https://start.askwonder.com/admin/ai/playground?workflowName=researchOS_v1_with_agents
This is the main workflow that generates the outputs for the clients that we usually see. It is a chain of client chat, research agents, and various other prompts.
ResearchOS takes in the client’s chat transcript as the initial input. From there, it breaks down various objectives that the client wants us to complete and launches individual research agents for each of those objectives. When all the research agents complete their individual research tasks after using their research tools, we get the final research report (i.e., reports from the research agents stitched together).
When using rOS for projects that did not go through Wonderpad (e.g., those sent in by email), there are two options for inputting the client’s ask: Chat Messages or Objectives Override.
The “Objectives Override” feature allows the user to enter individual objectives. When using this option, rOS will not evaluate and determine the best way to break the research paths up into objectives. Given this, it is critical that each objective enters represents a single research path.
When using this feature, keep the following in mind:
https://app.tango.us/app/embed/Running-ResearchOS---Agent-Workflow-for-the-Entire-Chat-c868a7a718f14910a96b2d53be7df671
Multi-pronged asks / Great when you don't know exactly what you want.
The main workflow that generates the full research reports' output is what we have traditionally seen in rOS reports/"instant insights" generated for clients. It is capable of handling all multi-objective use cases.