90 lines
2.9 KiB
TypeScript
Executable File
90 lines
2.9 KiB
TypeScript
Executable File
import type { BaseMessage } from '@langchain/core/messages';
|
|
import { HumanMessage } from '@langchain/core/messages';
|
|
import { Annotation, messagesStateReducer } from '@langchain/langgraph';
|
|
|
|
import type { SimpleWorkflow, WorkflowOperation } from './types/workflow';
|
|
import type { ChatPayload } from './workflow-builder-agent';
|
|
|
|
/**
|
|
* Reducer for collecting workflow operations from parallel tool executions.
|
|
* This reducer intelligently merges operations, avoiding duplicates and handling special cases.
|
|
*/
|
|
function operationsReducer(
|
|
current: WorkflowOperation[] | null,
|
|
update: WorkflowOperation[] | null | undefined,
|
|
): WorkflowOperation[] {
|
|
if (update === null) {
|
|
return [];
|
|
}
|
|
|
|
if (!update || update.length === 0) {
|
|
return current ?? [];
|
|
}
|
|
|
|
// For clear operations, we can reset everything
|
|
if (update.some((op) => op.type === 'clear')) {
|
|
return update.filter((op) => op.type === 'clear').slice(-1); // Keep only the last clear
|
|
}
|
|
|
|
if (!current && !update) {
|
|
return [];
|
|
}
|
|
// Otherwise, append new operations
|
|
return [...(current ?? []), ...update];
|
|
}
|
|
|
|
// Creates a reducer that trims the message history to keep only the last `maxUserMessages` HumanMessage instances
|
|
export function createTrimMessagesReducer(maxUserMessages: number) {
|
|
return (current: BaseMessage[]): BaseMessage[] => {
|
|
// Count HumanMessage instances and remember their indices
|
|
const humanMessageIndices: number[] = [];
|
|
current.forEach((msg, index) => {
|
|
if (msg instanceof HumanMessage) {
|
|
humanMessageIndices.push(index);
|
|
}
|
|
});
|
|
|
|
// If we have fewer than or equal to maxUserMessages, return as is
|
|
if (humanMessageIndices.length <= maxUserMessages) {
|
|
return current;
|
|
}
|
|
|
|
// Find the index of the first HumanMessage that we want to keep
|
|
const startHumanMessageIndex =
|
|
humanMessageIndices[humanMessageIndices.length - maxUserMessages];
|
|
|
|
// Slice from that HumanMessage onwards
|
|
return current.slice(startHumanMessageIndex);
|
|
};
|
|
}
|
|
|
|
export const WorkflowState = Annotation.Root({
|
|
messages: Annotation<BaseMessage[]>({
|
|
reducer: messagesStateReducer,
|
|
default: () => [],
|
|
}),
|
|
// // The original prompt from the user.
|
|
// The JSON representation of the workflow being built.
|
|
// Now a simple field without custom reducer - all updates go through operations
|
|
workflowJSON: Annotation<SimpleWorkflow>({
|
|
reducer: (x, y) => y ?? x,
|
|
default: () => ({ nodes: [], connections: {}, name: '' }),
|
|
}),
|
|
// Operations to apply to the workflow - processed by a separate node
|
|
workflowOperations: Annotation<WorkflowOperation[] | null>({
|
|
reducer: operationsReducer,
|
|
default: () => [],
|
|
}),
|
|
// Whether the user prompt is a workflow prompt.
|
|
// Latest workflow context
|
|
workflowContext: Annotation<ChatPayload['workflowContext'] | undefined>({
|
|
reducer: (x, y) => y ?? x,
|
|
}),
|
|
|
|
// Previous conversation summary (used for compressing long conversations)
|
|
previousSummary: Annotation<string>({
|
|
reducer: (x, y) => y ?? x, // Overwrite with the latest summary
|
|
default: () => 'EMPTY',
|
|
}),
|
|
});
|