Agent Cloud

Integrations

Blog

API Reference

Sign Up to Cloud

Home Page

Github

Retrieve a JSON list of all apps the user has access to. Will also return a list of all tasks, tools, agents, models and datasources as a front end helper

Get Apps

True or false check to determine if the agent is allowed to delegate tasks to other agents in the context of an app

A detailed description of what the LLM will be doing

A secondary model used to execute function calls (this links to a Model object), set to null if unused automatically by the API

The goal of the agent is fed into the LLM, this allows the LLM to know it's role in the RAG pipeline

IconAttachment object used to hold the attached icon used for the agent (this links to an IconAttachment object);

The linked ObjectId of the model being used by this agent (this links to a Model object)

Organisation the agent is linked to (generally the org of the user that created the agent)

Fed into the LLM to help it provide a more detailed and correct response

Team the agent is linked to (generally the team of the user that created the agent)

Array of the tools the agent can access to improve performance and abstract tool functionality from agent usage

True or false check to determine if custom verbosity is used in agent, higher verbosity requires agent to include more of the retrieved documents at the expense of longer answers, lower verbosity can result in shorter answers but can also ommit crucial details

Represents a data source configuration within the system, including connection details, status, chunking configurations, and record counts.

Configuration settings for chunking unstructured data, including partitioning and chunking strategies, character limits, and similarity thresholds.

The identifier of the connection associated with the datasource.

Configuration settings for the datasource connection.

The date and time when the datasource was created.

Schema discovered during the data source connection. The structure depends on the source type.

The field used for embedding within the datasource.

The name of the file associated with the datasource, if applicable.

Indicates whether the datasource is hidden from standard views.

The date and time when the datasource was last synced. Null indicates it has never been synced.

Identifier of the embedding model used, if applicable.

Identifier of the organization to which the datasource belongs.

The record count details for the datasource, including total, successful, and failed records.

Configuration settings for processing streams of data, breaking them into smaller chunks for more manageable processing.

Identifier of the team to which the datasource belongs.

A temporary field to limit CRON frequency based on the plan. This will be replaced with a more robust solution in the future.

The field used to apply time weighting within the datasource.

The identifier of the workspace associated with the datasource.

Configuration settings for a datasource connection, including scheduling, data residency, and namespace details.

Configuration settings for the datasource connection. Structure is dependent on the datasource type.

Specifies where the data should be stored geographically.

Defines how the namespace should be determined for the data.

The format of the namespace, can be null if not applicable.

Specifies the behavior for handling non-breaking schema updates.

Optional prefix to be added to the destination's namespace. Can be null.

Scheduling information for the datasource connection.

The CRON expression for scheduling, required if the schedule type is 'cron'.

The type of schedule for the datasource connection.

The status of the datasource connection. This should match the enum values defined by the Airbyte API and should allow creation in a paused state.

Represents the record count information for a datasource, including total, successful, and failed records.

The number of records that failed to process.

The number of successfully processed records.

Provides details about a specific field, including its description and type.

A mapping of field names to their descriptions.

Configuration for a form field within a task, including position, type, and optional settings.

An optional description for the form field.

Options available for fields like radio, select, or multiselect.

The position of the form field within the form layout.

A tooltip providing additional information about the form field.

Defines the properties of a function parameter in a tool.

Attachment, generally used for image upload

Filename of the attachment at the point of upload

this is NOT a unique id for the IconAttachment, this is a Mongo id that links to an Asset object (neet to implement assets in docs)

Information about a metadata field used in retriever configurations.

Represents an AI model used within a team or organization.

The length of the embeddings generated by the model.

Identifier of the organization to which the model belongs.

Identifier of the team to which the model belongs.

The general type of the model (e.g., embedding, language model).

Mongodb Object id, unique identifier, length of 24 characters fitting the following regex; [a-f0-9]{24}

Enum representing different types of retrievers used in tools.

Union type representing different retriever configurations.

Configuration for the self-query retriever.

Array of metadata field information objects.

Configuration settings for a specific stream, used to break down large volumes of data into smaller, manageable chunks for processing.

List of child stream identifiers that are checked for inclusion in the sync.

List of fields that act as the cursor for incremental syncs.

A map of field names to their descriptions.

List of fields that make up the primary key for the stream.

The synchronization mode used for the stream.

A mapping of stream names to their respective configuration settings.

Represents a task within the system, including its configuration, expected output, and optional form fields.

Identifier of the agent associated with the task.

Indicates if the task is executed asynchronously.

Contextual information related to the task.

Indicates if only the final output should be displayed.

Array of form field configurations associated with the task.

