references/lobe-chat
1
0
Fork 0
mirror of https://github.com/lobehub/lobe-chat.git synced 2026-06-13 19:20:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

♻️ refactor(userMemories): support resolving agent config from ServiceModel (#15138)

Browse Source 
* ♻️ refactor(userMemories): support resolving agent config from ServiceModel

* ♻️ refactor(userMemories): share memory analysis service model
This commit is contained in:
Neko
2026-05-25 04:06:50 +08:00
committed by GitHub
parent d71686ba88
commit f3d5d03cf5
17 changed files with 688 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files
locales/en-US/setting.json
+11
View File
@@ -503,6 +503,8 @@
"plugin.settings.tooltip": "Skill Configuration",
"plugin.store": "Skill Store",
"publishToCommunity": "Publish to Community",
"serviceModel.contextLimit.placeholder": "Context limit",
"serviceModel.memoryModels.title": "Memory Models",
"serviceModel.modelAssignments.title": "Model Assignments",
"serviceModel.optionalFeatures.title": "Optional Features",
"settingAgent.avatar.sizeExceeded": "Image size exceeds 1MB limit, please choose a smaller image",
@@ -850,6 +852,9 @@
"systemAgent.inputCompletion.label": "Model",
"systemAgent.inputCompletion.modelDesc": "Suggests text while you type. When enabled, this model generates the suggestions.",
"systemAgent.inputCompletion.title": "Input Suggestions",
"systemAgent.memoryAnalysisAgentConfig.label": "Model",
"systemAgent.memoryAnalysisAgentConfig.modelDesc": "Model used to decide whether conversations contain memory and extract identities, preferences, contexts, activities, and experiences.",
"systemAgent.memoryAnalysisAgentConfig.title": "Memory Analysis",
"systemAgent.promptRewrite.label": "Model",
"systemAgent.promptRewrite.modelDesc": "Improves prompts before generation. When enabled, this model rewrites the prompt.",
"systemAgent.promptRewrite.title": "Prompt Rewriting",
@@ -863,6 +868,12 @@
"systemAgent.translation.label": "Model",
"systemAgent.translation.modelDesc": "Model used to translate messages",
"systemAgent.translation.title": "Message Translation",
"systemAgent.userMemoryEmbedding.label": "Model",
"systemAgent.userMemoryEmbedding.modelDesc": "Model used to embed memory content for retrieval. The context limit caps each embedding input.",
"systemAgent.userMemoryEmbedding.title": "Memory Embedding",
"systemAgent.userMemoryPersonaWriter.label": "Model",
"systemAgent.userMemoryPersonaWriter.modelDesc": "Model used to write persona-oriented memory summaries.",
"systemAgent.userMemoryPersonaWriter.title": "Memory Persona Writer",
"tab.about": "About",
"tab.addAgentSkill": "Add Agent Skill",
"tab.addCustomMcp": "Add Custom MCP Skill",
locales/zh-CN/setting.json
+11
View File
@@ -503,6 +503,8 @@
"plugin.settings.tooltip": "技能配置",
"plugin.store": "技能商店",
"publishToCommunity": "发布到社区",
"serviceModel.contextLimit.placeholder": "上下文限制",
"serviceModel.memoryModels.title": "记忆模型",
"serviceModel.modelAssignments.title": "模型分配",
"serviceModel.optionalFeatures.title": "可选功能",
"settingAgent.avatar.sizeExceeded": "图片大小超过 1MB 限制，请选择更小的图片",
@@ -850,6 +852,9 @@
"systemAgent.inputCompletion.label": "模型",
"systemAgent.inputCompletion.modelDesc": "输入时生成文本建议。开启后，由该模型生成建议。",
"systemAgent.inputCompletion.title": "输入建议",
"systemAgent.memoryAnalysisAgentConfig.label": "模型",
"systemAgent.memoryAnalysisAgentConfig.modelDesc": "用于判断对话是否包含记忆，并提取身份、偏好、上下文、活动和经历。",
"systemAgent.memoryAnalysisAgentConfig.title": "记忆分析",
"systemAgent.promptRewrite.label": "模型",
"systemAgent.promptRewrite.modelDesc": "生成前优化提示词。开启后，由该模型改写提示词。",
"systemAgent.promptRewrite.title": "提示词改写",
@@ -863,6 +868,12 @@
"systemAgent.translation.label": "模型",
"systemAgent.translation.modelDesc": "用于翻译消息内容的模型",
"systemAgent.translation.title": "消息内容翻译",
"systemAgent.userMemoryEmbedding.label": "模型",
"systemAgent.userMemoryEmbedding.modelDesc": "用于为记忆内容生成向量以支持检索。上下文限制会约束每次向量化输入。",
"systemAgent.userMemoryEmbedding.title": "记忆向量化",
"systemAgent.userMemoryPersonaWriter.label": "模型",
"systemAgent.userMemoryPersonaWriter.modelDesc": "用于生成面向画像的记忆摘要。",
"systemAgent.userMemoryPersonaWriter.title": "记忆画像写入",
"tab.about": "关于",
"tab.addAgentSkill": "添加 Agent 技能",
"tab.addCustomMcp": "添加自定义 MCP 技能",
packages/const/src/settings/systemAgent.ts
+16 -4
View File
@@ -1,11 +1,15 @@
import { DEFAULT_MINI_PROVIDER, DEFAULT_PROVIDER } from '@lobechat/business-const';
import {
DEFAULT_EMBEDDING_PROVIDER,
DEFAULT_MINI_PROVIDER,
DEFAULT_PROVIDER,
} from '@lobechat/business-const';
import type {
PromptRewriteSystemAgent,
SystemAgentItem,
UserSystemAgentConfig,
UserServiceModelConfig,
} from '@lobechat/types';
import { DEFAULT_MINI_MODEL, DEFAULT_MODEL } from './llm';
import { DEFAULT_EMBEDDING_MODEL, DEFAULT_MINI_MODEL, DEFAULT_MODEL } from './llm';
export const DEFAULT_SYSTEM_AGENT_ITEM: SystemAgentItem = {
model: DEFAULT_MODEL,
@@ -35,12 +39,20 @@ export const DEFAULT_FOLLOW_UP_ACTION_SYSTEM_AGENT_ITEM: SystemAgentItem = {
provider: DEFAULT_MINI_SYSTEM_AGENT_ITEM.provider,
};
export const DEFAULT_SYSTEM_AGENT_CONFIG: UserSystemAgentConfig = {
export const DEFAULT_USER_MEMORY_EMBEDDING_SYSTEM_AGENT_ITEM: SystemAgentItem = {
model: DEFAULT_EMBEDDING_MODEL,
provider: DEFAULT_EMBEDDING_PROVIDER,
};
export const DEFAULT_SYSTEM_AGENT_CONFIG: UserServiceModelConfig = {
agentMeta: DEFAULT_SYSTEM_AGENT_ITEM,
followUpAction: DEFAULT_FOLLOW_UP_ACTION_SYSTEM_AGENT_ITEM,
generationTopic: DEFAULT_MINI_SYSTEM_AGENT_ITEM,
historyCompress: DEFAULT_SYSTEM_AGENT_ITEM,
inputCompletion: DEFAULT_INPUT_COMPLETION_SYSTEM_AGENT_ITEM,
memoryAnalysisAgentConfig: DEFAULT_MINI_SYSTEM_AGENT_ITEM,
userMemoryEmbedding: DEFAULT_USER_MEMORY_EMBEDDING_SYSTEM_AGENT_ITEM,
userMemoryPersonaWriter: DEFAULT_MINI_SYSTEM_AGENT_ITEM,
promptRewrite: DEFAULT_PROMPT_REWRITE_SYSTEM_AGENT_ITEM,
thread: DEFAULT_SYSTEM_AGENT_ITEM,
topic: DEFAULT_MINI_SYSTEM_AGENT_ITEM,
