houseassassin/twenty
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Restructure agent chat messages with parts-based architecture (#14749)

Browse Source 
Co-authored-by: Félix Malfait <felix@twenty.com>
This commit is contained in:
Abdul Rahman
2025-09-29 17:01:55 +05:30
committed by GitHub
parent 84cbd8e092
commit 2685f4a5b9
69 changed files with 1200 additions and 1539 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/twenty-server/src/engine/workspace-manager/dev-seeder/core/utils/seed-agents.util.ts
+129 -23
View File
@@ -11,6 +11,7 @@ const agentTableName = 'agent';
const workspaceTableName = 'workspace';
const agentChatThreadTableName = 'agentChatThread';
const agentChatMessageTableName = 'agentChatMessage';
const agentChatMessagePartTableName = 'agentChatMessagePart';
export const AGENT_DATA_SEED_IDS = {
APPLE_DEFAULT_AGENT: '20202020-0000-4000-8000-000000000001',
@@ -33,6 +34,17 @@ export const AGENT_CHAT_MESSAGE_DATA_SEED_IDS = {
YCOMBINATOR_MESSAGE_4: '20202020-0000-4000-8000-000000000034',
};
export const AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS = {
APPLE_MESSAGE_1_PART_1: '20202020-0000-4000-8000-000000000041',
APPLE_MESSAGE_2_PART_1: '20202020-0000-4000-8000-000000000042',
APPLE_MESSAGE_3_PART_1: '20202020-0000-4000-8000-000000000043',
APPLE_MESSAGE_4_PART_1: '20202020-0000-4000-8000-000000000044',
YCOMBINATOR_MESSAGE_1_PART_1: '20202020-0000-4000-8000-000000000051',
YCOMBINATOR_MESSAGE_2_PART_1: '20202020-0000-4000-8000-000000000052',
YCOMBINATOR_MESSAGE_3_PART_1: '20202020-0000-4000-8000-000000000053',
YCOMBINATOR_MESSAGE_4_PART_1: '20202020-0000-4000-8000-000000000054',
};
const seedAgentChatThreads = async (
dataSource: DataSource,
schemaName: string,
@@ -88,16 +100,24 @@ const seedAgentChatMessages = async (
threadId: string,
) => {
let messageIds: string[];
let partIds: string[];
let messages: Array<{
id: string;
threadId: string;
role: AgentChatMessageRole;
content: string;
createdAt: Date;
}>;
let messageParts: Array<{
id: string;
messageId: string;
orderIndex: number;
type: string;
textContent: string;
createdAt: Date;
}>;
const now = new Date();
const baseTime = new Date(now.getTime() - 24 * 60 * 60 * 1000); // 24 hours ago
const baseTime = new Date(now.getTime() - 24 * 60 * 60 * 1000);
if (workspaceId === SEED_APPLE_WORKSPACE_ID) {
messageIds = [
@@ -106,38 +126,74 @@ const seedAgentChatMessages = async (
AGENT_CHAT_MESSAGE_DATA_SEED_IDS.APPLE_MESSAGE_3,
AGENT_CHAT_MESSAGE_DATA_SEED_IDS.APPLE_MESSAGE_4,
];
partIds = [
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.APPLE_MESSAGE_1_PART_1,
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.APPLE_MESSAGE_2_PART_1,
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.APPLE_MESSAGE_3_PART_1,
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.APPLE_MESSAGE_4_PART_1,
];
messages = [
{
id: messageIds[0],
threadId,
role: AgentChatMessageRole.USER,
content:
'Hello! Can you help me understand our current product roadmap and key metrics?',
createdAt: new Date(baseTime.getTime()),
},
{
id: messageIds[1],
threadId,
role: AgentChatMessageRole.ASSISTANT,
content:
"Hello! I'd be happy to help you understand Apple's product roadmap and metrics. Based on your workspace data, I can see you have various projects and initiatives tracked. What specific aspect would you like to explore - product development timelines, user engagement metrics, or revenue targets?",
createdAt: new Date(baseTime.getTime() + 5 * 60 * 1000), // 5 minutes later
createdAt: new Date(baseTime.getTime() + 5 * 60 * 1000),
},
{
id: messageIds[2],
threadId,
role: AgentChatMessageRole.USER,
content:
"I'd like to focus on our user engagement metrics and how they're trending over the last quarter.",
createdAt: new Date(baseTime.getTime() + 10 * 60 * 1000), // 10 minutes later
createdAt: new Date(baseTime.getTime() + 10 * 60 * 1000),
},
{
id: messageIds[3],
threadId,
role: AgentChatMessageRole.ASSISTANT,
content:
createdAt: new Date(baseTime.getTime() + 15 * 60 * 1000),
},
];
messageParts = [
{
id: partIds[0],
messageId: messageIds[0],
orderIndex: 0,
type: 'text',
textContent:
'Hello! Can you help me understand our current product roadmap and key metrics?',
createdAt: new Date(baseTime.getTime()),
},
{
id: partIds[1],
messageId: messageIds[1],
orderIndex: 0,
type: 'text',
textContent:
"Hello! I'd be happy to help you understand Apple's product roadmap and metrics. Based on your workspace data, I can see you have various projects and initiatives tracked. What specific aspect would you like to explore - product development timelines, user engagement metrics, or revenue targets?",
createdAt: new Date(baseTime.getTime() + 5 * 60 * 1000),
},
{
id: partIds[2],
messageId: messageIds[2],
orderIndex: 0,
type: 'text',
textContent:
"I'd like to focus on our user engagement metrics and how they're trending over the last quarter.",
createdAt: new Date(baseTime.getTime() + 10 * 60 * 1000),
},
{
id: partIds[3],
messageId: messageIds[3],
orderIndex: 0,
type: 'text',
textContent:
'Great! Looking at your user engagement data, I can see several key trends from the last quarter. Your active user base has grown by 15%, with particularly strong engagement in the mobile app. Daily active users are averaging 2.3 million, and session duration has increased by 8%. Would you like me to dive deeper into any specific engagement metrics or create a detailed report?',
createdAt: new Date(baseTime.getTime() + 15 * 60 * 1000), // 15 minutes later
createdAt: new Date(baseTime.getTime() + 15 * 60 * 1000),
},
];
} else if (workspaceId === SEED_YCOMBINATOR_WORKSPACE_ID) {
@@ -147,38 +203,74 @@ const seedAgentChatMessages = async (
AGENT_CHAT_MESSAGE_DATA_SEED_IDS.YCOMBINATOR_MESSAGE_3,
AGENT_CHAT_MESSAGE_DATA_SEED_IDS.YCOMBINATOR_MESSAGE_4,
];
partIds = [
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.YCOMBINATOR_MESSAGE_1_PART_1,
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.YCOMBINATOR_MESSAGE_2_PART_1,
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.YCOMBINATOR_MESSAGE_3_PART_1,
AGENT_CHAT_MESSAGE_PART_DATA_SEED_IDS.YCOMBINATOR_MESSAGE_4_PART_1,
];
messages = [
{
id: messageIds[0],
threadId,
role: AgentChatMessageRole.USER,
content:
'What are the current startup trends and which companies in our portfolio are performing best?',
createdAt: new Date(baseTime.getTime()),
},
{
id: messageIds[1],
threadId,
role: AgentChatMessageRole.ASSISTANT,
content:
'Hello! I can help you analyze startup trends and portfolio performance. From your YCombinator workspace data, I can see strong performance in AI/ML startups, particularly in the B2B SaaS space. Several companies are showing 40%+ month-over-month growth. Would you like me to provide specific company performance metrics or focus on broader industry trends?',
