Merge branch '108' of https://github.com/HBAI-Ltd/Toonflow-app into 108

src/lib/initDB.ts

@@ -140,16 +140,36 @@ export default async (knex: Knex, forceInit: boolean = false): Promise<void> =>
             value: uuid().slice(0, 8),
           },
           {
-            key: "shortTermMemoryLength",
-            value: 10,
-          },
-          {
-            key: "searchTopK",
+            key: "messagesPerSummary",
             value: 3,
           },
           {
-            key: "similarityThreshold",
-            value: 0.3,
+            key: "shortTermLimit",
+            value: 5,
+          },
+          {
+            key: "summaryMaxLength",
+            value: 500,
+          },
+          {
+            key: "summaryLimit",
+            value: 10,
+          },
+          {
+            key: "ragLimit",
+            value: 3,
+          },
+          {
+            key: "deepRetrieveSummaryLimit",
+            value: 5,
+          },
+          {
+            key: "modelOnnxFile",
+            value: '["all-MiniLM-L6-v2", "onnx", "model_fp16.onnx"]',
+          },
+          {
+            key: "modelDtype",
+            value: "fp16",
           },
         ]);
       },

src/routes/setting/memoryConfig/getMemory.ts

@@ -4,14 +4,31 @@ import u from "@/utils";
 const router = express.Router();
 
 export default router.get("/", async (req, res) => {
-  const settingData = await u.db("o_setting").whereIn("key", ["shortTermMemoryLength", "searchTopK", "similarityThreshold"]);
+  const settingData = await u
+    .db("o_setting")
+    .whereIn("key", [
+      "messagesPerSummary",
+      "shortTermLimit",
+      "summaryMaxLength",
+      "summaryLimit",
+      "ragLimit",
+      "deepRetrieveSummaryLimit",
+      "modelOnnxFile",
+      "modelDtype",
+    ]);
 
   if (!settingData) return res.status(400).send(error(`获取记忆配置失败`));
-  const memoryObj: Record<string, number> = {};
+  const memoryObj: Record<string, number | string | string[]> = {};
 
   settingData.forEach((i) => {
     if (i.key && i.value) {
-      memoryObj[i.key] = Number(i.value);
+      let value: number | string | string[] = i.value;
+      if (i.key == "modelOnnxFile") {
+        value = JSON.parse(i.value);
+      } else if (i.key != "modelDtype") {
+        value = Number(value);
+      }
+      memoryObj[i.key] = value;
     }
   });
 

src/routes/setting/memoryConfig/sureMemory.ts

@@ -9,21 +9,37 @@ const router = express.Router();
 export default router.post(
   "/",
   validateFields({
-    shortTermMemoryLength: z.number(), //短期记忆长度
-    searchTopK: z.number(), //搜索记忆条数
-    similarityThreshold: z.number(), //记忆相似度阈值
+    messagesPerSummary: z.number(),
+    shortTermLimit: z.number(),
+    summaryMaxLength: z.number(),
+    summaryLimit: z.number(),
+    ragLimit: z.number(),
+    deepRetrieveSummaryLimit: z.number(),
+    modelOnnxFile: z.array(z.string()),
+    modelDtype: z.string(),
   }),
   async (req, res) => {
-    const { shortTermMemoryLength, searchTopK, similarityThreshold } = req.body;
-    await u.db("o_setting").where("key", "shortTermMemoryLength").update({
-      value: shortTermMemoryLength,
-    });
-    await u.db("o_setting").where("key", "searchTopK").update({
-      value: searchTopK,
-    });
-    await u.db("o_setting").where("key", "similarityThreshold").update({
-      value: similarityThreshold,
-    });
+    const { messagesPerSummary, shortTermLimit, summaryMaxLength, summaryLimit, ragLimit, deepRetrieveSummaryLimit, modelOnnxFile, modelDtype } =
+      req.body;
+
+    const upsert = async (key: string, value: string) => {
+      const exists = await u.db("o_setting").where("key", key).first();
+      if (exists) {
+        await u.db("o_setting").where("key", key).update({ value });
+      } else {
+        await u.db("o_setting").insert({ key, value });
+      }
+    };
+
+    await upsert("messagesPerSummary", messagesPerSummary);
+    await upsert("shortTermLimit", shortTermLimit);
+    await upsert("summaryMaxLength", summaryMaxLength);
+    await upsert("summaryLimit", summaryLimit);
+    await upsert("ragLimit", ragLimit);
+    await upsert("deepRetrieveSummaryLimit", deepRetrieveSummaryLimit);
+    await upsert("modelOnnxFile", JSON.stringify(modelOnnxFile));
+    await upsert("modelDtype", modelDtype);
+
     res.status(200).send(success("保存设置成功"));
   },
 );

src/utils/agent/embedding.ts

@@ -2,18 +2,24 @@ import { pipeline, env as transformersEnv, FeatureExtractionPipeline } from "@hu
 import path from "path";
 import fs from "fs";
 import getPath from "@/utils/getPath";
+import db from "@/utils/db";
 
 // ── 模型配置 ──
-const modelOnnxFile = ["all-MiniLM-L6-v2", "onnx", "model_fp16.onnx"]; // 模型文件路径
-const modelDtype = "fp16" as const; // 量化类型：fp32
-
+// const modelOnnxFile = ["all-MiniLM-L6-v2", "onnx", "model_fp16.onnx"]; // 模型文件路径
+// const modelDtype = "fp16" as const; // 量化类型：fp32
 let extractor: FeatureExtractionPipeline | null = null;
 
 export async function initEmbedding(): Promise<void> {
   if (extractor) return;
 
   //todo 模型配置放到这里
-
+  const modelConfigData = await db("o_setting").whereIn("key", ["modelOnnxFile", "modelDtype"]);
+  const modelObj: Record<string, string> = {};
+  Object.entries(modelConfigData).forEach(([key, value]) => {
+    modelObj[key] = value as string;
+  });
+  let modelOnnxFile = modelObj?.modelOnnxFile ? JSON.parse(modelObj.modelOnnxFile) : ["all-MiniLM-L6-v2", "onnx", "model_fp16.onnx"]; // 模型文件路径
+  let modelDtype = modelObj?.modelDtype ?? ("fp16" as const); // 量化类型：fp32
   const onnxPath = path.join(getPath("models"), ...modelOnnxFile);
   if (!fs.existsSync(onnxPath)) {
     throw new Error(`Embedding 模型文件不存在: ${onnxPath}`);