Coordinator#

class lumen.ai.coordinator.Coordinator(llm: Llm | None = None, interface: ChatInterface | None = None, agents: list[Agent | type[Agent]] | None = None, logs_db_path: str = '', **params)#

A Coordinator is responsible for coordinating the actions of a number of agents towards the user defined query by computing an execution graph and then executing each step along the graph.

Parameters#

agents

type: list
default: [<class 'lumen.ai.agents.ChatAgent'>]
List of agents to coordinate.

demo_inputs

type: list
default: ['What datasets are available?', 'Show me the the first dataset and its columns.', 'What are some interesting analyses I can do?', 'Perform an analysis using the first suggestion.', 'Show me a plot of these results.']
List of instructions to demo the Coordinator.

history

type: int
default: 3
bounds: None
Number of previous user-assistant interactions to include in the chat history.

interface

type: panel.chat.ChatInterface
default: None
The ChatInterface for the Coordinator to interact with.

llm

type: lumen.ai.llm.Llm
default: None
The LLM implementation to query.

logs_db_path

type: str
default: None
The path to the log file that will store the messages exchanged with the LLM.

memory

type: lumen.ai.memory._Memory
default: None
Local memory which will be used to provide the agent context.If None the global memory will be used.

name

type: str
default: 'Coordinator'
String identifier for this object.

prompts

type: dict
default: {}
A dictionary of prompts used by the actor, indexed by prompt name.Each prompt should be defined as a dictionary containing a template’template’ and optionally a ‘model’ and ‘tools’.

render_output

type: bool
default: True
Whether to write outputs to the ChatInterface.

suggestions

type: list
default: ['What datasets are available?', "What's interesting to analyze?", 'Can you visualize the data?']
Initial list of suggestions of actions the user can take.

template_overrides

type: dict
default: {}
Overrides the template’s ‘instructions’, ‘context’, ‘tools’, or ‘examples’ jinja2 blocks.Is a nested dictionary with the prompt name (e.g. main) as the keyand the block names as the inner keys with the new content as thevalues.

tools

type: list
default: []

Methods#

async Coordinator.respond(messages: list[Message], **kwargs: dict[str, Any]) → str#

Responds to the provided messages.

Coordinator.servable(title: str | None = None, location: bool | Location = True, area: str = 'main', target: str | None = None) → Viewable#

Serves the object or adds it to the configured pn.state.template if in a panel serve context, writes to the DOM if in a pyodide context and returns the Panel object to allow it to display itself in a notebook context.

Parameters:

  • title (str) – A string title to give the Document (if served as an app)

  • location (boolean or panel.io.location.Location) – Whether to create a Location component to observe and set the URL location.

  • area (str (deprecated)) – The area of a template to add the component too. Only has an effect if pn.config.template has been set.

  • target (str) – Target area to write to. If a template has been configured on pn.config.template this refers to the target area in the template while in pyodide this refers to the ID of the DOM node to write to.

Return type:

The Panel object itself

Coordinator.show(title: str | None = None, port: int = 0, address: str | None = None, websocket_origin: str | None = None, threaded: bool = False, verbose: bool = True, open: bool = True, location: bool | Location = True, **kwargs) → threading.Thread | Server#

Starts a Bokeh server and displays the Viewable in a new tab.

Parameters:

  • title (str | None) – A string title to give the Document (if served as an app)

  • port (int (optional, default=0)) – Allows specifying a specific port

  • address (str) – The address the server should listen on for HTTP requests.

  • websocket_origin (str or list(str) (optional)) – A list of hosts that can connect to the websocket. This is typically required when embedding a server app in an external web site. If None, “localhost” is used.

  • threaded (boolean (optional, default=False)) – Whether to launch the Server on a separate thread, allowing interactive use.

  • verbose (boolean (optional, default=True)) – Whether to print the address and port

  • open (boolean (optional, default=True)) – Whether to open the server in a new browser tab

  • location (boolean or panel.io.location.Location) – Whether to create a Location component to observe and set the URL location.

Returns:

server – Returns the Bokeh server instance or the thread the server was launched on (if threaded=True)

Return type:

bokeh.server.Server or panel.io.server.StoppableThread

Types#

DependencyResolver

DependencyResolver is a type of Coordinator that chooses the agent

DependencyResolver.html
Planner

The Planner develops a plan to solve the user query step-by-step

Planner.html