createdAt: new Date(baseTime.getTime() + 3 * 60 * 1000), // 3 minutes later
createdAt: new Date(baseTime.getTime() + 3 * 60 * 1000),
},
{
id: messageIds[2],
threadId,
role: AgentChatMessageRole.USER,
content:
'Please focus on our top 5 performing companies and their key metrics.',
createdAt: new Date(baseTime.getTime() + 8 * 60 * 1000), // 8 minutes later
createdAt: new Date(baseTime.getTime() + 8 * 60 * 1000),
},
{
id: messageIds[3],
threadId,
role: AgentChatMessageRole.ASSISTANT,
content:
createdAt: new Date(baseTime.getTime() + 12 * 60 * 1000),
},
];
messageParts = [
{
id: partIds[0],
messageId: messageIds[0],
orderIndex: 0,
type: 'text',
textContent:
'What are the current startup trends and which companies in our portfolio are performing best?',
createdAt: new Date(baseTime.getTime()),
},
{
id: partIds[1],
messageId: messageIds[1],
orderIndex: 0,
type: 'text',
textContent:
'Hello! I can help you analyze startup trends and portfolio performance. From your YCombinator workspace data, I can see strong performance in AI/ML startups, particularly in the B2B SaaS space. Several companies are showing 40%+ month-over-month growth. Would you like me to provide specific company performance metrics or focus on broader industry trends?',
createdAt: new Date(baseTime.getTime() + 3 * 60 * 1000),
},
{
id: partIds[2],
messageId: messageIds[2],
orderIndex: 0,
type: 'text',
textContent:
'Please focus on our top 5 performing companies and their key metrics.',
createdAt: new Date(baseTime.getTime() + 8 * 60 * 1000),
},
{
id: partIds[3],
messageId: messageIds[3],
orderIndex: 0,
type: 'text',
textContent:
'Here are your top 5 performing portfolio companies: 1) TechFlow AI - 45% MoM growth, $2M ARR, 2) DataSync Pro - 38% MoM growth, $1.5M ARR, 3) CloudOps Solutions - 35% MoM growth, $3.2M ARR, 4) SecureNet - 32% MoM growth, $1.8M ARR, 5) HealthTech Plus - 28% MoM growth, $2.5M ARR. All are showing strong customer retention (>95%) and expanding market share. Would you like detailed breakdowns for any specific company?',
createdAt: new Date(baseTime.getTime() + 12 * 60 * 1000), // 12 minutes later
createdAt: new Date(baseTime.getTime() + 12 * 60 * 1000),
},
];
} else {
@@ -194,12 +286,26 @@ const seedAgentChatMessages = async (
'id',
'threadId',
'role',
'content',
'createdAt',
])
.orIgnore()
.values(messages)
.execute();
await dataSource
.createQueryBuilder()
.insert()
.into(`${schemaName}.${agentChatMessagePartTableName}`, [
'id',
'messageId',
'orderIndex',
'type',
'textContent',
'createdAt',
])
.orIgnore()
.values(messageParts)
.execute();
};
export const seedAgents = async (
packages/twenty-server/src/engine/workspace-manager/workspace-sync-metadata/interfaces/partial-field-metadata.interface.ts
+4 -1
View File
@@ -1,4 +1,7 @@
import { type FieldMetadataType } from 'twenty-shared/types';
import {
type ExcludeFunctions,
type FieldMetadataType,
} from 'twenty-shared/types';
import { type WorkspaceDynamicRelationMetadataArgsFactory } from 'src/engine/twenty-orm/interfaces/workspace-dynamic-relation-metadata-args.interface';
packages/twenty-server/src/engine/workspace-manager/workspace-sync-metadata/utils/sync-metadata.util.spec.ts
+1
View File
@@ -1,4 +1,5 @@
import { FieldMetadataType } from 'twenty-shared/types';
import { type DeepPartial } from 'typeorm/common/DeepPartial';
import { type ObjectMetadataEntity } from 'src/engine/metadata-modules/object-metadata/object-metadata.entity';