packages/types/src/serverConfig.ts
+2 -2
View File
@@ -7,7 +7,7 @@ import type {
GlobalLLMProviderKey,
UserDefaultAgent,
UserImageConfig,
UserSystemAgentConfig,
UserServiceModelConfig,
} from './user/settings';
export type GlobalMemoryLayer = 'activity' | 'context' | 'experience' | 'identity' | 'preference';
@@ -76,7 +76,7 @@ export interface GlobalServerConfig {
image?: PartialDeep<UserImageConfig>;
memory?: GlobalMemoryConfig;
oAuthSSOProviders?: string[];
systemAgent?: PartialDeep<UserSystemAgentConfig>;
systemAgent?: PartialDeep<UserServiceModelConfig>;
telemetry: {
langfuse?: boolean;
};
packages/types/src/user/settings/index.ts
+2 -2
View File
@@ -9,7 +9,7 @@ import type { MarketAuthTokens } from './market';
import type { UserMemorySettings } from './memory';
import type { UserModelProviderConfig } from './modelProvider';
import type { NotificationSettings } from './notification';
import type { UserSystemAgentConfig } from './systemAgent';
import type { UserServiceModelConfig } from './systemAgent';
import type { UserToolConfig } from './tool';
import type { UserTTSConfig } from './tts';
@@ -42,7 +42,7 @@ export interface UserSettings {
market?: MarketAuthTokens;
memory?: UserMemorySettings;
notification?: NotificationSettings;
systemAgent: UserSystemAgentConfig;
systemAgent: UserServiceModelConfig;
tool: UserToolConfig;
tts: UserTTSConfig;
}
packages/types/src/user/settings/systemAgent.ts
+12
View File
@@ -1,4 +1,5 @@
export interface SystemAgentItem {
contextLimit?: number;
customPrompt?: string;
enabled?: boolean;
model: string;
@@ -21,4 +22,15 @@ export interface UserSystemAgentConfig {
translation: SystemAgentItem;
}
export interface UserMemoryServiceModelConfig {
memoryAnalysisAgentConfig: SystemAgentItem;
userMemoryEmbedding: SystemAgentItem;
userMemoryPersonaWriter: SystemAgentItem;
}
export interface UserServiceModelConfig
extends UserSystemAgentConfig, UserMemoryServiceModelConfig {}
export type UserSystemAgentConfigKey = keyof UserSystemAgentConfig;
export type UserMemoryServiceModelConfigKey = keyof UserMemoryServiceModelConfig;
export type UserServiceModelConfigKey = keyof UserServiceModelConfig;
src/features/ServiceModel/ModelAssignmentsForm.tsx
+56 -7
View File
@@ -1,7 +1,7 @@
'use client';
import type { FormGroupItemType, FormItemProps } from '@lobehub/ui';
import { Flexbox, Form, Icon, Skeleton } from '@lobehub/ui';
import { Flexbox, Form, Icon, InputNumber, Skeleton } from '@lobehub/ui';
import { Switch } from 'antd';
import isEqual from 'fast-deep-equal';
import { Loader2Icon } from 'lucide-react';
@@ -12,13 +12,14 @@ import { FORM_STYLE } from '@/const/layoutTokens';
import ModelSelect from '@/features/ModelSelect';
import { useUserStore } from '@/store/user';
import { settingsSelectors } from '@/store/user/selectors';
import type { SystemAgentItem, UserSystemAgentConfigKey } from '@/types/user/settings';
import type { SystemAgentItem, UserServiceModelConfigKey } from '@/types/user/settings';
interface SystemAgentModelItem {
key: UserSystemAgentConfigKey;
contextLimit?: boolean;
key: UserServiceModelConfigKey;
}
type LoadingKey = 'defaultAgent' | UserSystemAgentConfigKey;
type LoadingKey = 'defaultAgent' | UserServiceModelConfigKey;
const SYSTEM_AGENT_MODEL_ITEMS: SystemAgentModelItem[] = [
{ key: 'topic' },
@@ -34,6 +35,12 @@ const OPTIONAL_FEATURE_ITEMS: SystemAgentModelItem[] = [
{ key: 'promptRewrite' },
];
const MEMORY_MODEL_ITEMS: SystemAgentModelItem[] = [
{ contextLimit: true, key: 'memoryAnalysisAgentConfig' },
{ contextLimit: true, key: 'userMemoryPersonaWriter' },
{ contextLimit: true, key: 'userMemoryEmbedding' },
];
const ModelAssignmentsForm = memo(() => {
const { t } = useTranslation('setting');
const [defaultAgent, systemAgentSettings] = useUserStore(
@@ -69,7 +76,7 @@ const ModelAssignmentsForm = memo(() => {
};
const updateSystemAgentModel = async (
key: UserSystemAgentConfigKey,
key: UserServiceModelConfigKey,
value: Partial<SystemAgentItem>,
) => {
setLoadingKey(key);
@@ -121,6 +128,39 @@ const ModelAssignmentsForm = memo(() => {
} satisfies FormItemProps;
});
const memoryModelItems: FormItemProps[] = MEMORY_MODEL_ITEMS.map(({ contextLimit, key }) => {
const value = systemAgentSettings[key];
return {
children: (
<Flexbox direction="vertical" gap={8} style={{ width: 'min(100%, 448px)' }}>
<ModelSelect
showAbility={false}
style={{ minWidth: 0, width: '100%' }}
value={value}
onChange={(props) => updateSystemAgentModel(key, props)}
/>
{contextLimit && (
<InputNumber
min={1}
placeholder={t('serviceModel.contextLimit.placeholder')}
style={{ alignSelf: 'flex-end', width: 180 }}
value={value.contextLimit}
onChange={(contextLimit) =>
updateSystemAgentModel(key, {
contextLimit: typeof contextLimit === 'number' ? contextLimit : undefined,
})
}
/>
)}
</Flexbox>
),
desc: t(`systemAgent.${key}.modelDesc`),
label: t(`systemAgent.${key}.title`),
minWidth: undefined,
} satisfies FormItemProps;
});
const optionalFeatureItems: FormItemProps[] = OPTIONAL_FEATURE_ITEMS.map(({ key }) => {
const value = systemAgentSettings[key];
const disabled = value.enabled === false;
@@ -167,7 +207,8 @@ const ModelAssignmentsForm = memo(() => {
loadingKey === 'followUpAction' ||
loadingKey === 'inputCompletion' ||
loadingKey === 'promptRewrite';
const isModelAssignmentLoading = loadingKey && !isOptionalFeatureLoading;
const isMemoryModelLoading = MEMORY_MODEL_ITEMS.some(({ key }) => loadingKey === key);
const isModelAssignmentLoading = loadingKey && !isOptionalFeatureLoading && !isMemoryModelLoading;
const modelAssignments: FormGroupItemType = {
children: [defaultAgentItem, ...systemModelItems],
@@ -185,10 +226,18 @@ const ModelAssignmentsForm = memo(() => {
title: t('serviceModel.optionalFeatures.title'),
};
const memoryModels: FormGroupItemType = {
children: memoryModelItems,
extra: isMemoryModelLoading && (
<Icon spin icon={Loader2Icon} size={16} style={{ opacity: 0.5 }} />
