$ python research_agent.py

>>> What does Parallel's Search API do? Cite parallel.ai in one sentence.

Parallel’s Search API is a tool designed specifically for AI agents and LLMs to perform natural language web searches and retrieve structured, token-efficient data. Unlike traditional search engines that rely on keywords, it processes a "natural language objective" to return ranked URLs with extended excerpts optimized for model consumption, reducing hallucinations and improving accuracy in complex research tasks ([parallel.ai](https://docs.parallel.ai/search/search-quickstart)).

According to the developer documentation: "The Search API takes a natural language objective and returns relevant excerpts optimized for LLMs, replacing multiple keyword searches with a single call for broad or complex queries" ([parallel.ai](https://docs.parallel.ai/search/search-quickstart)).

### Key Features
*   **Search Modes:** Offers specialized modes including **one-shot** (comprehensive), **agentic** (concise for multi-step reasoning), and **fast** (latency-sensitive) ([parallel.ai](https://docs.parallel.ai/search/modes)).
*   **LLM Optimization:** Excerpts are curated to fit efficiently into a model's context window, lowering token costs while maintaining high relevance ([parallel.ai](https://parallel.ai/)).
*   **Accuracy Benchmarks:** Parallel claims its enterprise deep research API achieves up to 48% accuracy on complex benchmarks, significantly outperforming the native browsing capabilities of standard LLMs ([parallel.ai](https://parallel.ai/)).
*   **Evidence-Based:** Every output provides verifiability and provenance, ensuring AI agents can cite their sources accurately ([parallel.ai](https://parallel.ai/)).

--- Parallel call trace (1 calls) ---
  {"tool": "_web_search", "latency_s": 1.4036233751103282, "citation_count": 5}
