LLMonitor JS API Reference
Class: LLMonitor
Constructor
| Method | Description |
|---|
constructor() | Initializes the LLMonitor instance. |
Methods
| Method | Description |
|---|
init({ appId?: string, verbose?: boolean, apiUrl?: string }) | Initializes the LLMonitor with the provided options. |
trackEvent(type: string, event: string, data: RunEvent) | Tracks an event with the provided type, event name, and data. The data parameter is of type RunEvent, which has the following structure: |
| Property | Description |
|---|
type | The type of the event. It can be one of the following: "log", "tool", "agent", "llm", "convo", "chain", "retriever", "embed", "chat", "convo". |
event | The name of the event. It can be one of the following: "start", "end", "error", "info", "warn", "feedback". |
app | The ID of the app. |
timestamp | The timestamp of the event. |
userId | The ID of the user (optional). |
userProps | The properties of the user (optional). |
parentRunId | The ID of the parent run (optional). |
extra | Extra data (optional). |
tags | Tags associated with the event (optional). |
runtime | The runtime of the event (optional). |
error | The error object, if any, with properties message and stack (optional). |
runId | The ID of the run. |
input | The input data (optional). |
output | The output data (optional). |
tokensUsage | The token usage data (optional). |
| Method | Description |
| ----------------------------------------------------------- | ---------------------------------------------------------------------------------------- | ------------------------------------------------------ |
| wrapAgent(func: Function, params?: WrapParams) -> Promise | Wraps an agent's Promise to track its input, results, and any errors. Returns a Promise. |
| wrapTool(func: Function, params?: WrapParams) -> Promise | Wraps a tool's Promise to track its input, results, and any errors. Returns a Promise. |
| wrapModel(func: Function, params?: WrapParams) -> Promise | Wraps a model's Promise to track its input, results, and any errors. Returns a Promise. |
| info(message: string, extra?: any) | Logs an informational message with optional extra data. |
| log(message: string, extra?: any) | Alias for info. |
| warn(message: string, extra?: any) | Logs a warning message with optional extra data. |
| error(message: string | any, error?: any) | Reports any errors that occur during the conversation. |
| flush() -> Promise | Ensures the event queue is flushed before exiting the program. Returns a Promise. |
| trackFeedback(messageId: string, feedback: cJSON) | Tracks feedback for a specific message. |
| startThread() -> Thread | Starts a new chat conversation and returns a Thread instance. |
| resumeThread(id: string) -> Thread | Resumes a chat conversation with the given thread ID and returns the Thread instance. |
Class: Thread
Constructor
constructor(id?: string): Initializes the Thread instance with a provided LLMonitor.
Methods
-
trackUserMessage(text: string, props?: cJSON, customId?: string) -> string: Tracks a new message from the user and returns the message ID.
-
trackBotMessage(replyToId: string, text: string, props?: cJSON): Tracks a new message from the bot.
(React Specific Exports) Hooks
-
useChatMonitor() -> { startThread: Function, trackUserMessage: Function, trackBotMessage: Function, trackFeedback: Function }: A React hook that provides functions for starting a new thread, tracking user messages, tracking bot messages, and tracking feedback.
-
useMonitorVercelAI(props) -> { ...props, trackFeedback: Function }: A React hook that provides a function for tracking feedback in addition to the properties passed in.
RunEvent Structure
Questions? We're here to help.
