Project Agent
Learn how the Project Agent system works, which takes Projects to the next level.
Projects are both containers of things (Notes, Files, etc) as well as Agents. This is a superpower. Below you can learn about the Project Agent, but its useful to have already read about the Sapience object graph, here:
Key Things To Know
- All Projects are also Agents;
- The Project Agent is available to all User roles, and you open it from the Project Dashboard > Actions > Chat with Project;
- Project Agents do not show in the Agent Browser;
- A Project Agent has access to all of the things inside a Project, including the core Project Meta-data like Success Criteria, Target Completion Date etc, as well as the Goals, Tasks, Files and Notes in the Project.
- It won’t go outside the sandbox of the Project. I.e. it won’t read Files and Notes outside of the Project. That’s deliberate and good!
How it Works
When you start a conversation with the Project Agent, it first reads into its Memory, all of the core Project Metadata, as well as the full contents of all the Goals, Tasks and Notes in the Project.
For Files, the Project Agent has the filename, summary and description of each file, and can open them and read them if either you tell it to, or it thinks it is necessary based on what you ask it (we don’t auto-load file contents, as this causes context bloat).
To get started, we recommend using some of the canned questions. In the screenshot above, we clicked on the “What tasks are associated with this project?” and you can see that it answered purely based on the context of the Project itself.
Project Agent Tools
Customizing the Project Agent
You can make the Project Agent even more powerful, and smarter, by configuring Project-specific facts and behavior preferences for it. That’s covered in this separate article: