@minded-ai/mindedjs 1.0.101 → 1.0.103-beta-1

package/src/platform/models/mindedChatOpenAI.ts

@@ -1,49 +0,0 @@
-import { ChatOpenAI, ChatOpenAIFields } from '@langchain/openai';
-import { getConfig } from '../config';
-
-export interface BaseParallelChatFields {
-  /**
-   * Number of parallel requests to make. Defaults to 1 (no parallelization).
-   * When > 1, multiple identical requests are sent and the fastest response is used.
-   */
-  numParallelRequests?: number;
-
-  /**
-   * Whether to log timing information for parallel requests
-   */
-  logTimings?: boolean;
-}
-
-export interface MindedChatOpenAIFields extends ChatOpenAIFields, BaseParallelChatFields {}
-
-export class MindedChatOpenAI extends ChatOpenAI {
-  private numParallelRequests: number;
-  private logTimings: boolean;
-
-  constructor(fields?: MindedChatOpenAIFields) {
-    const { token, baseUrl } = getConfig();
-    const mindedBaseUrl = `${baseUrl}/sdk/llmGateway/chatOpenAI`;
-    if (!token) {
-      throw new Error('Minded token not found');
-    }
-
-    // Store parallel configuration
-    const numParallelRequests = fields?.numParallelRequests || 1;
-    const logTimings = fields?.logTimings || false;
-
-    super({
-      ...fields,
-      apiKey: token,
-      configuration: {
-        baseURL: mindedBaseUrl,
-        defaultHeaders: {
-          'X-Parallel-Requests': numParallelRequests.toString(),
-          'X-Log-Timings': logTimings.toString(),
-        },
-      },
-    });
-
-    this.numParallelRequests = numParallelRequests;
-    this.logTimings = logTimings;
-  }
-}

package/src/platform/models/parallelWrapper.ts

@@ -1,141 +0,0 @@
-import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { BaseMessage } from '@langchain/core/messages';
-import { ChatGenerationChunk, ChatResult } from '@langchain/core/outputs';
-import { CallbackManagerForLLMRun } from '@langchain/core/callbacks/manager';
-import { logger } from '../../utils/logger';
-
-export interface BaseParallelChatFields {
-  /**
-   * Number of parallel requests to make. Defaults to 1 (no parallelization).
-   * When > 1, multiple identical requests are sent and the fastest response is used.
-   */
-  numParallelRequests?: number;
-
-  /**
-   * Whether to log timing information for parallel requests
-   */
-  logTimings?: boolean;
-}
-
-/**
- * Creates a wrapped version of a chat model with parallel request capabilities
- */
-export function createParallelWrapper<T extends BaseChatModel>(model: T, fields?: BaseParallelChatFields): T {
-  const numParallelRequests = fields?.numParallelRequests || 1;
-  const logTimings = fields?.logTimings || false;
-
-  // If parallel requests are disabled, return the original model
-  if (numParallelRequests <= 1) {
-    return model;
-  }
-
-  // Helper function that returns a promise that resolves with the second fulfilled promise
-  const promiseSecondFulfillment = <T>(promises: Promise<T>[]): Promise<T> => {
-    return new Promise((resolve) => {
-      let fulfillCount = 0;
-
-      for (const p of promises) {
-        Promise.resolve(p)
-          .then((value) => {
-            fulfillCount++;
-            if (fulfillCount === 2) {
-              resolve(value);
-            }
-          })
-          .catch(() => {
-            // Ignoring rejections for second fulfillment tracking
-          });
-      }
-    });
-  };
-
-  // Store original methods
-  const originalGenerate = model._generate.bind(model);
-  const originalStream = model._streamResponseChunks.bind(model);
-
-  // Override _generate
-  model._generate = async function (messages: BaseMessage[], options?: any, runManager?: CallbackManagerForLLMRun): Promise<ChatResult> {
-    const startTime = Date.now();
-
-    // Create array of identical requests
-    const requests = Array.from({ length: numParallelRequests }, () => originalGenerate(messages, options, runManager));
-
-    let fastestRequestTime = 0;
-    let secondFastestRequestTime = 0;
-
-    // Race all requests and return the fastest
-    const racePromise = Promise.race(requests).then((result) => {
-      fastestRequestTime = (Date.now() - startTime) / 1000.0;
-
-      if (logTimings) {
-        logger.debug({
-          msg: '[Model] Fastest request completed',
-          requestTime: fastestRequestTime,
-          numParallelRequests: numParallelRequests,
-        });
-      }
-
-      return result;
-    });
-
-    // Track second fastest if we have 3+ requests
-    if (requests.length >= 3) {
-      promiseSecondFulfillment(requests)
-        .then(() => {
-          secondFastestRequestTime = (Date.now() - startTime) / 1000.0;
-        })
-        .catch(() => {
-          // Ignore errors in timing tracking
-        });
-    }
-
-    // Track all completion times if we have 2+ requests
-    if (requests.length >= 2 && logTimings) {
-      Promise.all(requests)
-        .then(() => {
-          const allFinishTime = (Date.now() - startTime) / 1000.0;
-          const timeSaved = allFinishTime - fastestRequestTime;
-          const timeSavedFromSecond = secondFastestRequestTime > 0 ? secondFastestRequestTime - fastestRequestTime : 0;
-
-          logger.debug({
-            msg: '[Model] Time saved using parallel requests',
-            fastestRequestTime,
-            secondFastestRequestTime: secondFastestRequestTime || 'N/A',
-            allFinishTime,
-            timeSaved,
-            timeSavedFromSecond,
-            numParallelRequests: numParallelRequests,
-          });
-        })
-        .catch(() => {
-          // Ignore errors in timing tracking
-        });
-    }
-
-    return racePromise;
-  };
-
-  // Override _streamResponseChunks
-  model._streamResponseChunks = async function* (
-    messages: BaseMessage[],
-    options?: any,
-    runManager?: CallbackManagerForLLMRun,
-  ): AsyncGenerator<ChatGenerationChunk> {
-    // For streaming, we'll use the original implementation
-    // Parallel streaming is more complex and may not provide the same benefits
-    if (numParallelRequests > 1 && logTimings) {
-      logger.debug({
-        msg: '[Model] Streaming mode - using single request',
-        reason: 'Parallel streaming not implemented',
-      });
-    }
-
-    yield* originalStream(messages, options, runManager);
-  };
-
-  // Add properties for testing
-  (model as any).numParallelRequests = numParallelRequests;
-  (model as any).logTimings = logTimings;
-
-  return model;
-}