),
title: t('serviceModel.memoryModels.title'),
};
return (
<Form
collapsible={false}
items={[modelAssignments, optionalFeatures]}
items={[modelAssignments, memoryModels, optionalFeatures]}
itemsType={'group'}
variant={'filled'}
{...FORM_STYLE}
src/locales/default/setting.ts
+14
View File
@@ -812,6 +812,8 @@ export default {
'settingSystem.oauth.signout.success': 'Sign out successful',
'settingSystem.title': 'System Settings',
'serviceModel.modelAssignments.title': 'Model Assignments',
'serviceModel.contextLimit.placeholder': 'Context limit',
'serviceModel.memoryModels.title': 'Memory Models',
'serviceModel.optionalFeatures.title': 'Optional Features',
'settingSystemTools.appEnvironment.chromium.desc': 'Chromium browser engine version',
'settingSystemTools.appEnvironment.desc': 'Built-in runtime versions in the desktop app',
@@ -991,6 +993,18 @@ When I am ___, I need ___
'systemAgent.inputCompletion.modelDesc':
'Suggests text while you type. When enabled, this model generates the suggestions.',
'systemAgent.inputCompletion.title': 'Input Suggestions',
'systemAgent.userMemoryEmbedding.label': 'Model',
'systemAgent.userMemoryEmbedding.modelDesc':
'Model used to embed memory content for retrieval. The context limit caps each embedding input.',
'systemAgent.userMemoryEmbedding.title': 'Memory Embedding',
'systemAgent.memoryAnalysisAgentConfig.label': 'Model',
'systemAgent.memoryAnalysisAgentConfig.modelDesc':
'Model used to decide whether conversations contain memory and extract identities, preferences, contexts, activities, and experiences.',
'systemAgent.memoryAnalysisAgentConfig.title': 'Memory Analysis',
'systemAgent.userMemoryPersonaWriter.label': 'Model',
'systemAgent.userMemoryPersonaWriter.modelDesc':
'Model used to write persona-oriented memory summaries.',
'systemAgent.userMemoryPersonaWriter.title': 'Memory Persona Writer',
'systemAgent.promptRewrite.label': 'Model',
'systemAgent.promptRewrite.modelDesc':
'Improves prompts before generation. When enabled, this model rewrites the prompt.',
src/server/globalConfig/parseSystemAgent.test.ts
+19
View File
@@ -103,6 +103,25 @@ describe('parseSystemAgent', () => {
expect(result.agentMeta).toEqual({ provider: 'ollama', model: 'deepseek-v3' });
expect(result.historyCompress).toEqual({ provider: 'ollama', model: 'deepseek-v3' });
expect(result.thread).toEqual({ provider: 'ollama', model: 'deepseek-v3' });
expect(result.userMemoryEmbedding).toBeUndefined();
expect(result.memoryAnalysisAgentConfig).toBeUndefined();
expect(result.userMemoryPersonaWriter).toBeUndefined();
});
it('should parse memory service model assignments explicitly', () => {
const envValue =
'memoryAnalysisAgentConfig=lobehub/gpt-5.4-mini,userMemoryEmbedding=openai/text-embedding-3-large';
const result = parseSystemAgent(envValue);
expect(result.memoryAnalysisAgentConfig).toEqual({
provider: 'lobehub',
model: 'gpt-5.4-mini',
});
expect(result.userMemoryEmbedding).toEqual({
provider: 'openai',
model: 'text-embedding-3-large',
});
});
it('should override default setting with specific settings', () => {
src/server/globalConfig/parseSystemAgent.ts
+13 -7
View File
@@ -1,14 +1,20 @@
import { DEFAULT_SYSTEM_AGENT_CONFIG } from '@/const/settings';
import { type UserSystemAgentConfig } from '@/types/user/settings';
import { type UserServiceModelConfig } from '@/types/user/settings';
const protectedKeys = Object.keys(DEFAULT_SYSTEM_AGENT_CONFIG);
const defaultTrueLey = new Set(['promptRewrite', 'autoSuggestion']);
const memoryServiceModelKeys = new Set([
'memoryAnalysisAgentConfig',
'userMemoryEmbedding',
'userMemoryPersonaWriter',
]);
const defaultModelAssignmentKeys = protectedKeys.filter((key) => !memoryServiceModelKeys.has(key));
export const parseSystemAgent = (envString: string = ''): Partial<UserSystemAgentConfig> => {
export const parseSystemAgent = (envString: string = ''): Partial<UserServiceModelConfig> => {
if (!envString) return {};
const config: Partial<UserSystemAgentConfig> = {};
const config: Partial<UserServiceModelConfig> = {};
// Handle full-width commas and extra spaces
const envValue = envString.replaceAll('，', ',').trim();
@@ -39,7 +45,7 @@ export const parseSystemAgent = (envString: string = ''): Partial<UserSystemAgen
}
if (protectedKeys.includes(key)) {
config[key as keyof UserSystemAgentConfig] = {
config[key as keyof UserServiceModelConfig] = {
enabled: defaultTrueLey.has(key) ? true : undefined,
model: model.trim(),
provider: provider.trim(),
@@ -52,9 +58,9 @@ export const parseSystemAgent = (envString: string = ''): Partial<UserSystemAgen
// If there are default settings, apply them to all unconfigured system agents
if (defaultSetting) {
for (const key of protectedKeys) {
if (!config[key as keyof UserSystemAgentConfig]) {
config[key as keyof UserSystemAgentConfig] = {
for (const key of defaultModelAssignmentKeys) {
if (!config[key as keyof UserServiceModelConfig]) {
config[key as keyof UserServiceModelConfig] = {
enabled: defaultTrueLey.has(key) ? true : undefined,
model: defaultSetting.model,
provider: defaultSetting.provider,
src/server/services/memory/userMemory/__tests__/extract.runtime.test.ts
+200 -5
View File
@@ -57,7 +57,202 @@ const createExecutor = (privateOverrides?: Partial<MemoryExtractionPrivateConfig
});
};
const resolveRuntimeKeyVaults = async (
executor: MemoryExtractionExecutor,
runtimeState: AiProviderRuntimeState,
) => {
const memoryServiceConfig = (executor as any).resolveUserMemoryServiceConfig();
return (executor as any).resolveRuntimeKeyVaults(runtimeState, memoryServiceConfig);
};
describe('MemoryExtractionExecutor.resolveRuntimeKeyVaults', () => {
it('drops fallback credentials when user memory provider is overridden', () => {
const executor = createExecutor({
embedding: {
apiKey: 'openai-system-key',
baseURL: 'https://openai.example.com',
model: 'embed-1',
