working es

application/api/answer/routes.py

@@ -13,8 +13,8 @@
 
 
 from application.core.settings import settings
+from application.vectorstore.vector_creator import VectorCreator
 from application.llm.llm_creator import LLMCreator
-from application.vectorstore.faiss import FaissStore
 from application.error import bad_request
 
 
@@ -226,7 +226,7 @@
         vectorstore = get_vectorstore({"active_docs": data["active_docs"]})
     else:
         vectorstore = ""
-    docsearch = FaissStore(vectorstore, embeddings_key)
+    docsearch = VectorCreator.create_vectorstore(settings.VECTOR_STORE, vectorstore, embeddings_key)
 
     return Response(
         complete_stream(question, docsearch,
@@ -260,7 +260,7 @@
         vectorstore = get_vectorstore(data)
         # loading the index and the store and the prompt template
         # Note if you have used other embeddings than OpenAI, you need to change the embeddings
-        docsearch = FaissStore(vectorstore, embeddings_key)
+        docsearch = VectorCreator.create_vectorstore(settings.VECTOR_STORE, vectorstore, embeddings_key)
 
 
         llm = LLMCreator.create_llm(settings.LLM_NAME, api_key=api_key)

application/api/internal/routes.py

@@ -34,25 +34,26 @@
     if "name" not in request.form:
         return {"status": "no name"}
     job_name = secure_filename(request.form["name"])
-    if "file_faiss" not in request.files:
-        print("No file part")
-        return {"status": "no file"}
-    file_faiss = request.files["file_faiss"]
-    if file_faiss.filename == "":
-        return {"status": "no file name"}
-    if "file_pkl" not in request.files:
-        print("No file part")
-        return {"status": "no file"}
-    file_pkl = request.files["file_pkl"]
-    if file_pkl.filename == "":
-        return {"status": "no file name"}
-
-    # saves index files
     save_dir = os.path.join(current_dir, "indexes", user, job_name)
-    if not os.path.exists(save_dir):
-        os.makedirs(save_dir)
-    file_faiss.save(os.path.join(save_dir, "index.faiss"))
-    file_pkl.save(os.path.join(save_dir, "index.pkl"))
+    if settings.VECTOR_STORE == "faiss":
+        if "file_faiss" not in request.files:
+            print("No file part")
+            return {"status": "no file"}
+        file_faiss = request.files["file_faiss"]
+        if file_faiss.filename == "":
+            return {"status": "no file name"}
+        if "file_pkl" not in request.files:
+            print("No file part")
+            return {"status": "no file"}
+        file_pkl = request.files["file_pkl"]
+        if file_pkl.filename == "":
+            return {"status": "no file name"}
+        # saves index files
+
+        if not os.path.exists(save_dir):
+            os.makedirs(save_dir)
+        file_faiss.save(os.path.join(save_dir, "index.faiss"))
+        file_pkl.save(os.path.join(save_dir, "index.pkl"))
     # create entry in vectors_collection
     vectors_collection.insert_one(
         {

application/api/user/routes.py

@@ -11,6 +11,8 @@
 from application.api.user.tasks import ingest
 
 from application.core.settings import settings
+from application.vectorstore.vector_creator import VectorCreator
+
 mongo = MongoClient(settings.MONGO_URI)
 db = mongo["docsgpt"]
 conversations_collection = db["conversations"]
@@ -90,10 +92,17 @@
         return {"status": "error"}
     path_clean = "/".join(dirs)
     vectors_collection.delete_one({"location": path})
-    try:
-        shutil.rmtree(path_clean)
-    except FileNotFoundError:
-        pass
+    if settings.VECTOR_STORE == "faiss":
+        try:
+            shutil.rmtree(os.path.join(current_dir, path_clean))
+        except FileNotFoundError:
+            pass
+    else:
+        vetorstore = VectorCreator.create_vectorstore(
+            settings.VECTOR_STORE, path=os.path.join(current_dir, path_clean)
+        )
+        vetorstore.delete_index()
+
     return {"status": "ok"}
 