Indicates if the task is hidden from standard views.

Icon associated with the task, either an attachment or an object containing the icon details.

Indicates if the output of the task is structured.

Identifier of the organization to which the task belongs.

Indicates if the task requires human input.

Identifier of the team to which the task belongs.

List of tool identifiers associated with the task.

Configuration for the time-weighted retriever.

Decay rate applied to the time weighting.

Represents a tool within the system, including its configuration, state, and associated retrievers.

Data related to the tool, including runtime, environment variables, and more.

Indicates if the tool is a built-in feature.

Environment variables required by the tool.

Key used to match OpenAPI specifications.

Identifier of the datasource associated with the tool.

Identifier of the function associated with the tool.

Logs related to the function's execution.

Identifier of the organization to which the tool belongs.

Configuration settings for the retriever.

Identifier of the team to which the tool belongs.

Enum representing the possible states of a tool.

Enum representing different types of tools.

Configuration for chunking unstructured data, including partitioning and chunking strategies, character limits, and similarity thresholds.

The maximum number of characters allowed per chunk.

The number of characters after which a new chunk is created.

The number of characters to overlap between chunks.

Indicates whether to apply overlap to all chunks or only between adjacent chunks.

The partitioning strategy used for unstructured data.

The chunking strategy used for unstructured data.

Enum representing the strategy used for unstructured data chunking.

Enum representing the strategy used for unstructured data partitioning.

Test the endpoints on your own instance of Agent Cloud!

Retrieve a JSON object of the session from path sessionId. Will also return a list of all tasks, tools, agents, models and datasources as a front end helper

Retrieve Session

Retrieve a JSON list of all sessions the user has access to. Will also return a list of all tasks, tools, agents, models and datasources as a front end helper

Retrieve Sessions

Get all messages associated to the path sessionId

Retrieve Messages

Create a new chat session from the parameter appId

Add a session

Cancel a session by Id, best practice to cancel a session once it has completed. The session and it's associated fields will still exist within the database under the same sessionId. The session can be completley removed from the database using the delete route AFTER this has been executed.

Cancel a session

Delete a session

Retrieve a JSON object of the retrieved app. Will also return a list of all tasks, tools, agents, models and datasources as a front end helper

Get App

Retrieve a JSON object of a single agent from agentId and the resourceSlug

Retrieve Agent

Retrieve a JSON list of all agents the user has access to

Retrieve Agents

Add an agent

Edit An Agent

Delete An Agent

Retrieve a JSON object with a singular task to be retrieved by id.

Retrieve Task

Retrieve a JSON list of all tasks the user has access to

Retrieve Tasks

Add A New Task

Edit A Task

Delete A Task

Retrieve a JSON object of a single tool from toolId and the resourceSlug

Retrieve Tool

Retrieve a JSON list of all tools the user has access to

Retrieve Tools

Add A New Tool

Edit an existing tool by its ID within the specified team.

Edit A Tool

Delete a tool by its ID within the specified team.

Delete A Tool

Retrieve a JSON object of a single model within the specified team by modelId.

Retrieve Model

Retrieve a JSON list of all models the user has access to within the specified team.

Retrieve Models

Add A New Model

Edit an existing model by its ID within the specified team.

Edit A Model

Delete a model by its ID within the specified team.

Delete A Model

Retrieve A Datasource

Retrieve a JSON list of all Datasources the user has access to within the specified team.

Retrieve Datasources

Edit Schedule In Datasource - This is marked as Uncomplete

Edit Datasource Streams - This Is Marked As Unfinished

Initialise an Airbyte synchronisation for an existing datasource object

Sync Datasource

Test a new datasource. This is required to gather the datasource schema, tables and values to continue to the next stage of addition. This MUST be executed first when adding a new datasource

Test A Datasource To Be Added

Second stage of adding Datasource (check test datasource for the first step), this connects the datasource to a vectordb, creates a connection between the datasource and airbyte, starts a job to synchronise the rows and finally creates a tool for the datasource. A tool for the datasource is required as within agents a datasource is used as a 'tool' just the same as any other tool, this is required for the 'RAG' process.

Add A Datasource

Introduction

Agent Cloud - Talk to Your Data

Quickstart

How to Create Your First Agent Cloud App

Local Setup

Get Started with Agent Cloud on a Local Deployment

How to use Agent Cloud on your local development machine.

Local Dev with CLI

Agent Cloud Platform

Agents

Apps

Datasources

Models

Sessions

Tasks

Tools

Variables

Step by step guide to setup a conversational chat app to RAG a Google Big Query datasource (or any other data source)

Reference

Get Apps

Path Parameters

Response