provider: 'openai',
},
});
const memoryServiceConfig = (executor as any).resolveUserMemoryServiceConfig({
userMemoryEmbedding: {
model: 'embed-2',
provider: 'anthropic',
},
});
expect(memoryServiceConfig.agents.embedding).toMatchObject({
model: 'embed-2',
provider: 'anthropic',
});
expect(memoryServiceConfig.agents.embedding.apiKey).toBeUndefined();
expect(memoryServiceConfig.agents.embedding.baseURL).toBeUndefined();
});
it('keeps fallback credentials when user memory provider is unchanged', () => {
const executor = createExecutor({
embedding: {
apiKey: 'openai-system-key',
baseURL: 'https://openai.example.com',
model: 'embed-1',
provider: 'openai',
},
});
const memoryServiceConfig = (executor as any).resolveUserMemoryServiceConfig({
userMemoryEmbedding: {
model: 'embed-2',
provider: 'openai',
},
});
expect(memoryServiceConfig.agents.embedding).toMatchObject({
apiKey: 'openai-system-key',
baseURL: 'https://openai.example.com',
model: 'embed-2',
provider: 'openai',
});
});
it('shares ServiceModel memory analysis config between gatekeeper and layer extractor', () => {
const executor = createExecutor({
agentGateKeeper: {
apiKey: 'gate-system-key',
baseURL: 'https://gate.example.com',
model: 'gate-1',
provider: 'provider-gate',
},
agentLayerExtractor: {
apiKey: 'layer-system-key',
baseURL: 'https://layer.example.com',
contextLimit: 2048,
layers: {
activity: 'layer-act',
context: 'layer-ctx',
experience: 'layer-exp',
identity: 'layer-id',
preference: 'layer-pref',
},
model: 'layer-1',
provider: 'provider-layer',
},
});
const memoryServiceConfig = (executor as any).resolveUserMemoryServiceConfig({
memoryAnalysisAgentConfig: {
contextLimit: 4096,
model: 'analysis-1',
provider: 'provider-analysis',
},
});
expect(memoryServiceConfig.agents.gatekeeper).toMatchObject({
model: 'analysis-1',
provider: 'provider-analysis',
});
expect(memoryServiceConfig.agents.layerExtractor).toMatchObject({
contextLimit: 4096,
model: 'analysis-1',
provider: 'provider-analysis',
});
expect(memoryServiceConfig.agents.gatekeeper.apiKey).toBeUndefined();
expect(memoryServiceConfig.agents.layerExtractor.apiKey).toBeUndefined();
expect(memoryServiceConfig.modelConfig.gateModel).toBe('analysis-1');
expect(memoryServiceConfig.modelConfig.layerModels).toEqual({
activity: 'analysis-1',
context: 'analysis-1',
experience: 'analysis-1',
identity: 'analysis-1',
preference: 'analysis-1',
});
});
it('uses ServiceModel provider before env preferred providers when provider is overridden', async () => {
const executor = createExecutor({
agentGateKeeper: {
model: 'gate-1',
provider: 'provider-g',
},
agentLayerExtractor: {
contextLimit: 2048,
layers: {
activity: 'layer-1',
context: 'layer-1',
experience: 'layer-1',
identity: 'layer-1',
preference: 'layer-1',
},
model: 'layer-1',
provider: 'provider-l',
},
embedding: {
apiKey: 'openai-system-key',
baseURL: 'https://openai.example.com',
model: 'embed-1',
provider: 'openai',
},
embeddingPreferredProviders: ['provider-b'],
});
const memoryServiceConfig = (executor as any).resolveUserMemoryServiceConfig({
userMemoryEmbedding: {
model: 'embed-2',
provider: 'provider-a',
},
});
const runtimeState = createRuntimeState(
[
{
abilities: {},
enabled: true,
id: 'gate-1',
providerId: 'provider-g',
type: 'chat',
},
{
abilities: {},
enabled: true,
id: 'layer-1',
providerId: 'provider-l',
type: 'chat',
},
{
abilities: {},
enabled: true,
id: 'embed-2',
providerId: 'provider-a',
type: 'embedding',
},
{
abilities: {},
enabled: true,
id: 'embed-2',
providerId: 'provider-b',
type: 'embedding',
},
],
{
'provider-a': { apiKey: 'a-key' },
'provider-b': { apiKey: 'b-key' },
'provider-g': { apiKey: 'g-key' },
'provider-l': { apiKey: 'l-key' },
},
);
const keyVaults = await (executor as any).resolveRuntimeKeyVaults(
runtimeState,
memoryServiceConfig,
);
expect(keyVaults).toMatchObject({
'provider-a': { apiKey: 'a-key' },
});
expect(keyVaults).not.toHaveProperty('provider-b');
});
it('prefers configured providers/models for gatekeeper, embedding, and layer extractors', async () => {
const executor = createExecutor({
embeddingPreferredProviders: ['provider-c', 'provider-a'],
@@ -119,7 +314,7 @@ describe('MemoryExtractionExecutor.resolveRuntimeKeyVaults', () => {
},
);
const keyVaults = await (executor as any).resolveRuntimeKeyVaults(runtimeState);
const keyVaults = await resolveRuntimeKeyVaults(executor, runtimeState);
expect(keyVaults).toMatchObject({
'provider-a': { apiKey: 'a-key' },
@@ -182,7 +377,7 @@ describe('MemoryExtractionExecutor.resolveRuntimeKeyVaults', () => {
},
);
const keyVaults = await (executor as any).resolveRuntimeKeyVaults(runtimeState);
const keyVaults = await resolveRuntimeKeyVaults(executor, runtimeState);
expect(keyVaults).toMatchObject({
'provider-b': { apiKey: 'b-key' },
@@ -222,7 +417,7 @@ describe('MemoryExtractionExecutor.resolveRuntimeKeyVaults', () => {
},
);
const keyVaults = await (executor as any).resolveRuntimeKeyVaults(runtimeState);
const keyVaults = await resolveRuntimeKeyVaults(executor, runtimeState);
expect(keyVaults).toMatchObject({
'provider-a': { apiKey: 'a-key' },
@@ -253,7 +448,7 @@ describe('MemoryExtractionExecutor.resolveRuntimeKeyVaults', () => {
},
);
const keyVaults = await (executor as any).resolveRuntimeKeyVaults(runtimeState);
const keyVaults = await resolveRuntimeKeyVaults(executor, runtimeState);
expect(keyVaults).toMatchObject({
'provider-b': { apiKey: 'b-key' }, // picks first preferred provider
@@ -271,7 +466,7 @@ describe('MemoryExtractionExecutor.resolveRuntimeKeyVaults', () => {
'provider-fallback': { apiKey: 'fb-key' },
});
const keyVaults = await (executor as any).resolveRuntimeKeyVaults(runtimeState);
const keyVaults = await resolveRuntimeKeyVaults(executor, runtimeState);
expect(keyVaults).toMatchObject({
'provider-fallback': { apiKey: 'fb-key' },
src/server/services/memory/userMemory/extract.ts
+217 -57
View File
