mirror of
https://github.com/infiniflow/ragflow.git
synced 2026-04-21 19:27:47 +08:00
feat: Adds the tenant model ID field to the interface definition. (#13274)
### What problem does this PR solve? feat: Adds the tenant model ID field to the interface definition ### Type of change - [x] New Feature (non-breaking change which adds functionality)
This commit is contained in:
@ -108,6 +108,7 @@ export interface IGenerateForm {
|
||||
cite?: boolean;
|
||||
prompt: number;
|
||||
llm_id: string;
|
||||
tenant_llm_id?: string;
|
||||
parameters: { key: string; component_id: string };
|
||||
}
|
||||
|
||||
@ -143,6 +144,7 @@ export interface IRetrievalForm {
|
||||
top_n?: number;
|
||||
top_k?: number;
|
||||
rerank_id?: string;
|
||||
tenant_rerank_id?: string;
|
||||
empty_response?: string;
|
||||
kb_ids: string[];
|
||||
}
|
||||
|
||||
@ -35,6 +35,7 @@ export interface Variable {
|
||||
temperature?: number;
|
||||
top_p?: number;
|
||||
llm_id?: string;
|
||||
tenant_llm_id?: string;
|
||||
}
|
||||
|
||||
export interface IDialog {
|
||||
@ -48,6 +49,7 @@ export interface IDialog {
|
||||
kb_names: string[];
|
||||
language: string;
|
||||
llm_id: string;
|
||||
tenant_llm_id?: string;
|
||||
llm_setting: Variable;
|
||||
llm_setting_type: string;
|
||||
name: string;
|
||||
|
||||
@ -117,6 +117,13 @@ export interface ITenantInfo {
|
||||
speech2text_id: string;
|
||||
rerank_id?: string;
|
||||
tts_id: string;
|
||||
// Tenant model IDs
|
||||
tenant_asr_id?: string;
|
||||
tenant_embd_id?: string;
|
||||
tenant_img2txt_id?: string;
|
||||
tenant_llm_id?: string;
|
||||
tenant_rerank_id?: string;
|
||||
tenant_tts_id?: string;
|
||||
}
|
||||
|
||||
export type ChunkDocType = 'image' | 'table' | 'text';
|
||||
|
||||
59
web/src/utils/llm-cache.ts
Normal file
59
web/src/utils/llm-cache.ts
Normal file
@ -0,0 +1,59 @@
|
||||
// LLM list cache utility
|
||||
|
||||
interface LlmCache {
|
||||
data: Record<string, any>;
|
||||
timestamp: number;
|
||||
}
|
||||
|
||||
const CACHE_KEY = 'ragflow_llm_list_cache';
|
||||
const CACHE_DURATION = 5 * 60 * 1000; // 5 minutes
|
||||
|
||||
// Get cached LLM list
|
||||
export function getCachedLlmList(): Record<string, any> | null {
|
||||
try {
|
||||
const cached = localStorage.getItem(CACHE_KEY);
|
||||
if (!cached) return null;
|
||||
|
||||
const parsed: LlmCache = JSON.parse(cached);
|
||||
const now = Date.now();
|
||||
|
||||
// Check if cache is expired
|
||||
if (now - parsed.timestamp > CACHE_DURATION) {
|
||||
clearLlmCache();
|
||||
return null;
|
||||
}
|
||||
|
||||
return parsed.data;
|
||||
} catch (error) {
|
||||
console.error('Error getting cached LLM list:', error);
|
||||
clearLlmCache();
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
// Set LLM list to cache
|
||||
export function setCachedLlmList(data: Record<string, any>): void {
|
||||
try {
|
||||
const cache: LlmCache = {
|
||||
data,
|
||||
timestamp: Date.now(),
|
||||
};
|
||||
localStorage.setItem(CACHE_KEY, JSON.stringify(cache));
|
||||
} catch (error) {
|
||||
console.error('Error setting cached LLM list:', error);
|
||||
}
|
||||
}
|
||||
|
||||
// Clear LLM list cache
|
||||
export function clearLlmCache(): void {
|
||||
try {
|
||||
localStorage.removeItem(CACHE_KEY);
|
||||
} catch (error) {
|
||||
console.error('Error clearing LLM cache:', error);
|
||||
}
|
||||
}
|
||||
|
||||
// Check if LLM list is cached and not expired
|
||||
export function isLlmListCached(): boolean {
|
||||
return getCachedLlmList() !== null;
|
||||
}
|
||||
@ -1,4 +1,5 @@
|
||||
import { IThirdOAIModel } from '@/interfaces/database/llm';
|
||||
