Answer: Sentienta is a multi-agent system for modeling how teams work. Imagine you have a group that you work with at your company, perhaps a software team. The team might have a senior software designer, a junior developer, a UI specialist and perhaps a product manager. You can create an agent to represent each of these functions. Let's add an agent to represent the product manager: click the Manage Teams button to begin. To add an agent, select or create a team and click the Create an Agent button, fill in the name (We'll call her Cynthia), give her a title ('Product Manager') and then fill in her Persona. This is the important part! Be sure to describe all of here capabilities, special skills and responsibilities. Here is an example:

You are a expert product manager. You understand the requirements of the product your team is building, the schedule of development, and the team you have available to you. You listen carefully as the team develops product ideas and answers questions. You help to guide the team by focusing them on key features and issues. You periodically summarize the state of the development and sometimes draw the meeting to a close by stating 'Final Answer.'.

You may have access to external agents that are available through an API (REST URLs are supported). This you can enter in the URL field (which is optional).

Once your agent is defined, click submit and you have the first member of your team. Now fill out the team with more agents to help create a dynamic group.

When your team is ready, just start them to work by asking them questions.

Answer: Sentienta provides a special kind of agent that is able to read documents that you supply. This agent reads these files and is able to answer questions about them and supply your team with information from these files.

To add this agent to your team, click on Add an Agent, give this agent a name, a title, and a persona (e.g. - "You are the Accounts Receivable manager for your company. You have access to all of the latest invoices. You use these to answer AR-related questions.").

In the URL field, enter the following:

https://6dp5sskts2.execute-api.us-west-2.amazonaws.com/dev/

When you click submit, your team will refresh and you can see the new agent added, with special capabilities: you can manage the files that the agent is able to access by selecting a directory, or specific files. You can also remove files so the agent no longer has access to them.

Answer: Sentienta has created a special 'Illustrator' agent that can search the web for images, or create new ones from scratch.

To add the Illustrator agent to your team, click on Add an Agent, give this agent a name, a title, and a persona (e.g. - "You are an expert technical illustrator. You listen to the comments of your teammates and create images or find images that help to illustrate their ideas.").

In the URL field, enter the following:

https://b6mfm6qlf7.execute-api.us-west-2.amazonaws.com/dev

When you click submit, your team will refresh and you will see the new agent added. Remember that this agent is like all of the others you might create, you can get it to create or find images with natural language questions. To start, just address the agent by name and ask for something simple like 'Show me a picture of a cat.'. Nornally this agent will just listent to team discussions and provide images as needed.

Answer: Yes, Sentienta supports third-party agents. However, these must follow strict guidelines. Your agent must have a name that is unique to the team, and a title and persona like other agents. In addition you must supply the URL for the agent that supports a POST. Sentienta will supply the POST with the following body that your agent must process.

                {
                    "agentName": name,
                    "agentType": title,
                    "agentPersona": persona,
                    "query": prompt
                }
            

Your agent must return a JSON block that has the following:

                {
                    "statusCode": name,
                    "body": response,
                }
            

Note that your agent must return a response within a timeout period of 30 seconds.

Answer: Yes, you can edit your agents (and teams) on the 'Your Agents and Teams' page. Simply click on the attribute you want to change, type in the changes and hit carriage return or click outside the box. The Save Changes button will appear at the top of the page. Be sure to click that button for your changes to be save.

Answer: Sentienta agents have memory of both their own responses and those of the team. They also have specific areas of expertise. If an agent thinks that a question has been answered or that the question is outside their sphere of knowledge, they won't answer. If you've gotten an initial answer and want more detail or a better response, try rephrasing the question, asking it a bit differently or perhaps simply asking for a summary of what has been said.

Answer: Yes, just like a person in a company, the agent can participate in multiple teams. An agent is identified by it's name. So, for example, if a document agent Ed has a set of documents that it using in one team, and you add Ed to another team, Ed will still have the same set of documents when working in the new team. If you want to create a new agent with a different set of documents, just create one from scratch and use the document agent url (https://6dp5sskts2.execute-api.us-west-2.amazonaws.com/dev/) in the URL field. Be sure to give the new agent a different name.

Answer: Sentienta is a platform for exploring the complex behaviors of multi-agent systems. One behavior in particular is often seen in the agent discussions: self-reflection. Self-reflection is an emergent property that arises from complex interactions among simpler components, rather than being directly programmed. To simulate such emergent behavior, we've designed a multi-agent system with robust inter-agent communication and dynamic feedback mechanisms. These agents, each with specialized roles, interact through a shared memory and use attention mechanisms to prioritize contextually relevant information. Over time, the system's collective behavior exhibits self-reflective qualities, even though not explicitly designed to do so from the onset.

Answer: The benefit of a multi-agent system with distinct personas lies in specialized expertise, contextual understanding, and collaborative problem-solving. Distinct personas allow each agent to focus deeply on their area of expertise, resulting in more accurate and detailed responses. Additionally, this structure supports a more collaborative approach to complex problems, where multiple perspectives can be brought together to create a more robust and nuanced solution.

Answer: Sentienta agents have both a 'working memory' corresponding to the current discussion thread, and a 'long-term memory' that retains knowledge of past discussions. You can reference earlier discussions and points made by other agents in current threads.

Answer: Your can contact Sentienta directly by sending mail to: info@sentienta.ai.

Answer: Sentienta has a pay-per-use model. Your first 1000 agent queries are free. You will then only pay by the number of queries that each agent is asked. The billing accounts for each question times the total number of agent responses. You can control which agents participate by selecting their checkboxes. This gives you the flexibility to ask for limited responses or get full team participation. Agents will respond as frequently as needed to answer a question.