LibreChat/packages/data-provider/src/parsers.ts
Marco Beretta 650e9b4f6c
📜 refactor: Optimize Conversation History Nav with Cursor Pagination (#5785)
*  feat: improve Nav/Conversations/Convo/NewChat component performance

*  feat: implement cursor-based pagination for conversations API

* 🔧 refactor: remove createdAt from conversation selection in API and type definitions

* 🔧 refactor: include createdAt in conversation selection and update related types

*  fix: search functionality and bugs with loadMoreConversations

* feat: move ArchivedChats to cursor and DataTable standard

* 🔧 refactor: add InfiniteQueryObserverResult type import in Nav component

* feat: enhance conversation listing with pagination, sorting, and search capabilities

* 🔧 refactor: remove unnecessary comment regarding lodash/debounce in ArchivedChatsTable

* 🔧 refactor: remove unused translation keys for archived chats and search results

* 🔧 fix: Archived Chats, Delete Convo, Duplicate Convo

* 🔧 refactor: improve conversation components with layout adjustments and new translations

* 🔧 refactor: simplify archive conversation mutation and improve unarchive handling; fix: update fork mutation

* 🔧 refactor: decode search query parameter in conversation route; improve error handling in unarchive mutation; clean up DataTable component styles

* 🔧 refactor: remove unused translation key for empty archived chats

* 🚀 fix: `archivedConversation` query key not updated correctly while archiving

* 🧠 feat: Bedrock Anthropic Reasoning & Update Endpoint Handling (#6163)

* feat: Add thinking and thinkingBudget parameters for Bedrock Anthropic models

* chore: Update @librechat/agents to version 2.1.8

* refactor: change region order in params

* refactor: Add maxTokens parameter to conversation preset schema

* refactor: Update agent client to use bedrockInputSchema and improve error handling for model parameters

* refactor: streamline/optimize llmConfig initialization and saving for bedrock

* fix: ensure config titleModel is used for all endpoints

* refactor: enhance OpenAIClient and agent initialization to support endpoint checks for OpenRouter

* chore: bump @google/generative-ai

*  feat: improve Nav/Conversations/Convo/NewChat component performance

* 🔧 refactor: remove unnecessary comment regarding lodash/debounce in ArchivedChatsTable

* 🔧 refactor: update translation keys for clarity; simplify conversation query parameters and improve sorting functionality in SharedLinks component

* 🔧 refactor: optimize conversation loading logic and improve search handling in Nav component

* fix: package-lock

* fix: package-lock 2

* fix: package lock 3

* refactor: remove unused utility files and exports to clean up the codebase

* refactor: remove i18n and useAuthRedirect modules to streamline codebase

* refactor: optimize Conversations component and remove unused ToggleContext

* refactor(Convo): add RenameForm and ConvoLink components; enhance Conversations component with responsive design

* fix: add missing @azure/storage-blob dependency in package.json

* refactor(Search): add error handling with toast notification for search errors

* refactor: make createdAt and updatedAt fields of tConvoUpdateSchema less restrictive if timestamps are missing

* chore: update @azure/storage-blob dependency to version 12.27.0, ensure package-lock is correct

* refactor(Search): improve conversation handling server side

* fix: eslint warning and errors

* refactor(Search): improved search loading state and overall UX

* Refactors conversation cache management

Centralizes conversation mutation logic into dedicated utility functions for adding, updating, and removing conversations from query caches.

Improves reliability and maintainability by:
- Consolidating duplicate cache manipulation code
- Adding type safety for infinite query data structures
- Implementing consistent cache update patterns across all conversation operations
- Removing obsolete conversation helper functions in favor of standardized utilities

* fix: conversation handling and SSE event processing

- Optimizes conversation state management with useMemo and proper hook ordering
- Improves SSE event handler documentation and error handling
- Adds reset guard flag for conversation changes
- Removes redundant navigation call
- Cleans up cursor handling logic and document structure

Improves code maintainability and prevents potential race conditions in conversation state updates

* refactor: add type for SearchBar `onChange`

* fix: type tags

* style: rounded to xl all Header buttons

* fix: activeConvo in Convo not working

* style(Bookmarks): improved UI

* a11y(AccountSettings): fixed hover style not visible when using light theme

* style(SettingsTabs): improved tab switchers and dropdowns

* feat: add translations keys for Speech

* chore: fix package-lock

* fix(mutations): legacy import after rebase

* feat: refactor conversation navigation for accessibility

* fix(search): convo and message create/update date not returned

* fix(search): show correct iconURL and endpoint for searched messages

* fix: small UI improvements

* chore: console.log cleanup

* chore: fix tests

* fix(ChatForm): improve conversation ID handling and clean up useMemo dependencies