import { getCachedLlmList } from './llm-cache';
|
||||
|
||||
export const getLLMIconName = (fid: string, llm_name: string) => {
|
||||
if (fid === 'FastEmbed') {
|
||||
@ -22,3 +23,117 @@ export function getRealModelName(llmName: string) {
|
||||
export function buildLlmUuid(llm: IThirdOAIModel) {
|
||||
return `${llm.llm_name}@${llm.fid}`;
|
||||
}
|
||||
|
||||
// Get tenant model ID from LLM list by model name and factory ID
|
||||
export function getTenantModelId(
|
||||
llmList: Record<string, any>,
|
||||
modelName: string,
|
||||
factoryId: string,
|
||||
): string {
|
||||
// Iterate through all providers in the LLM list
|
||||
for (const [provider, data] of Object.entries(llmList)) {
|
||||
if (data.llm && Array.isArray(data.llm)) {
|
||||
// Handle /v1/llm/my_llms format
|
||||
const model = data.llm.find(
|
||||
(m: any) => m.name === modelName && provider === factoryId,
|
||||
);
|
||||
if (model && model.id) {
|
||||
return model.id;
|
||||
}
|
||||
} else if (Array.isArray(data)) {
|
||||
// Handle /v1/llm/list format
|
||||
const model = data.find(
|
||||
(m: any) => m.llm_name === modelName && m.fid === factoryId,
|
||||
);
|
||||
if (model && model.id) {
|
||||
return model.id;
|
||||
}
|
||||
}
|
||||
}
|
||||
return '';
|
||||
}
|
||||
|
||||
// Extract model name and factory ID from a model UUID (e.g., "model_name@factory_id")
|
||||
export function parseModelUuid(uuid: string): {
|
||||
modelName: string;
|
||||
factoryId: string;
|
||||
} {
|
||||
const [modelName, factoryId] = uuid.split('@');
|
||||
return { modelName, factoryId };
|
||||
}
|
||||
|
||||
// Model parameter to tenant parameter mapping
|
||||
type ModelParamMap = {
|
||||
[key: string]: string;
|
||||
};
|
||||
|
||||
const modelParamMap: ModelParamMap = {
|
||||
llm_id: 'tenant_llm_id',
|
||||
embd_id: 'tenant_embd_id',
|
||||
asr_id: 'tenant_asr_id',
|
||||
tts_id: 'tenant_tts_id',
|
||||
img2txt_id: 'tenant_img2txt_id',
|
||||
rerank_id: 'tenant_rerank_id',
|
||||
};
|
||||
|
||||
// API endpoint whitelist - only these endpoints will have tenant parameters added
|
||||
const API_WHITELIST = [
|
||||
'/v1/user/set_tenant_info',
|
||||
'/v1/dialog/set',
|
||||
'/v1/canvas/set',
|
||||
'/v1/canvas/setting',
|
||||
'/v1/search/update',
|
||||
'/api/v1/memories',
|
||||
'/v1/kb/create',
|
||||
'/v1/kb/update',
|
||||
'/v1/dataflow/set',
|
||||
];
|
||||
|
||||
// Check if the URL is in the whitelist
|
||||
export function isUrlInWhitelist(url: string): boolean {
|
||||
return API_WHITELIST.some((endpoint) => url.includes(endpoint));
|
||||
}
|
||||
|
||||
// Add tenant model ID parameters to request data
|
||||
export function addTenantParams(data: any, url?: string): any {
|
||||
if (!data || typeof data !== 'object') return data;
|
||||
|
||||
// If URL is provided and not in whitelist, return original data
|
||||
if (url && !isUrlInWhitelist(url)) {
|
||||
return data;
|
||||
}
|
||||
|
||||
const llmList = getCachedLlmList();
|
||||
if (!llmList) return data;
|
||||
|
||||
// Handle arrays
|
||||
if (Array.isArray(data)) {
|
||||
return data.map((item) => addTenantParams(item, url));
|
||||
}
|
||||
|
||||
const newData = { ...data };
|
||||
|
||||
// Iterate through model parameters and add corresponding tenant parameters
|
||||
for (const [paramName, tenantParamName] of Object.entries(modelParamMap)) {
|
||||
if (newData[paramName]) {
|
||||
try {
|
||||
const { modelName, factoryId } = parseModelUuid(newData[paramName]);
|
||||
const tenantModelId = getTenantModelId(llmList, modelName, factoryId);
|
||||
if (tenantModelId) {
|
||||
newData[tenantParamName] = tenantModelId;