@@ -45,6 +45,7 @@ import type {
MemoryExtractionAgentCallTrace,
MemoryExtractionTraceError,
MemoryExtractionTracePayload,
UserServiceModelConfig,
} from '@lobechat/types';
import { RequestTrigger } from '@lobechat/types';
import { type FlowControl } from '@upstash/qstash';
@@ -583,6 +584,38 @@ type RuntimeBundle = {
layerExtractor: ModelRuntime;
};
interface MemoryExtractionModelConfig {
embeddingsModel: string;
gateModel: string;
layerModels: Partial<Record<LayersEnum, string>>;
observabilityS3: MemoryExtractionConfig['observabilityS3'];
}
interface ResolvedMemoryServiceConfig {
agents: {
embedding: MemoryAgentConfig;
gatekeeper: MemoryAgentConfig;
layerExtractor: MemoryAgentConfig;
};
embeddingContextLimit?: number;
extractorContextLimit?: number;
modelConfig: MemoryExtractionModelConfig;
overrides: {
embedding: {
model: boolean;
provider: boolean;
};
gatekeeper: {
model: boolean;
provider: boolean;
};
layerExtractor: {
model: boolean;
provider: boolean;
};
};
}
export interface TopicExtractionJob {
asyncTaskId?: string;
forceAll: boolean;
@@ -623,14 +656,7 @@ export class MemoryExtractionExecutor {
private readonly layerPreferredModels?: string[];
private readonly layerPreferredProviders?: string[];
private readonly privateConfig: MemoryExtractionConfig;
private readonly modelConfig: {
embeddingsModel: string;
gateModel: string;
layerModels: Partial<Record<LayersEnum, string>>;
observabilityS3: MemoryExtractionConfig['observabilityS3'];
};
private readonly embeddingContextLimit?: number;
private readonly modelConfig: MemoryExtractionModelConfig;
private readonly runtimeCache = new Map<string, RuntimeBundle>();
private readonly db = getServerDB();
@@ -659,9 +685,6 @@ export class MemoryExtractionExecutor {
),
observabilityS3: privateConfig.observabilityS3,
};
this.embeddingContextLimit =
privateConfig.embedding?.contextLimit ?? privateConfig.agentLayerExtractor.contextLimit;
}
static async create() {
@@ -673,6 +696,95 @@ export class MemoryExtractionExecutor {
return new MemoryExtractionExecutor(serverConfig, privateConfig);
}
private resolveUserMemoryAgent(
systemAgent: Partial<UserServiceModelConfig> | undefined,
key: keyof Pick<
UserServiceModelConfig,
'userMemoryEmbedding' | 'memoryAnalysisAgentConfig' | 'userMemoryPersonaWriter'
>,
fallback: MemoryAgentConfig,
): MemoryAgentConfig {
const override = systemAgent?.[key];
const provider = override?.provider || fallback.provider;
const shouldInheritCredentials =
!override?.provider ||
normalizeProvider(override.provider) === normalizeProvider(fallback.provider || 'openai');
const contextLimit =
typeof override?.contextLimit === 'number' &&
Number.isFinite(override.contextLimit) &&
override.contextLimit > 0
? Math.floor(override.contextLimit)
: fallback.contextLimit;
return {
apiKey: shouldInheritCredentials ? fallback.apiKey : undefined,
baseURL: shouldInheritCredentials ? fallback.baseURL : undefined,
contextLimit,
language: fallback.language,
model: override?.model || fallback.model,
provider,
};
}
private resolveUserMemoryServiceConfig(
systemAgent?: Partial<UserServiceModelConfig>,
): ResolvedMemoryServiceConfig {
const gatekeeper = this.resolveUserMemoryAgent(
systemAgent,
'memoryAnalysisAgentConfig',
this.privateConfig.agentGateKeeper,
);
const layerExtractor = this.resolveUserMemoryAgent(
systemAgent,
'memoryAnalysisAgentConfig',
this.privateConfig.agentLayerExtractor,
);
const embedding = this.resolveUserMemoryAgent(
systemAgent,
'userMemoryEmbedding',
this.privateConfig.embedding,
);
const layerModels = systemAgent?.memoryAnalysisAgentConfig?.model
? {
activity: layerExtractor.model,
context: layerExtractor.model,
experience: layerExtractor.model,
identity: layerExtractor.model,
preference: layerExtractor.model,
}
: resolveLayerModels(undefined, this.privateConfig.agentLayerExtractor.layers);
return {
agents: {
embedding,
gatekeeper,
layerExtractor,
},
embeddingContextLimit: embedding.contextLimit ?? layerExtractor.contextLimit,
extractorContextLimit: layerExtractor.contextLimit,
modelConfig: {
embeddingsModel: embedding.model,
gateModel: gatekeeper.model,
layerModels,
observabilityS3: this.privateConfig.observabilityS3,
},
overrides: {
embedding: {
model: Boolean(systemAgent?.userMemoryEmbedding?.model),
provider: Boolean(systemAgent?.userMemoryEmbedding?.provider),
},
gatekeeper: {
model: Boolean(systemAgent?.memoryAnalysisAgentConfig?.model),
provider: Boolean(systemAgent?.memoryAnalysisAgentConfig?.provider),
},
layerExtractor: {
model: Boolean(systemAgent?.memoryAnalysisAgentConfig?.model),
provider: Boolean(systemAgent?.memoryAnalysisAgentConfig?.provider),
},
},
};
}
private buildBaseMetadata(
job: MemoryExtractionJob,
messageIds: string[],
@@ -1431,10 +1543,16 @@ export class MemoryExtractionExecutor {
userModel.getUserState(KeyVaultsGateKeeper.getUserKeyVaults),
this.getAiProviderRuntimeState(job.userId),
]);
const keyVaults = await this.resolveRuntimeKeyVaults(aiProviderRuntimeState);
const memoryServiceConfig = this.resolveUserMemoryServiceConfig(
userState.settings?.systemAgent as Partial<UserServiceModelConfig> | undefined,
);
const keyVaults = await this.resolveRuntimeKeyVaults(
aiProviderRuntimeState,
memoryServiceConfig,
);
const language = userState.settings?.general?.responseLanguage;
const runtimes = await this.getRuntime(job.userId, keyVaults);
const runtimes = await this.getRuntime(job.userId, memoryServiceConfig, keyVaults);
const conversations = await this.listConversationsForTopic(
job.userId,
@@ -1454,8 +1572,9 @@ export class MemoryExtractionExecutor {
};
}
const extractorContextLimit = this.privateConfig.agentLayerExtractor.contextLimit;
const embeddingContextLimit = this.embeddingContextLimit ?? extractorContextLimit;
const extractorContextLimit = memoryServiceConfig.extractorContextLimit;
const embeddingContextLimit =
memoryServiceConfig.embeddingContextLimit ?? extractorContextLimit;