* chore: improve typing

* chore: improve typing

* fix(useSSE): clear conversation ID on submission to prevent draft restoration

* refactor(OpenAIClient): clean up abort handler

* refactor(abortMiddleware): change handleAbort to use function expression

* feat: add PENDING_CONVO constant and update conversation ID checks

* fix: final event handling on abort

* fix: improve title sync and query cache sync on final event

* fix: prevent overwriting cached conversation data if it already exists

---------

Co-authored-by: Danny Avila <danny@librechat.ai>
2025-04-15 04:04:00 -04:00

420 lines
12 KiB
TypeScript
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

import type { ZodIssue } from 'zod';
import type * as a from './types/assistants';
import type * as s from './schemas';
import type * as t from './types';
import { ContentTypes } from './types/runs';
import {
openAISchema,
googleSchema,
EModelEndpoint,
anthropicSchema,
assistantSchema,
gptPluginsSchema,
// agentsSchema,
compactAgentsSchema,
compactGoogleSchema,
compactPluginsSchema,
compactAssistantSchema,
} from './schemas';
import { bedrockInputSchema } from './bedrock';
import { extractEnvVariable } from './utils';
import { alternateName } from './config';
type EndpointSchema =
| typeof openAISchema
| typeof googleSchema
| typeof anthropicSchema
| typeof gptPluginsSchema
| typeof assistantSchema
| typeof compactAgentsSchema
| typeof bedrockInputSchema;
export type EndpointSchemaKey = Exclude<EModelEndpoint, EModelEndpoint.chatGPTBrowser>;
const endpointSchemas: Record<EndpointSchemaKey, EndpointSchema> = {
[EModelEndpoint.openAI]: openAISchema,
[EModelEndpoint.azureOpenAI]: openAISchema,
[EModelEndpoint.custom]: openAISchema,
[EModelEndpoint.google]: googleSchema,
[EModelEndpoint.anthropic]: anthropicSchema,
[EModelEndpoint.gptPlugins]: gptPluginsSchema,
[EModelEndpoint.assistants]: assistantSchema,
[EModelEndpoint.azureAssistants]: assistantSchema,
[EModelEndpoint.agents]: compactAgentsSchema,
[EModelEndpoint.bedrock]: bedrockInputSchema,
};
// const schemaCreators: Record<EModelEndpoint, (customSchema: DefaultSchemaValues) => EndpointSchema> = {
// [EModelEndpoint.google]: createGoogleSchema,
// };
/** Get the enabled endpoints from the `ENDPOINTS` environment variable */
export function getEnabledEndpoints() {
const defaultEndpoints: string[] = [
EModelEndpoint.openAI,
EModelEndpoint.agents,
EModelEndpoint.assistants,
EModelEndpoint.azureAssistants,
EModelEndpoint.azureOpenAI,
EModelEndpoint.google,
EModelEndpoint.chatGPTBrowser,
EModelEndpoint.gptPlugins,
EModelEndpoint.anthropic,
EModelEndpoint.bedrock,
];
const endpointsEnv = process.env.ENDPOINTS ?? '';
let enabledEndpoints = defaultEndpoints;
if (endpointsEnv) {
enabledEndpoints = endpointsEnv
.split(',')
.filter((endpoint) => endpoint.trim())
.map((endpoint) => endpoint.trim());
}
return enabledEndpoints;
}
/** Orders an existing EndpointsConfig object based on enabled endpoint/custom ordering */
export function orderEndpointsConfig(endpointsConfig: t.TEndpointsConfig) {
if (!endpointsConfig) {
return {};
}
const enabledEndpoints = getEnabledEndpoints();
const endpointKeys = Object.keys(endpointsConfig);
const defaultCustomIndex = enabledEndpoints.indexOf(EModelEndpoint.custom);
return endpointKeys.reduce(
(accumulatedConfig: Record<string, t.TConfig | null | undefined>, currentEndpointKey) => {
const isCustom = !(currentEndpointKey in EModelEndpoint);
const isEnabled = enabledEndpoints.includes(currentEndpointKey);
if (!isEnabled && !isCustom) {
return accumulatedConfig;
}
const index = enabledEndpoints.indexOf(currentEndpointKey);
if (isCustom) {
accumulatedConfig[currentEndpointKey] = {
order: defaultCustomIndex >= 0 ? defaultCustomIndex : 9999,
...(endpointsConfig[currentEndpointKey] as Omit<t.TConfig, 'order'> & { order?: number }),
};
} else if (endpointsConfig[currentEndpointKey]) {
accumulatedConfig[currentEndpointKey] = {