 @user.route("/api/upload", methods=["POST"])
@@ -173,11 +182,11 @@
                 "location": "local",
             }
         )
-
-    data_remote = requests.get("https://d3dg1063dc54p9.cloudfront.net/combined.json").json()
-    for index in data_remote:
-        index["location"] = "remote"
-        data.append(index)
+    if settings.VECTOR_STORE == "faiss":
+        data_remote = requests.get("https://d3dg1063dc54p9.cloudfront.net/combined.json").json()
+        for index in data_remote:
+            index["location"] = "remote"
+            data.append(index)
 
     return jsonify(data)
 

application/core/settings.py

@@ -13,6 +13,7 @@ class Settings(BaseSettings):
     TOKENS_MAX_HISTORY: int = 150
     SELF_HOSTED_MODEL: bool = False
     UPLOAD_FOLDER: str = "inputs"
+    VECTOR_STORE: str = "elasticsearch"  # "faiss" or "elasticsearch"
 
     API_URL: str = "http://localhost:7091"  # backend url for celery worker
 
@@ -23,6 +24,13 @@ class Settings(BaseSettings):
     AZURE_DEPLOYMENT_NAME: str = None  # azure deployment name for answering
     AZURE_EMBEDDINGS_DEPLOYMENT_NAME: str = None  # azure deployment name for embeddings
 
+    # elasticsearch
+    ELASTIC_CLOUD_ID: str = None # cloud id for elasticsearch
+    ELASTIC_USERNAME: str = None # username for elasticsearch
+    ELASTIC_PASSWORD: str = None # password for elasticsearch
+    ELASTIC_URL: str = None # url for elasticsearch
+    ELASTIC_INDEX: str = "docsgpt" # index name for elasticsearch
+
 
 path = Path(__file__).parent.parent.absolute()
 settings = Settings(_env_file=path.joinpath(".env"), _env_file_encoding="utf-8")

application/parser/open_ai_func.py

@@ -1,8 +1,8 @@
 import os
 
 import tiktoken
-from langchain.embeddings import OpenAIEmbeddings
-from langchain.vectorstores import FAISS
+from application.vectorstore.vector_creator import VectorCreator
+from application.core.settings import settings
 from retry import retry
 
 
@@ -33,12 +33,23 @@
         os.makedirs(f"{folder_name}")
 
     from tqdm import tqdm
-    docs_test = [docs[0]]
-    docs.pop(0)
     c1 = 0
-
-    store = FAISS.from_documents(docs_test, OpenAIEmbeddings(openai_api_key=os.getenv("EMBEDDINGS_KEY")))
-
+    if settings.VECTOR_STORE == "faiss":
+        docs_init = [docs[0]]
+        docs.pop(0)
+
+        store = VectorCreator.create_vectorstore(
+            settings.VECTOR_STORE,
+            docs_init = docs_init,
+            path=f"{folder_name}",
+            embeddings_key=os.getenv("EMBEDDINGS_KEY")
+        )
+    else:
+        store = VectorCreator.create_vectorstore(
+            settings.VECTOR_STORE,
+            path=f"{folder_name}",
+            embeddings_key=os.getenv("EMBEDDINGS_KEY")
+        )
     # Uncomment for MPNet embeddings
     # model_name = "sentence-transformers/all-mpnet-base-v2"
     # hf = HuggingFaceEmbeddings(model_name=model_name)
@@ -57,7 +68,8 @@
             store.save_local(f"{folder_name}")
             break
         c1 += 1
-    store.save_local(f"{folder_name}")
+    if settings.VECTOR_STORE == "faiss":
+        store.save_local(f"{folder_name}")
 
 
 def get_user_permission(docs, folder_name):

application/requirements.txt

@@ -22,6 +22,7 @@ decorator==5.1.1
 dill==0.3.6
 dnspython==2.3.0
 ecdsa==0.18.0
+elasticsearch==8.9.0
 entrypoints==0.4
 faiss-cpu==1.7.3
 filelock==3.9.0

application/vectorstore/base.py

@@ -19,7 +19,7 @@ def search(self, *args, **kwargs):
     def is_azure_configured(self):
         return settings.OPENAI_API_BASE and settings.OPENAI_API_VERSION and settings.AZURE_DEPLOYMENT_NAME
 
-    def _get_docsearch(self, embeddings_name, embeddings_key=None):
+    def _get_embeddings(self, embeddings_name, embeddings_key=None):
         embeddings_factory = {
             "openai_text-embedding-ada-002": OpenAIEmbeddings,
             "huggingface_sentence-transformers/all-mpnet-base-v2": HuggingFaceHubEmbeddings,