const extractorConversations = await this.trimConversationsToTokenLimit(
conversations,
extractorContextLimit,
@@ -1508,7 +1627,7 @@ export class MemoryExtractionExecutor {
searchResult = await this.listRelevantUserMemories(
extractionJob,
runtimes.embeddings,
this.modelConfig.embeddingsModel,
memoryServiceConfig.modelConfig.embeddingsModel,
job.userId,
embeddingConversations,
embeddingContextLimit,
@@ -1621,7 +1740,7 @@ export class MemoryExtractionExecutor {
};
const service = new MemoryExtractionService({
config: this.modelConfig,
config: memoryServiceConfig.modelConfig,
db,
language,
runtimes,
@@ -1676,6 +1795,7 @@ export class MemoryExtractionExecutor {
messageIds,
extraction,
runtimes,
memoryServiceConfig,
db,
);
if (retrievalErrors.length > 0) {
@@ -2023,6 +2143,7 @@ export class MemoryExtractionExecutor {
messageIds: string[],
extraction: MemoryExtractionResult,
runtimes: RuntimeBundle,
memoryServiceConfig: ResolvedMemoryServiceConfig,
db: Awaited<ReturnType<typeof getServerDB>>,
): Promise<PersistedMemoryResult> {
const createdIds: string[] = [];
@@ -2108,8 +2229,8 @@ export class MemoryExtractionExecutor {
messageIds,
activityOutput.data,
runtimes.embeddings,
this.modelConfig.embeddingsModel,
this.embeddingContextLimit,
memoryServiceConfig.modelConfig.embeddingsModel,
memoryServiceConfig.embeddingContextLimit,
db,
),
);
@@ -2126,8 +2247,8 @@ export class MemoryExtractionExecutor {
messageIds,
contextOutput.data,
runtimes.embeddings,
this.modelConfig.embeddingsModel,
this.embeddingContextLimit,
memoryServiceConfig.modelConfig.embeddingsModel,
memoryServiceConfig.embeddingContextLimit,
db,
),
);
@@ -2144,8 +2265,8 @@ export class MemoryExtractionExecutor {
messageIds,
experienceOutput.data,
runtimes.embeddings,
this.modelConfig.embeddingsModel,
this.embeddingContextLimit,
memoryServiceConfig.modelConfig.embeddingsModel,
memoryServiceConfig.embeddingContextLimit,
db,
),
);
@@ -2162,8 +2283,8 @@ export class MemoryExtractionExecutor {
messageIds,
preferenceOutput.data,
runtimes.embeddings,
this.modelConfig.embeddingsModel,
this.embeddingContextLimit,
memoryServiceConfig.modelConfig.embeddingsModel,
memoryServiceConfig.embeddingContextLimit,
db,
),
);
@@ -2180,8 +2301,8 @@ export class MemoryExtractionExecutor {
messageIds,
identityOutput.data,
runtimes.embeddings,
this.modelConfig.embeddingsModel,
this.embeddingContextLimit,
memoryServiceConfig.modelConfig.embeddingsModel,
memoryServiceConfig.embeddingContextLimit,
db,
),
);
@@ -2210,6 +2331,7 @@ export class MemoryExtractionExecutor {
private async resolveRuntimeKeyVaults(
runtimeState: AiProviderRuntimeState,
memoryServiceConfig: ResolvedMemoryServiceConfig,
): Promise<ProviderKeyVaultMap> {
const normalizedRuntimeConfig = Object.fromEntries(
Object.entries(runtimeState.runtimeConfig || {}).map(([providerId, config]) => [
@@ -2221,11 +2343,15 @@ export class MemoryExtractionExecutor {
const keyVaults: ProviderKeyVaultMap = {};
const gatekeeperProvider = await AiInfraRepos.tryMatchingProviderFrom(runtimeState, {
fallbackProvider: this.privateConfig.agentGateKeeper.provider,
fallbackProvider: memoryServiceConfig.agents.gatekeeper.provider,
label: 'gatekeeper',
modelId: this.modelConfig.gateModel,
preferredModels: this.gatekeeperPreferredModels,
preferredProviders: this.gatekeeperPreferredProviders,
modelId: memoryServiceConfig.modelConfig.gateModel,
preferredModels: memoryServiceConfig.overrides.gatekeeper.model
? undefined
: this.gatekeeperPreferredModels,
preferredProviders: memoryServiceConfig.overrides.gatekeeper.provider
? undefined
: this.gatekeeperPreferredProviders,
});
const gatekeeperRuntime = normalizedRuntimeConfig[gatekeeperProvider];
if (gatekeeperRuntime?.keyVaults) {
@@ -2233,25 +2359,33 @@ export class MemoryExtractionExecutor {
}
const embeddingProvider = await AiInfraRepos.tryMatchingProviderFrom(runtimeState, {
fallbackProvider: this.privateConfig.embedding.provider,
fallbackProvider: memoryServiceConfig.agents.embedding.provider,
label: 'embedding',
modelId: this.modelConfig.embeddingsModel,
preferredModels: this.embeddingPreferredModels,
preferredProviders: this.embeddingPreferredProviders,
modelId: memoryServiceConfig.modelConfig.embeddingsModel,
preferredModels: memoryServiceConfig.overrides.embedding.model
? undefined
: this.embeddingPreferredModels,
preferredProviders: memoryServiceConfig.overrides.embedding.provider
? undefined
: this.embeddingPreferredProviders,
});
const embeddingRuntime = normalizedRuntimeConfig[embeddingProvider];
if (embeddingRuntime?.keyVaults) {
keyVaults[embeddingProvider] = embeddingRuntime.keyVaults;
}
for (const model of Object.values(this.modelConfig.layerModels)) {
for (const model of Object.values(memoryServiceConfig.modelConfig.layerModels)) {
if (!model) continue;
const providerId = await AiInfraRepos.tryMatchingProviderFrom(runtimeState, {
fallbackProvider: this.privateConfig.agentLayerExtractor.provider,
fallbackProvider: memoryServiceConfig.agents.layerExtractor.provider,
label: 'layer extractor',
modelId: model,
preferredModels: this.layerPreferredModels,
preferredProviders: this.layerPreferredProviders,
preferredModels: memoryServiceConfig.overrides.layerExtractor.model
? undefined
: this.layerPreferredModels,
preferredProviders: memoryServiceConfig.overrides.layerExtractor.provider
? undefined
: this.layerPreferredProviders,
});
const runtime = normalizedRuntimeConfig[providerId];
if (runtime?.keyVaults) {
@@ -2264,6 +2398,7 @@ export class MemoryExtractionExecutor {
private async getRuntime(
userId: string,
memoryServiceConfig: ResolvedMemoryServiceConfig,
keyVaults?: ProviderKeyVaultMap,
): Promise<RuntimeBundle> {
// TODO: implement a better cache eviction strategy
@@ -2272,33 +2407,51 @@ export class MemoryExtractionExecutor {
this.runtimeCache.clear();
}
const cached = this.runtimeCache.get(userId);
const cacheKey = [
userId,
memoryServiceConfig.agents.embedding.provider,
memoryServiceConfig.agents.gatekeeper.provider,
memoryServiceConfig.agents.layerExtractor.provider,
].join(':');
const cached = this.runtimeCache.get(cacheKey);
if (cached) return cached;
const embeddingOptions: RuntimeResolveOptions = {
fallback: {
apiKey: this.privateConfig.embedding.apiKey,
baseURL: this.privateConfig.embedding.baseURL,
apiKey: memoryServiceConfig.agents.embedding.apiKey,
baseURL: memoryServiceConfig.agents.embedding.baseURL,
},
preferred: {
providerIds: memoryServiceConfig.overrides.embedding.provider
? undefined
: this.embeddingPreferredProviders,
},
preferred: { providerIds: this.embeddingPreferredProviders },
userId,
};
const gatekeeperOptions: RuntimeResolveOptions = {
fallback: {
apiKey: this.privateConfig.agentGateKeeper.apiKey,
baseURL: this.privateConfig.agentGateKeeper.baseURL,
apiKey: memoryServiceConfig.agents.gatekeeper.apiKey,
baseURL: memoryServiceConfig.agents.gatekeeper.baseURL,
},
preferred: {
providerIds: memoryServiceConfig.overrides.gatekeeper.provider
? undefined
: this.gatekeeperPreferredProviders,
},
preferred: { providerIds: this.gatekeeperPreferredProviders },
userId,
};
const layerExtractorOptions: RuntimeResolveOptions = {
fallback: {
apiKey: this.privateConfig.agentLayerExtractor.apiKey,
baseURL: this.privateConfig.agentLayerExtractor.baseURL,
apiKey: memoryServiceConfig.agents.layerExtractor.apiKey,
baseURL: memoryServiceConfig.agents.layerExtractor.baseURL,
},
preferred: {
providerIds: memoryServiceConfig.overrides.layerExtractor.provider
? undefined
: this.layerPreferredProviders,
},
preferred: { providerIds: this.layerPreferredProviders },
userId,
};
@@ -2306,26 +2459,26 @@ export class MemoryExtractionExecutor {
const runtimes: RuntimeBundle = {
embeddings: await resolveRuntimeAgentConfig(
{ ...this.privateConfig.embedding },
memoryServiceConfig.agents.embedding,
keyVaults,
embeddingOptions,
hooks,
),
gatekeeper: await resolveRuntimeAgentConfig(
{ ...this.privateConfig.agentGateKeeper },
memoryServiceConfig.agents.gatekeeper,
keyVaults,
gatekeeperOptions,
hooks,
),
layerExtractor: await resolveRuntimeAgentConfig(
{ ...this.privateConfig.agentLayerExtractor },
memoryServiceConfig.agents.layerExtractor,
keyVaults,
layerExtractorOptions,
hooks,
),
};
this.runtimeCache.set(userId, runtimes);
this.runtimeCache.set(cacheKey, runtimes);
return runtimes;
}
@@ -2361,10 +2514,16 @@ export class MemoryExtractionExecutor {
userModel.getUserState(KeyVaultsGateKeeper.getUserKeyVaults),
this.getAiProviderRuntimeState(params.userId),
]);
const keyVaults = await this.resolveRuntimeKeyVaults(aiProviderRuntimeState);
const memoryServiceConfig = this.resolveUserMemoryServiceConfig(
userState.settings?.systemAgent as Partial<UserServiceModelConfig> | undefined,
);
const keyVaults = await this.resolveRuntimeKeyVaults(
aiProviderRuntimeState,
memoryServiceConfig,
);
const language = params.language || userState.settings?.general?.responseLanguage;
const runtimes = await this.getRuntime(params.userId, keyVaults);
const runtimes = await this.getRuntime(params.userId, memoryServiceConfig, keyVaults);
const contextProvider =
params.contextProvider ||
new BenchmarkLocomoContextProvider({
@@ -2393,7 +2552,7 @@ export class MemoryExtractionExecutor {
};
const builtContext = await contextProvider.buildContext(extractionJob.userId);
const extractorContextLimit = this.privateConfig.agentLayerExtractor.contextLimit;
const extractorContextLimit = memoryServiceConfig.extractorContextLimit;
const trimmedContext = await this.trimTextToTokenLimit(
builtContext.context,
extractorContextLimit,
@@ -2431,7 +2590,7 @@ export class MemoryExtractionExecutor {
};
const service = new MemoryExtractionService({
config: this.modelConfig,
config: memoryServiceConfig.modelConfig,
db,
language,
runtimes,
@@ -2476,6 +2635,7 @@ export class MemoryExtractionExecutor {
[],
extraction,
runtimes,
memoryServiceConfig,
db,
);
src/server/services/memory/userMemory/persona/__tests__/service.test.ts
+44 -1
View File
@@ -1,11 +1,12 @@
// @vitest-environment node
import { type LobeChatDatabase } from '@lobechat/database';
import { users } from '@lobechat/database/schemas';
import { users, userSettings } from '@lobechat/database/schemas';
import { getTestDB } from '@lobechat/database/test-utils';
import { beforeEach, describe, expect, it, vi } from 'vitest';
import { UserPersonaModel } from '@/database/models/userMemory/persona';
import type * as AiInfraReposModule from '@/database/repositories/aiInfra';
import { resolveRuntimeAgentConfig } from '@/server/services/memory/userMemory/extract';
import { UserPersonaService } from '../service';
@@ -134,4 +135,46 @@ describe('UserPersonaService', () => {
}),
);
});
it('drops fallback credentials when persona writer provider is overridden', async () => {
await db.insert(userSettings).values({
id: userId,
systemAgent: {
userMemoryPersonaWriter: {
model: 'claude-mock',
provider: 'anthropic',
},
},
});
aiInfraMocks.tryMatchingProviderFrom.mockResolvedValue('anthropic');
aiInfraMocks.getAiProviderRuntimeState.mockResolvedValue({
enabledAiModels: [
{ abilities: {}, enabled: true, id: 'claude-mock', providerId: 'anthropic', type: 'chat' },
],
enabledAiProviders: [],
enabledChatAiProviders: [],
enabledImageAiProviders: [],
runtimeConfig: {},
});
const service = new UserPersonaService(db);
await service.composeWriting({ userId, username: 'User' });
expect(resolveRuntimeAgentConfig).toHaveBeenLastCalledWith(
expect.objectContaining({
apiKey: undefined,
baseURL: undefined,
model: 'claude-mock',
provider: 'anthropic',
}),
expect.any(Object),
expect.objectContaining({
fallback: {
apiKey: undefined,
baseURL: undefined,
},
}),
undefined,
);
});
});
src/server/services/memory/userMemory/persona/service.ts
+49 -7
View File
@@ -9,9 +9,11 @@ import {
RetrievalUserMemoryIdentitiesProvider,