...endpointsConfig[currentEndpointKey],
order: index,
};
}
return accumulatedConfig;
},
{},
);
}
/** Converts an array of Zod issues into a string. */
export function errorsToString(errors: ZodIssue[]) {
return errors
.map((error) => {
const field = error.path.join('.');
const message = error.message;
return `${field}: ${message}`;
})
.join(' ');
}
/** Resolves header values to env variables if detected */
export function resolveHeaders(headers: Record<string, string> | undefined) {
const resolvedHeaders = { ...(headers ?? {}) };
if (headers && typeof headers === 'object' && !Array.isArray(headers)) {
Object.keys(headers).forEach((key) => {
resolvedHeaders[key] = extractEnvVariable(headers[key]);
});
}
return resolvedHeaders;
}
export function getFirstDefinedValue(possibleValues: string[]) {
let returnValue;
for (const value of possibleValues) {
if (value) {
returnValue = value;
break;
}
}
return returnValue;
}
export function getNonEmptyValue(possibleValues: string[]) {
for (const value of possibleValues) {
if (value && value.trim() !== '') {
return value;
}
}
return undefined;
}
export type TPossibleValues = {
models: string[];
secondaryModels?: string[];
};
export const parseConvo = ({
endpoint,
endpointType,
conversation,
possibleValues,
}: {
endpoint: EndpointSchemaKey;
endpointType?: EndpointSchemaKey | null;
conversation: Partial<s.TConversation | s.TPreset> | null;
possibleValues?: TPossibleValues;
// TODO: POC for default schema
// defaultSchema?: Partial<EndpointSchema>,
}) => {
let schema = endpointSchemas[endpoint] as EndpointSchema | undefined;
if (!schema && !endpointType) {
throw new Error(`Unknown endpoint: ${endpoint}`);
} else if (!schema && endpointType) {
schema = endpointSchemas[endpointType];
}
// if (defaultSchema && schemaCreators[endpoint]) {
// schema = schemaCreators[endpoint](defaultSchema);
// }
const convo = schema?.parse(conversation) as s.TConversation | undefined;
const { models, secondaryModels } = possibleValues ?? {};
if (models && convo) {
convo.model = getFirstDefinedValue(models) ?? convo.model;
}
if (secondaryModels && convo?.agentOptions) {
convo.agentOptions.model = getFirstDefinedValue(secondaryModels) ?? convo.agentOptions.model;
}
return convo;
};
/** Match GPT followed by digit, optional decimal, and optional suffix
*
* Examples: gpt-4, gpt-4o, gpt-4.5, gpt-5a, etc. */
const extractGPTVersion = (modelStr: string): string => {
const gptMatch = modelStr.match(/gpt-(\d+(?:\.\d+)?)([a-z])?/i);
if (gptMatch) {
const version = gptMatch[1];
const suffix = gptMatch[2] || '';
return `GPT-${version}${suffix}`;
}
return '';
};
/** Match omni models (o1, o3, etc.), "o" followed by a digit, possibly with decimal */
const extractOmniVersion = (modelStr: string): string => {
const omniMatch = modelStr.match(/\bo(\d+(?:\.\d+)?)\b/i);
if (omniMatch) {
const version = omniMatch[1];
return `o${version}`;
}
return '';
};
export const getResponseSender = (endpointOption: t.TEndpointOption): string => {
const {
model: _m,
endpoint,
endpointType,
modelDisplayLabel: _mdl,
chatGptLabel: _cgl,
modelLabel: _ml,
} = endpointOption;
const model = _m ?? '';
const modelDisplayLabel = _mdl ?? '';
const chatGptLabel = _cgl ?? '';
const modelLabel = _ml ?? '';
if (
[
EModelEndpoint.openAI,
EModelEndpoint.bedrock,
EModelEndpoint.gptPlugins,
EModelEndpoint.azureOpenAI,
EModelEndpoint.chatGPTBrowser,
].includes(endpoint)
) {
if (chatGptLabel) {
return chatGptLabel;
} else if (modelLabel) {
return modelLabel;
} else if (model && extractOmniVersion(model)) {
return extractOmniVersion(model);
} else if (model && (model.includes('mistral') || model.includes('codestral'))) {
return 'Mistral';
} else if (model && model.includes('gpt-')) {
const gptVersion = extractGPTVersion(model);
return gptVersion || 'GPT';
}
return (alternateName[endpoint] as string | undefined) ?? 'ChatGPT';
}
if (endpoint === EModelEndpoint.anthropic) {
return modelLabel || 'Claude';
}
