[webui] 更新部分组件

packages/core/src/bot.ts

@@ -52,12 +52,7 @@ export interface FailedResponse {
 type Response = SuccessResponse | SkipResponse | FailedResponse;
 
 export class Bot {
-  private memorySize: number;
-
-  private summarySize: number;         // 上下文达到多少时进行总结
   private contextSize: number = 20;    // 以对话形式给出的上下文长度
-  private retainedContextSize: number; // 进行总结时保留的上下文长度，用于保持记忆连贯性
-  private maxSelfMemoryCharacters: 5000; //
 
   private minTriggerCount: number;
   private maxTriggerCount: number;
@@ -69,10 +64,8 @@ export class Bot {
   private template: Template;
   private extensions: { [key: string]: Extension & Function } = {};
   private toolsSchema: ToolSchema[] = [];
-  private lastModified: Date = new Date();
 
   private emojiManager: EmojiManager;
-  private embedder: EmbeddingBase;
   readonly verifier: ResponseVerifier;
   readonly imageViewer: ImageViewer;
 
@@ -89,7 +82,6 @@ export class Bot {
     );
     if (config.Embedding.Enabled) {
       this.emojiManager = new EmojiManager(config.Embedding);
-      this.embedder = getEmbedding(config.Embedding);
     };
     if (config.Verifier.Enabled) this.verifier = new ResponseVerifier(config);
 
@@ -103,11 +95,6 @@ export class Bot {
     }
   }
 
-  updateConfig(config: Config) {
-    this.config = config;
-    this.adapterSwitcher.updateConfig(config.API.APIList, config.Parameters);
-  }
-
   setSystemPrompt(content: string) {
     this.prompt = content;
   }

packages/core/src/extensions/ext_memory.ts

@@ -5,9 +5,11 @@ import { Description, Extension, Name, Param } from "./base";
 @Name("addCoreMemory")
 @Description("Append to the contents of core memory.")
 @Param("content", SchemaNode.String("Content to write to the memory. All unicode (including emojis) are supported."))
+@Param("topic", SchemaNode.String("The topic of the memory.", ""))
+@Param("keywords", SchemaNode.Array("The keywords of the memory.", ""))
 export class AddCoreMemory extends Extension {
-  async apply(content: string) {
-    return await this.ctx.memory.addCoreMemory(content);
+  async apply(content: string, topic?: string, keywords?: string[]) {
+    return await this.ctx.memory.addCoreMemory(content, topic, keywords);
   }
 }
 

packages/memory/src/config.ts

@@ -28,8 +28,7 @@ export const EmbeddingConfig: Schema<EmbeddingConfig> = Schema.intersect([
       .default(300)
       .experimental()
       .description("文本分词长度"),
-
-  }),
+  }).description("Embedding API 配置"),
   Schema.union([
     Schema.object({
       APIType: Schema.const("OpenAI"),
@@ -58,12 +57,3 @@ export const EmbeddingConfig: Schema<EmbeddingConfig> = Schema.intersect([
     }),
   ])
 ])
-
-
-export interface Config {
-  embedding: EmbeddingConfig;
-}
-
-export const Config: Schema<Config> = Schema.object({
-  embedding: EmbeddingConfig,
-});

packages/memory/src/index.ts

@@ -1,11 +1,9 @@
 import { Context, Schema, Service } from "koishi";
-import { ChatMessage } from "koishi-plugin-yesimbot";
-import { EmbeddingBase, calculateCosineSimilarity } from "koishi-plugin-yesimbot/embeddings";
-import { getEmbedding } from "koishi-plugin-yesimbot/utils";
+import { ChatMessage, EmbeddingBase, calculateCosineSimilarity, getEmbedding } from "koishi-plugin-yesimbot";
 import { EmbeddingConfig } from "./config";
 import { MemoryItem, MemoryType } from "./model";
-import { MemoryMetadata, MemoryVectorStore } from "./vectorStore";
 import { MEMORY_PROMPT } from "./prompt";
+import { MemoryMetadata, MemoryVectorStore } from "./vectorStore";
 
 declare module "koishi" {
   interface Context {
@@ -17,8 +15,6 @@ export const inject = {
   required: ["yesimbot", "database"],
 };
 
-export { Config } from "./config";
-
 class Memory extends Service {
   private vectorStore: MemoryVectorStore;
   private embedder: EmbeddingBase;
@@ -51,13 +47,29 @@ class Memory extends Service {
     this.vectorStore.clear();
   }
 
-  async addCoreMemory(content: string) {
+  async modifyMemoryById(memoryId: string, content: string, type?: MemoryType, topic?: string, keywords?: string[]) {
+    const memory = this.vectorStore.get(memoryId);
+    if (memory) {
+      const embedding = (content == memory.content) ? memory.embedding : await this.embedder.embed(content);
+      const metadata: MemoryMetadata = {
+        content,
+        type: type || memory.type,
+        topic: topic || memory.topic,
+        keywords: keywords || memory.keywords,
+        createdAt: memory.createdAt,
+        updatedAt: new Date(),
+      };
+      this.vectorStore.update(memoryId, embedding, metadata);
+    }
+  }
+
+  async addCoreMemory(content: string, topic?: string, keywords?: string[]) {
     const embedding = await this.embedder.embed(content);
     const metadata: MemoryMetadata = {
       content,
-      topic: "核心记忆",
-      keywords: [],
       type: MemoryType.Core,
+      topic,
+      keywords,
       createdAt: new Date(),
       updatedAt: new Date(),
     };
@@ -124,14 +136,14 @@ class Memory extends Service {
     })
   }
 