|
||||
}
|
||||
} catch (error) {
|
||||
console.error(`Error processing ${paramName}:`, error);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Recursively process nested objects
|
||||
for (const [key, value] of Object.entries(newData)) {
|
||||
if (value && typeof value === 'object' && !modelParamMap[key]) {
|
||||
newData[key] = addTenantParams(value, url);
|
||||
}
|
||||
}
|
||||
|
||||
return newData;
|
||||
}
|
||||
|
||||
@ -8,6 +8,8 @@ import authorizationUtil, {
|
||||
import notification from '@/utils/notification';
|
||||
import axios from 'axios';
|
||||
import { convertTheKeysOfTheObjectToSnake } from './common-util';
|
||||
import { setCachedLlmList } from './llm-cache';
|
||||
import { addTenantParams } from './llm-util';
|
||||
|
||||
const FAILED_TO_FETCH = 'Failed to fetch';
|
||||
|
||||
@ -82,9 +84,13 @@ request.interceptors.request.use(
|
||||
const data = convertTheKeysOfTheObjectToSnake(config.data);
|
||||
const params = convertTheKeysOfTheObjectToSnake(config.params);
|
||||
|
||||
const newConfig = { ...config, data, params };
|
||||
// Add tenant parameters to data
|
||||
const dataWithTenantParams = addTenantParams(data, config.url);
|
||||
|
||||
if (!newConfig.skipToken) {
|
||||
const newConfig = { ...config, data: dataWithTenantParams, params };
|
||||
|
||||
// Skip token if explicitly requested
|
||||
if (!(newConfig as any).skipToken) {
|
||||
newConfig.headers.set(Authorization, getAuthorization());
|
||||
}
|
||||
|
||||
@ -105,6 +111,15 @@ request.interceptors.response.use(
|
||||
return response;
|
||||
}
|
||||
const data = response?.data;
|
||||
|
||||
// Update LLM list cache when fetching my_llm or llm_list
|
||||
if (data?.code === 0 && data?.data) {
|
||||
const url = response?.config?.url || '';
|
||||
if (url.includes('/v1/llm/my_llms') || url.includes('/v1/llm/list')) {
|
||||
setCachedLlmList(data.data);
|
||||
}
|
||||
}
|
||||
|
||||
if (data?.code === 100) {
|
||||
message.error(data?.message);
|
||||
} else if (data?.code === 401) {
|
||||
|
||||
@ -9,6 +9,8 @@ import authorizationUtil, {
|
||||
import notification from '@/utils/notification';
|
||||
import { RequestMethod, extend } from 'umi-request';
|
||||
import { convertTheKeysOfTheObjectToSnake } from './common-util';
|
||||
import { setCachedLlmList } from './llm-cache';
|
||||
import { addTenantParams } from './llm-util';
|
||||
|
||||
const FAILED_TO_FETCH = 'Failed to fetch';
|
||||
|
||||
@ -82,11 +84,14 @@ request.interceptors.request.use((url: string, options: any) => {
|
||||
const data = convertTheKeysOfTheObjectToSnake(options.data);
|
||||
const params = convertTheKeysOfTheObjectToSnake(options.params);
|
||||
|
||||
// Add tenant parameters to data
|
||||
const dataWithTenantParams = addTenantParams(data, url);
|
||||
|
||||
return {
|
||||
url,
|
||||
options: {
|
||||
...options,
|
||||
data,
|
||||
data: dataWithTenantParams,
|
||||
params,
|
||||
headers: {
|
||||
...(options.skipToken
|
||||
@ -109,6 +114,15 @@ request.interceptors.response.use(async (response: Response, options) => {
|
||||
}
|
||||
|
||||
const data: ResponseType = await response?.clone()?.json();
|
||||
|
||||
// Update LLM list cache when fetching my_llm or llm_list
|
||||
if (data?.code === 0 && data?.data) {
|
||||
const url = response?.url || '';
|
||||
if (url.includes('/v1/llm/my_llms') || url.includes('/v1/llm/list')) {
|
||||
setCachedLlmList(data.data);
|
||||
}
|
||||
}
|
||||
|
||||
if (data?.code === 100) {
|
||||
message.error(data?.message);
|
||||
} else if (data?.code === 401) {
|
||||
|
||||
Reference in New Issue
Block a user