DeepResearch requires structured input that defines the scope and objectives of the research task. Users must provide a clear research topic or question, along with optional parameters like source types (e.g., academic papers, news articles), timeframes, or specific domains (e.g., AI, healthcare). For example, a user might input a query like "Analyze recent advancements in quantum computing hardware since 2020, focusing on academic journals and industry whitepapers." The system also supports context through additional constraints, such as excluding certain sources or prioritizing data from specific organizations. Inputs should avoid ambiguity to ensure the tool retrieves relevant data and generates accurate insights.
To initiate a DeepResearch session, users typically interact with an interface (e.g., command-line tool, API, or web dashboard) and submit their query using a predefined format. For instance, a developer might use a CLI command like deepresearch start --query "impact of GPT-4 on code generation" --sources "arXiv, GitHub repositories" --timeframe "2022-2023". Alternatively, an API request could include a JSON payload specifying the query, filters, and output preferences. The system validates the input to ensure required parameters are present and formats the request for processing. Once validated, the session begins, and the user receives a session ID to track progress or retrieve results later.
The workflow often involves iterative refinement. For example, a user might start with a broad query like "machine learning in climate science," review initial findings, and then narrow the scope by adding filters like "peer-reviewed studies published in 2023" or "focus on carbon capture projects." DeepResearch may also prompt users for clarifications if inputs are too vague or lack sufficient context. This approach ensures the tool aligns with the user’s goals while minimizing irrelevant data. Developers integrating DeepResearch into applications can automate session initiation by programmatically passing user inputs to the API, enabling seamless embedding in workflows like data pipelines or research platforms.