UserPersonaExtractor,
} from '@lobechat/memory-user-memory';
import type { UserServiceModelConfig } from '@lobechat/types';
import { desc, eq } from 'drizzle-orm';
import { getBusinessModelRuntimeHooks } from '@/business/server/model-runtime';
import { UserModel } from '@/database/models/user';
import { UserMemoryModel } from '@/database/models/userMemory';
import { UserPersonaModel } from '@/database/models/userMemory/persona';
import { AiInfraRepos } from '@/database/repositories/aiInfra';
@@ -47,6 +49,14 @@ interface UserPersonaAgentResult {
document: UserPersonaDocument;
}
const resolvePositiveInteger = (value?: number) => {
if (typeof value !== 'number' || !Number.isFinite(value) || value <= 0) return undefined;
return Math.floor(value);
};
const normalizeProvider = (provider: string) => provider.toLowerCase();
export class UserPersonaService {
private readonly preferredLanguage?: string;
private readonly db: LobeChatDatabase;
@@ -60,15 +70,40 @@ export class UserPersonaService {
this.agentConfig = agentPersonaWriter;
}
private async resolveAgentConfig(userId: string): Promise<MemoryAgentConfig> {
const userModel = new UserModel(this.db, userId);
const settings = await userModel.getUserSettings();
const userMemoryPersonaWriter = (
settings?.systemAgent as Partial<UserServiceModelConfig> | undefined
)?.userMemoryPersonaWriter;
const provider = userMemoryPersonaWriter?.provider || this.agentConfig.provider;
const shouldInheritCredentials =
!userMemoryPersonaWriter?.provider ||
normalizeProvider(userMemoryPersonaWriter.provider) ===
normalizeProvider(this.agentConfig.provider || 'openai');
return {
apiKey: shouldInheritCredentials ? this.agentConfig.apiKey : undefined,
baseURL: shouldInheritCredentials ? this.agentConfig.baseURL : undefined,
contextLimit:
resolvePositiveInteger(userMemoryPersonaWriter?.contextLimit) ??
this.agentConfig.contextLimit,
language: this.agentConfig.language,
model: userMemoryPersonaWriter?.model || this.agentConfig.model,
provider,
};
}
async composeWriting(payload: UserPersonaAgentPayload): Promise<UserPersonaAgentResult> {
const agentConfig = await this.resolveAgentConfig(payload.userId);
const aiInfraRepos = new AiInfraRepos(this.db, payload.userId, {});
const runtimeState = await aiInfraRepos.getAiProviderRuntimeState(
KeyVaultsGateKeeper.getUserKeyVaults,
);
const providerId = await AiInfraRepos.tryMatchingProviderFrom(runtimeState, {
fallbackProvider: this.agentConfig.provider,
fallbackProvider: agentConfig.provider,
label: 'persona writer',
modelId: this.agentConfig.model,
modelId: agentConfig.model,
});
const keyVaults: ProviderKeyVaultMap = Object.entries(runtimeState.runtimeConfig || {}).reduce(
@@ -82,12 +117,12 @@ export class UserPersonaService {
const hooks = getBusinessModelRuntimeHooks(payload.userId, 'lobehub');
const runtime = await resolveRuntimeAgentConfig(
{ ...this.agentConfig },
agentConfig,
keyVaults,
{
fallback: {
apiKey: this.agentConfig.apiKey,
baseURL: this.agentConfig.baseURL,
apiKey: agentConfig.apiKey,
baseURL: agentConfig.baseURL,
},
preferred: { providerIds: [providerId] },
userId: payload.userId,
@@ -101,7 +136,7 @@ export class UserPersonaService {
const extractor = new UserPersonaExtractor({
agent: 'user-persona',
model: this.agentConfig.model,
model: agentConfig.model,
modelRuntime: runtime,
});
@@ -136,7 +171,14 @@ export const buildUserPersonaJobInput = async (db: LobeChatDatabase, userId: str
const personaModel = new UserPersonaModel(db, userId);
const latestPersona = await personaModel.getLatestPersonaDocument();
const { agentPersonaWriter } = parseMemoryExtractionConfig();
const personaContextLimit = agentPersonaWriter.contextLimit;
const userModel = new UserModel(db, userId);
const settings = await userModel.getUserSettings();
const userMemoryPersonaWriter = (
settings?.systemAgent as Partial<UserServiceModelConfig> | undefined
)?.userMemoryPersonaWriter;
const personaContextLimit =
resolvePositiveInteger(userMemoryPersonaWriter?.contextLimit) ??
agentPersonaWriter.contextLimit;
const userMemoryModel = new UserMemoryModel(db, userId);
src/store/user/slices/settings/action.ts
+2 -1
View File
@@ -11,6 +11,7 @@ import type {
SystemAgentItem,
UserGeneralConfig,
UserKeyVaults,
UserServiceModelConfigKey,
UserSettings,
UserSystemAgentConfigKey,
} from '@/types/user/settings';
@@ -237,7 +238,7 @@ export class UserSettingsActionImpl {
};
updateSystemAgent = async (
key: UserSystemAgentConfigKey,
key: UserServiceModelConfigKey,
value: Partial<SystemAgentItem>,
): Promise<void> => {
await this.#get().setSettings({
src/store/user/slices/settings/selectors/__snapshots__/settings.test.ts.snap
+12
View File
@@ -73,6 +73,10 @@ exports[`settingsSelectors > currentSystemAgent > should merge DEFAULT_SYSTEM_AG
"model": "gpt-5.4-mini",
"provider": "openai",
},
"memoryAnalysisAgentConfig": {
"model": "gpt-5.4-mini",
"provider": "openai",
},
"promptRewrite": {
"enabled": true,
"model": "gpt-5.4-mini",
@@ -91,6 +95,14 @@ exports[`settingsSelectors > currentSystemAgent > should merge DEFAULT_SYSTEM_AG
"model": "gpt-5.4-mini",
"provider": "openai",
},
"userMemoryEmbedding": {
"model": "text-embedding-3-small",
"provider": "openai",
},
"userMemoryPersonaWriter": {
"model": "gpt-5.4-mini",
"provider": "openai",
},
}
`;
vitest.config.mts
+8
View File
@@ -16,6 +16,14 @@ const alias = {
__dirname,
'./packages/business/model-runtime/src/index.ts',
),
'@lobechat/business-model-bank/model-config': resolve(
__dirname,
'./packages/business/model-bank/src/model-config.ts',
),
'@lobechat/business-model-bank': resolve(
__dirname,
'./packages/business/model-bank/src/index.ts',
),
'@emoji-mart/data': resolve(__dirname, './tests/mocks/emojiMartData.ts'),
'@emoji-mart/react': resolve(__dirname, './tests/mocks/emojiMartReact.tsx'),
'@/database/_deprecated': resolve(__dirname, './src/database/_deprecated'),