fix(fastapi): use fastembed-vectorestore package

.dockerignore

@@ -7,4 +7,5 @@ build
 .git
 .gitignore
 tmp
-debug
+debug
+.fastembed_cache

.gitignore

@@ -8,4 +8,5 @@ node_modules
 out
 user_data
 tmp
-debug
+debug
+.fastembed_cache

servers/fastapi/image_processor/icons_finder.py

@@ -5,17 +5,17 @@ from ppt_generator.models.query_and_prompt_models import (
     IconCategoryEnum,
     IconQueryCollectionWithData,
 )
-from langchain_core.vectorstores import InMemoryVectorStore
+from fastembed_vectorstore import FastembedVectorstore
 
 
 async def get_icon(
-    vector_store: InMemoryVectorStore,
+    vector_store: FastembedVectorstore,
     input: IconQueryCollectionWithData,
 ) -> str:
     try:
         query = input.icon_query
-        results = vector_store.similarity_search(query=query, k=1)
-        icon_name = results[0].page_content
+        results = vector_store.search(query, 1)
+        icon_name = results[0][0].split("||")[0]
         return get_resource(f"assets/icons/bold/{icon_name}.png")
     except Exception as e:
         print("Error finding icon: ", e)
@@ -23,7 +23,7 @@ async def get_icon(
 
 
 async def get_icons(
-    vector_store: InMemoryVectorStore,
+    vector_store: FastembedVectorstore,
     query: str,
     page: int,
     limit: int,
@@ -31,7 +31,7 @@ async def get_icons(
     temp_dir: str,
 ) -> List[str]:
 
-    results = await vector_store.asimilarity_search(query=query, k=limit)
-    icon_names = [result.page_content for result in results]
+    results = vector_store.search(query, limit)
+    icon_names = [result[0].split("||")[0] for result in results]
 
     return [get_resource(f"assets/icons/bold/{each}.png") for each in icon_names]

servers/fastapi/image_processor/icons_vectorstore_utils.py

@@ -1,34 +1,26 @@
 import json
 import os
-from fastembed import TextEmbedding
 
 from api.utils.utils import get_resource
+from fastembed_vectorstore import FastembedVectorstore, FastembedEmbeddingModel
 
 
 def get_icons_vectorstore():
     vector_store_path = get_resource("assets/icons_vectorstore.json")
+    embedding_model = FastembedEmbeddingModel.BGESmallENV15
 
-    embedding_model = TextEmbedding()
-
-    # if os.path.exists(vector_store_path):
-    #     vector_store = InMemoryVectorStore.load(vector_store_path, embeddings)
-    #     return vector_store
+    if os.path.exists(vector_store_path):
+        return FastembedVectorstore.load(embedding_model, vector_store_path)
 
+    vector_store = FastembedVectorstore(embedding_model)
     with open(get_resource("assets/icons.json"), "r") as f:
         icons = json.load(f)
+    documents = []
+    for each in icons["icons"]:
+        if each["name"].split("-")[-1] == "bold":
+            documents.append(f"{each['name']}||{each['tags']}")
 
-    icon_names = [icon["name"] for icon in icons["icons"]]
-    bold_icon_names = []
-    for each in icon_names:
-        if each.split("-")[-1] == "bold":
-            bold_icon_names.append(each)
+    vector_store.embed_documents(documents)
+    vector_store.save(vector_store_path)
 
-    documents_and_embeddings = {
-        "documents": bold_icon_names,
-        "embeddings": embedding_model.embed(bold_icon_names),
-    }
-
-    with open(vector_store_path, "w") as f:
-        json.dump(documents_and_embeddings, f)
-
-    return documents_and_embeddings
+    return vector_store

servers/fastapi/image_processor/images_finder.py

@@ -3,7 +3,6 @@ import base64
 import os
 import uuid
 import aiohttp
-from langchain_google_genai import ChatGoogleGenerativeAI
 from openai import OpenAI
 
 from ppt_generator.models.query_and_prompt_models import (
@@ -67,22 +66,23 @@ async def generate_image_openai(prompt: str, output_directory: str) -> str:
 
 
 async def generate_image_google(prompt: str, output_directory: str) -> str:
-    response = await ChatGoogleGenerativeAI(
-        model="gemini-2.0-flash-preview-image-generation"
-    ).ainvoke([prompt], generation_config={"response_modalities": ["TEXT", "IMAGE"]})
+    # response = await ChatGoogleGenerativeAI(
+    #     model="gemini-2.0-flash-preview-image-generation"
+    # ).ainvoke([prompt], generation_config={"response_modalities": ["TEXT", "IMAGE"]})
 
-    image_block = next(
-        block
-        for block in response.content
-        if isinstance(block, dict) and block.get("image_url")
-    )
+    # image_block = next(
+    #     block
+    #     for block in response.content
+    #     if isinstance(block, dict) and block.get("image_url")
+    # )
 
-    base64_image = image_block["image_url"].get("url").split(",")[-1]
-    image_path = os.path.join(output_directory, f"{str(uuid.uuid4())}.jpg")
-    with open(image_path, "wb") as f:
-        f.write(base64.b64decode(base64_image))
+    # base64_image = image_block["image_url"].get("url").split(",")[-1]
+    # image_path = os.path.join(output_directory, f"{str(uuid.uuid4())}.jpg")
+    # with open(image_path, "wb") as f:
+    #     f.write(base64.b64decode(base64_image))
 
-    return image_path
+    # return image_path
+    return ""
 
 
 async def get_image_from_pexels(prompt: str, output_directory: str) -> str:

servers/fastapi/requirements.txt

@@ -19,7 +19,7 @@ dnspython==2.7.0
 email_validator==2.2.0
 fastapi==0.115.12
 fastapi-cli==0.0.7
-fastembed==0.7.0
+fastembed_vectorstore==0.1.5
 filelock==3.18.0
 filetype==1.2.0
 flatbuffers==25.2.10