OpenHands/agenthub

Agent Hub

In this folder, there may exist multiple implementations of Agent that will be used by the framework.

For example, agenthub/codeact_agent, etc. Contributors from different backgrounds and interests can choose to contribute to any (or all!) of these directions.

Constructing an Agent

The abstraction for an agent can be found here.

Agents are run inside of a loop. At each iteration, agent.step() is called with a State input, and the agent must output an Action.

Every agent also has a self.llm which it can use to interact with the LLM configured by the user. See the LiteLLM docs for self.llm.completion.

State

The state contains:

• A history of actions taken by the agent, as well as any observations (e.g. file content, command output) from those actions
• A list of actions/observations that have happened since the most recent step
• A root_task, which contains a plan of action
  • The agent can add and modify subtasks through the AddTaskAction and ModifyTaskAction

Actions

Here is a list of available Actions, which can be returned by agent.step():

• CmdRunAction - Runs a command inside a sandboxed terminal
• IPythonRunCellAction - Execute a block of Python code interactively (in Jupyter notebook) and receives CmdOutputObservation. Requires setting up jupyter plugin as a requirement.
• FileReadAction - Reads the content of a file
• FileWriteAction - Writes new content to a file
• BrowseURLAction - Gets the content of a URL
• AddTaskAction - Adds a subtask to the plan
• ModifyTaskAction - Changes the state of a subtask.
• AgentFinishAction - Stops the control loop, allowing the user/delegator agent to enter a new task
• AgentRejectAction - Stops the control loop, allowing the user/delegator agent to enter a new task
• AgentFinishAction - Stops the control loop, allowing the user to enter a new task
• MessageAction - Represents a message from an agent or the user

You can use action.to_dict() and action_from_dict to serialize and deserialize actions.

Observations

There are also several types of Observations. These are typically available in the step following the corresponding Action. But they may also appear as a result of asynchronous events (e.g. a message from the user).

Here is a list of available Observations:

• CmdOutputObservation
• BrowserOutputObservation
• FileReadObservation
• FileWriteObservation
• ErrorObservation
• SuccessObservation

You can use observation.to_dict() and observation_from_dict to serialize and deserialize observations.

Interface

Every agent must implement the following methods:

step

def step(self, state: "State") -> "Action"

step moves the agent forward one step towards its goal. This probably means sending a prompt to the LLM, then parsing the response into an Action.