Add bce-embedding and fastembed (#383)

### What problem does this PR solve?


Issue link:#326

### Type of change

- [x] New Feature (non-breaking change which adds functionality)

README.md

@@ -55,6 +55,8 @@
 
 ## 📌 Latest Features
 
+- 2024-04-16 Add an embedding model 'bce-embedding-base_v1' from [QAnything](https://github.com/netease-youdao/QAnything).
+- 2024-04-16 Add [FastEmbed](https://github.com/qdrant/fastembed) is designed for light and speeding embedding.
 - 2024-04-11 Support [Xinference](./docs/xinference.md) for local LLM deployment.
 - 2024-04-10 Add a new layout recognization model for analyzing Laws documentation.
 - 2024-04-08 Support [Ollama](./docs/ollama.md) for local LLM deployment.

README_ja.md

@@ -55,6 +55,8 @@
 
 ## 📌 最新の機能
 
+- 2024-04-16 [QAnything](https://github.com/netease-youdao/QAnything) から埋め込みモデル「bce-embedding-base_v1」を追加します。
+- 2024-04-16 [FastEmbed](https://github.com/qdrant/fastembed) は、軽量かつ高速な埋め込み用に設計されています。
 - 2024-04-11 ローカル LLM デプロイメント用に [Xinference](./docs/xinference.md) をサポートします。
 - 2024-04-10 メソッド「Laws」に新しいレイアウト認識モデルを追加します。
 - 2024-04-08 [Ollama](./docs/ollama.md) を使用した大規模モデルのローカライズされたデプロイメントをサポートします。

README_zh.md

@@ -55,6 +55,8 @@
 
 ## 📌 新增功能
 
+- 2024-04-16 添加嵌入模型 [QAnything的bce-embedding-base_v1](https://github.com/netease-youdao/QAnything) 。
+- 2024-04-16 添加 [FastEmbed](https://github.com/qdrant/fastembed) 专为轻型和高速嵌入而设计。
 - 2024-04-11 支持用 [Xinference](./docs/xinference.md) 本地化部署大模型。
 - 2024-04-10 为‘Laws’版面分析增加了底层模型。
 - 2024-04-08 支持用 [Ollama](./docs/ollama.md) 本地化部署大模型。

api/apps/chunk_app.py

@@ -252,7 +252,7 @@ def retrieval_test():
             return get_data_error_result(retmsg="Knowledgebase not found!")
 
         embd_mdl = TenantLLMService.model_instance(
-            kb.tenant_id, LLMType.EMBEDDING.value)
+            kb.tenant_id, LLMType.EMBEDDING.value, llm_name=kb.embd_id)
         ranks = retrievaler.retrieval(question, embd_mdl, kb.tenant_id, [kb_id], page, size, similarity_threshold,
                                       vector_similarity_weight, top, doc_ids)
         for c in ranks["chunks"]:

api/apps/document_app.py

@@ -15,6 +15,7 @@
 #
 
 import base64
+import os
 import pathlib
 import re
 

api/apps/llm_app.py

@@ -28,7 +28,7 @@ from rag.llm import EmbeddingModel, ChatModel
 def factories():
     try:
         fac = LLMFactoriesService.get_all()
-        return get_json_result(data=[f.to_dict() for f in fac])
+        return get_json_result(data=[f.to_dict() for f in fac if f.name not in ["QAnything", "FastEmbed"]])
     except Exception as e:
         return server_error_response(e)
 
@@ -174,7 +174,7 @@ def list():
         llms = [m.to_dict()
                 for m in llms if m.status == StatusEnum.VALID.value]
         for m in llms:
-            m["available"] = m["fid"] in facts or m["llm_name"].lower() == "flag-embedding"
+            m["available"] = m["fid"] in facts or m["llm_name"].lower() == "flag-embedding" or m["fid"] in ["QAnything","FastEmbed"]
 
         llm_set = set([m["llm_name"] for m in llms])
         for o in objs:

