with LLM limitations (e.g., via summarization)
Using internal and external search engines to power your AI agent
Real-world use cases like competitive product analysis
Guardrails to ensure safe, controlled agent behavior
Future outlook: collaborative human-AI research agents
đź’ˇ Takeaway: With just a small prompt and minimal code, you can build powerful agentic systems. While production systems are more complex, this lightweight framework is often enough to solve meaningful enterprise problems.