if (endpoint === EModelEndpoint.bedrock) {
return modelLabel || alternateName[endpoint];
}
if (endpoint === EModelEndpoint.google) {
if (modelLabel) {
return modelLabel;
} else if (model && (model.includes('gemini') || model.includes('learnlm'))) {
return 'Gemini';
} else if (model && model.includes('code')) {
return 'Codey';
}
return 'PaLM2';
}
if (endpoint === EModelEndpoint.custom || endpointType === EModelEndpoint.custom) {
if (modelLabel) {
return modelLabel;
} else if (chatGptLabel) {
return chatGptLabel;
} else if (model && extractOmniVersion(model)) {
return extractOmniVersion(model);
} else if (model && (model.includes('mistral') || model.includes('codestral'))) {
return 'Mistral';
} else if (model && model.includes('gpt-')) {
const gptVersion = extractGPTVersion(model);
return gptVersion || 'GPT';
} else if (modelDisplayLabel) {
return modelDisplayLabel;
}
return 'AI';
}
return '';
};
type CompactEndpointSchema =
| typeof openAISchema
| typeof compactAssistantSchema
| typeof compactAgentsSchema
| typeof compactGoogleSchema
| typeof anthropicSchema
| typeof bedrockInputSchema
| typeof compactPluginsSchema;
const compactEndpointSchemas: Record<EndpointSchemaKey, CompactEndpointSchema> = {
[EModelEndpoint.openAI]: openAISchema,
[EModelEndpoint.azureOpenAI]: openAISchema,
[EModelEndpoint.custom]: openAISchema,
[EModelEndpoint.assistants]: compactAssistantSchema,
[EModelEndpoint.azureAssistants]: compactAssistantSchema,
[EModelEndpoint.agents]: compactAgentsSchema,
[EModelEndpoint.google]: compactGoogleSchema,
[EModelEndpoint.bedrock]: bedrockInputSchema,
[EModelEndpoint.anthropic]: anthropicSchema,
[EModelEndpoint.gptPlugins]: compactPluginsSchema,
};
export const parseCompactConvo = ({
endpoint,
endpointType,
conversation,
possibleValues,
}: {
endpoint?: EndpointSchemaKey;
endpointType?: EndpointSchemaKey | null;
conversation: Partial<s.TConversation | s.TPreset>;
possibleValues?: TPossibleValues;
// TODO: POC for default schema
// defaultSchema?: Partial<EndpointSchema>,
}) => {
if (!endpoint) {
throw new Error(`undefined endpoint: ${endpoint}`);
}
let schema = compactEndpointSchemas[endpoint] as CompactEndpointSchema | undefined;
if (!schema && !endpointType) {
throw new Error(`Unknown endpoint: ${endpoint}`);
} else if (!schema && endpointType) {
schema = compactEndpointSchemas[endpointType];
}
if (!schema) {
throw new Error(`Unknown endpointType: ${endpointType}`);
}
const convo = schema.parse(conversation) as s.TConversation | null;
// const { models, secondaryModels } = possibleValues ?? {};
const { models } = possibleValues ?? {};
if (models && convo) {
convo.model = getFirstDefinedValue(models) ?? convo.model;
}
// if (secondaryModels && convo.agentOptions) {
// convo.agentOptionmodel = getFirstDefinedValue(secondaryModels) ?? convo.agentOptionmodel;
// }
return convo;
};
export function parseTextParts(
contentParts: a.TMessageContentParts[],
skipReasoning: boolean = false,
): string {
let result = '';
for (const part of contentParts) {
if (!part.type) {
continue;
}
if (part.type === ContentTypes.TEXT) {
const textValue = typeof part.text === 'string' ? part.text : part.text.value;
if (
result.length > 0 &&
textValue.length > 0 &&
result[result.length - 1] !== ' ' &&
textValue[0] !== ' '
) {
result += ' ';
}
result += textValue;
} else if (part.type === ContentTypes.THINK && !skipReasoning) {
const textValue = typeof part.think === 'string' ? part.think : '';
if (
result.length > 0 &&
textValue.length > 0 &&
result[result.length - 1] !== ' ' &&
textValue[0] !== ' '
) {
result += ' ';
}
result += textValue;
}
}
return result;
}
export const SEPARATORS = ['.', '?', '!', '۔', '。', '‥', ';', '¡', '¿', '\n', '```'];
export function findLastSeparatorIndex(text: string, separators = SEPARATORS): number {
let lastIndex = -1;
for (const separator of separators) {
const index = text.lastIndexOf(separator);
if (index > lastIndex) {
lastIndex = index;
}
}
return lastIndex;
}