api/db/init_data.py

@@ -18,7 +18,7 @@ import time
 import uuid
 
 from api.db import LLMType, UserTenantRole
-from api.db.db_models import init_database_tables as init_web_db, LLMFactories, LLM
+from api.db.db_models import init_database_tables as init_web_db, LLMFactories, LLM, TenantLLM
 from api.db.services import UserService
 from api.db.services.llm_service import LLMFactoriesService, LLMService, TenantLLMService, LLMBundle
 from api.db.services.user_service import TenantService, UserTenantService
@@ -114,12 +114,16 @@ factory_infos = [{
     "logo": "",
     "tags": "TEXT EMBEDDING",
     "status": "1",
-},
+}, {
-    {
     "name": "Xinference",
     "logo": "",
     "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
         "status": "1",
+},{
+    "name": "QAnything",
+    "logo": "",
+    "tags": "LLM,TEXT EMBEDDING,SPEECH2TEXT,MODERATION",
+        "status": "1",
 },
     # {
     #     "name": "文心一言",
@@ -254,12 +258,6 @@ def init_llm_factory():
             "tags": "LLM,CHAT,",
             "max_tokens": 7900,
             "model_type": LLMType.CHAT.value
-        }, {
-            "fid": factory_infos[4]["name"],
-            "llm_name": "flag-embedding",
-            "tags": "TEXT EMBEDDING,",
-            "max_tokens": 128 * 1000,
-            "model_type": LLMType.EMBEDDING.value
         }, {
             "fid": factory_infos[4]["name"],
             "llm_name": "moonshot-v1-32k",
@@ -325,6 +323,14 @@ def init_llm_factory():
             "max_tokens": 2147483648,
             "model_type": LLMType.EMBEDDING.value
         },
+        # ------------------------ QAnything -----------------------
+        {
+            "fid": factory_infos[7]["name"],
+            "llm_name": "maidalun1020/bce-embedding-base_v1",
+            "tags": "TEXT EMBEDDING,",
+            "max_tokens": 512,
+            "model_type": LLMType.EMBEDDING.value
+        },
     ]
     for info in factory_infos:
         try:
@@ -337,8 +343,10 @@ def init_llm_factory():
         except Exception as e:
             pass
 
-    LLMFactoriesService.filter_delete([LLMFactories.name=="Local"])
+    LLMFactoriesService.filter_delete([LLMFactories.name == "Local"])
-    LLMService.filter_delete([LLM.fid=="Local"])
+    LLMService.filter_delete([LLM.fid == "Local"])
+    LLMService.filter_delete([LLM.fid == "Moonshot", LLM.llm_name == "flag-embedding"])
+    TenantLLMService.filter_delete([TenantLLM.llm_factory == "Moonshot", TenantLLM.llm_name == "flag-embedding"])
 
     """
     drop table llm;

api/db/services/dialog_service.py

@@ -80,8 +80,12 @@ def chat(dialog, messages, **kwargs):
             raise LookupError("LLM(%s) not found" % dialog.llm_id)
         max_tokens = 1024
     else: max_tokens = llm[0].max_tokens
+    kbs = KnowledgebaseService.get_by_ids(dialog.kb_ids)
+    embd_nms = list(set([kb.embd_id for kb in kbs]))
+    assert len(embd_nms) == 1, "Knowledge bases use different embedding models."
+
     questions = [m["content"] for m in messages if m["role"] == "user"]
-    embd_mdl = LLMBundle(dialog.tenant_id, LLMType.EMBEDDING)
+    embd_mdl = LLMBundle(dialog.tenant_id, LLMType.EMBEDDING, embd_nms[0])
     chat_mdl = LLMBundle(dialog.tenant_id, LLMType.CHAT, dialog.llm_id)
 
     prompt_config = dialog.prompt_config

api/db/services/llm_service.py

@@ -66,7 +66,7 @@ class TenantLLMService(CommonService):
             raise LookupError("Tenant not found")
 
         if llm_type == LLMType.EMBEDDING.value:
-            mdlnm = tenant.embd_id
+            mdlnm = tenant.embd_id if not llm_name else llm_name
         elif llm_type == LLMType.SPEECH2TEXT.value:
             mdlnm = tenant.asr_id
         elif llm_type == LLMType.IMAGE2TEXT.value:
@@ -77,9 +77,14 @@ class TenantLLMService(CommonService):
             assert False, "LLM type error"
 
         model_config = cls.get_api_key(tenant_id, mdlnm)
+        if model_config: model_config = model_config.to_dict()
         if not model_config:
-            raise LookupError("Model({}) not authorized".format(mdlnm))
+            if llm_type == LLMType.EMBEDDING.value:
-        model_config = model_config.to_dict()
+                llm = LLMService.query(llm_name=llm_name)
+                if llm and llm[0].fid in ["QAnything", "FastEmbed"]:
+                    model_config = {"llm_factory": llm[0].fid, "api_key":"", "llm_name": llm_name, "api_base": ""}
+            if not model_config: raise LookupError("Model({}) not authorized".format(mdlnm))
+
         if llm_type == LLMType.EMBEDDING.value:
             if model_config["llm_factory"] not in EmbeddingModel:
                 return

api/db/services/task_service.py

@@ -41,7 +41,7 @@ class TaskService(CommonService):
             Document.size,
             Knowledgebase.tenant_id,
             Knowledgebase.language,
-            Tenant.embd_id,
+            Knowledgebase.embd_id,
             Tenant.img2txt_id,
             Tenant.asr_id,
             cls.model.update_time]

rag/llm/__init__.py

@@ -24,8 +24,8 @@ EmbeddingModel = {
     "Xinference": XinferenceEmbed,
     "Tongyi-Qianwen": HuEmbedding, #QWenEmbed,
     "ZHIPU-AI": ZhipuEmbed,
-    "Moonshot": HuEmbedding,
+    "FastEmbed": FastEmbed,
-    "FastEmbed": FastEmbed
+    "QAnything": QAnythingEmbed
 }
 