-  async searchArchivalMemory(query: string, options: { type?: MemoryType, topic?: string, keywords?: string[]; limit?: number }): Promise<string[]> {
+  async searchArchivalMemory(query: string, type?: MemoryType, topic?: string, keywords?: string[], limit?: number): Promise<string[]> {
     const contextEmbedding = await this.embedder.embed(query);
 
     // 1. 主题与关键词过滤
     let filteredMemory = this.vectorStore.filter(item => {
-      const topicMatch = options.topic ? item.topic === options.topic : true;
-      const keywordMatch = options.keywords
-        ? options.keywords.some(keyword => item.keywords.includes(keyword))
+      const topicMatch = topic ? item.topic === topic : true;
+      const keywordMatch = keywords
+        ? keywords.some(keyword => item.keywords.includes(keyword))
         : true;
       return topicMatch && keywordMatch;
     });
@@ -145,29 +157,28 @@ class Memory extends Service {
     // 3. 排序并限制结果数
     const sortedMemory = scoredMemory
       .sort((a, b) => b.similarity - a.similarity) // 按相似度降序排序
-      .slice(0, options.limit || 5); // 限制返回结果数
+      .slice(0, limit || 5); // 限制返回结果数
 
     return sortedMemory.map(item => item.content);
   }
 
-
   /**
-   * Searches for conversation messages from a specific user that are semantically similar to a given query.
+   * 搜索与给定查询语义相似且来自特定用户的对话消息。
    *
-   * This method performs the following steps:
-   * 1. Embeds the query into a vector representation using the configured embedder.
-   * 2. Retrieves the most recent chat messages sent by the specified user from the database.
-   * 3. Computes the cosine similarity between the query embedding and the embedding of each chat message.
-   * 4. Sorts the messages by similarity in descending order and limits the results to the specified count.
-   * 5. Re-sorts the limited results by the original send time in ascending order to maintain chronological order.
-   * 6. Returns the content of the most relevant messages.
+   * 该方法执行以下步骤：
+   * 1. 使用配置的嵌入器将查询嵌入为向量表示。
+   * 2. 从数据库中检索指定用户发送的最新聊天消息。
+   * 3. 计算查询嵌入与每条聊天消息嵌入之间的余弦相似度。
+   * 4. 按相似度降序排序消息，并将结果限制为指定的数量。
+   * 5. 将限制后的结果按原始发送时间升序重新排序，以保持时间顺序。
+   * 6. 返回最相关消息的内容。
    *
-   * @param query - The search query string to find similar messages.
-   * @param userId - The ID of the user whose messages are to be searched.
-   * @param count - The maximum number of messages to return. Defaults to 10.
-   * @returns A promise that resolves to an array of message contents sorted by relevance and chronological order.
+   * @param query - 用于查找相似消息的搜索查询字符串。
+   * @param userId - 要搜索消息的用户ID。
+   * @param count - 要返回的最大消息数量。默认为10。
+   * @returns 一个Promise，解析为按相关性和时间顺序排序的消息内容数组。
    */
-  async searchConversation(query: string, userId: string, count: number = 10): Promise<string[]>{
+  async searchConversation(query: string, userId: string, count: number = 10): Promise<string[]> {
     let embedding = await this.embedder.embed(query);
 
     let chatMessages = await this.ctx.database
@@ -199,9 +210,17 @@ class Memory extends Service {
 
 namespace Memory {
   export interface Config {
+    memorySize: number;              // 记忆容量
+    summarySize: number;             // 上下文达到多少时进行总结
+    retainedContextSize: number;     // 进行总结时保留的上下文长度，用于保持记忆连贯性
+    maxCoreMemoryCharacters: number; // 最大记忆字符数
     embedding: EmbeddingConfig;
   }
   export const Config: Schema<Config> = Schema.object({
+    memorySize: Schema.number().default(1000),
+    summarySize: Schema.number().default(100),
+    retainedContextSize: Schema.number().default(10),
+    maxCoreMemoryCharacters: Schema.number().default(5000),
     embedding: EmbeddingConfig,
   });
 }

packages/memory/src/vectorStore.ts

@@ -4,7 +4,7 @@ import { Context } from "koishi";
 import { defineAccessor } from "@satorijs/core";
 
 import { CacheManager } from "koishi-plugin-yesimbot";
-import { calculateCosineSimilarity } from "koishi-plugin-yesimbot/embeddings";
+import { calculateCosineSimilarity } from "koishi-plugin-yesimbot";
 import { MemoryItem, MemoryType } from "./model";
 
 export interface MemoryMetadata {
@@ -17,13 +17,6 @@ export interface MemoryMetadata {
   updatedAt: Date;
 }
 
-
-export interface  MemoryVectorStore {
-  get(id: string): MemoryItem;
-  delete(id: string): boolean;
-  clear(): void;
-}
-
 export class MemoryVectorStore {
   readonly store: CacheManager<MemoryItem>;
 
@@ -32,6 +25,14 @@ export class MemoryVectorStore {
     this.store = new CacheManager(vectorsFilePath, true);
   }
 
+  delete(id: string): boolean {
+    return this.store.delete(id);
+  }
+
+  clear(): void {
+    this.store.clear();
+  }
+
   get(id: string): MemoryItem | undefined {
     return this.store.get(id);
   }
@@ -136,10 +137,6 @@ export class MemoryVectorStore {
   }
 }
 
-defineAccessor(MemoryVectorStore.prototype, "get", ["store", "get"]);
-defineAccessor(MemoryVectorStore.prototype, "clear", ["store", "clear"]);
-defineAccessor(MemoryVectorStore.prototype, "delete", ["store", "delete"]);
-
 /**
  * 获取向量的模
  * @param vector