 

rag/llm/embedding_model.py

@@ -20,7 +20,6 @@ from abc import ABC
 from ollama import Client
 import dashscope
 from openai import OpenAI
-from fastembed import TextEmbedding
 from FlagEmbedding import FlagModel
 import torch
 import numpy as np
@@ -28,6 +27,7 @@ import numpy as np
 from api.utils.file_utils import get_project_base_directory
 from rag.utils import num_tokens_from_string
 
+
 try:
     flag_model = FlagModel(os.path.join(
         get_project_base_directory(),
@@ -82,8 +82,10 @@ class HuEmbedding(Base):
 
 
 class OpenAIEmbed(Base):
-    def __init__(self, key, model_name="text-embedding-ada-002", base_url="https://api.openai.com/v1"):
+    def __init__(self, key, model_name="text-embedding-ada-002",
-        if not base_url: base_url="https://api.openai.com/v1"
+                 base_url="https://api.openai.com/v1"):
+        if not base_url:
+            base_url = "https://api.openai.com/v1"
         self.client = OpenAI(api_key=key, base_url=base_url)
         self.model_name = model_name
 
@@ -183,10 +185,12 @@ class FastEmbed(Base):
         threads: Optional[int] = None,
         **kwargs,
     ):
+        from fastembed import TextEmbedding
         self._model = TextEmbedding(model_name, cache_dir, threads, **kwargs)
 
     def encode(self, texts: list, batch_size=32):
-        # Using the internal tokenizer to encode the texts and get the total number of tokens
+        # Using the internal tokenizer to encode the texts and get the total
+        # number of tokens
         encodings = self._model.model.tokenizer.encode_batch(texts)
         total_tokens = sum(len(e) for e in encodings)
 
@@ -195,7 +199,8 @@ class FastEmbed(Base):
         return np.array(embeddings), total_tokens
 
     def encode_queries(self, text: str):
-        # Using the internal tokenizer to encode the texts and get the total number of tokens
+        # Using the internal tokenizer to encode the texts and get the total
+        # number of tokens
         encoding = self._model.model.tokenizer.encode(text)
         embedding = next(self._model.query_embed(text)).tolist()
 
@@ -218,3 +223,33 @@ class XinferenceEmbed(Base):
                                             model=self.model_name)
         return np.array(res.data[0].embedding), res.usage.total_tokens
 
+
+class QAnythingEmbed(Base):
+    _client = None
+
+    def __init__(self, key=None, model_name="maidalun1020/bce-embedding-base_v1", **kwargs):
+        from BCEmbedding import EmbeddingModel as qanthing
+        if not QAnythingEmbed._client:
+            try:
+                print("LOADING BCE...")
+                QAnythingEmbed._client = qanthing(model_name_or_path=os.path.join(
+                    get_project_base_directory(),
+                    "rag/res/bce-embedding-base_v1"))
+            except Exception as e:
+                QAnythingEmbed._client = qanthing(
+                    model_name_or_path=model_name.replace(
+                        "maidalun1020", "InfiniFlow"))
+
+    def encode(self, texts: list, batch_size=10):
+        res = []
+        token_count = 0
+        for t in texts:
+            token_count += num_tokens_from_string(t)
+        for i in range(0, len(texts), batch_size):
+            embds = QAnythingEmbed._client.encode(texts[i:i + batch_size])
+            res.extend(embds)
+        return np.array(res), token_count
+
+    def encode_queries(self, text):
+        embds = QAnythingEmbed._client.encode([text])
+        return np.array(embds[0]), num_tokens_from_string(text)

rag/nlp/search.py

@@ -46,7 +46,7 @@ class Dealer:
             "k": topk,
             "similarity": sim,
             "num_candidates": topk * 2,
-            "query_vector": list(qv)
+            "query_vector": [float(v) for v in qv]
         }
 
     def search(self, req, idxnm, emb_mdl=None):

rag/svr/task_executor.py

@@ -244,8 +244,9 @@ def main(comm, mod):
     for _, r in rows.iterrows():
         callback = partial(set_progress, r["id"], r["from_page"], r["to_page"])
         try:
-            embd_mdl = LLMBundle(r["tenant_id"], LLMType.EMBEDDING)
+            embd_mdl = LLMBundle(r["tenant_id"], LLMType.EMBEDDING, llm_name=r["embd_id"], lang=r["language"])
         except Exception as e:
+            traceback.print_stack(e)
             callback(prog=-1, msg=str(e))
             continue
 

requirements.txt

@@ -132,3 +132,5 @@ xpinyin==0.7.6
 xxhash==3.4.1
 yarl==1.9.4
 zhipuai==2.0.1
+BCEmbedding
+loguru==0.7.2