feat: updated docker file to use the open-text-embeddings package from pypi and clean up

Dockerfile

@@ -19,7 +19,7 @@ RUN chmod +x *.sh && \
 
 # Stage 3 - final runtime image
 # Grab a fresh copy of the Python image
-FROM python:3.10-slim
+FROM python:3.11-slim
 
 # Include global args in this stage of the build
 ARG MODEL
@@ -31,11 +31,9 @@ ENV HOST=0.0.0.0
 ENV PORT=7860
 
 COPY --from=build-image ${MODEL} ${MODEL}
-COPY open/text/embeddings ./open/text/embeddings
-COPY server-requirements.txt ./
 COPY ./start_server.sh ./
 COPY ./index.html ./
-RUN pip install --no-cache-dir -r server-requirements.txt && \
+RUN pip install --no-cache-dir open-text-embeddings[server] && \
     chmod +x ./start_server.sh
 
 # Expose a port for the server

index.html

@@ -7,7 +7,7 @@
     <h1>BAAI/bge-large-en OpenAI API-Compatible Endpoint</h1>
     <p>
       With the utilization of the
-      <a href="https://github.com/limcheekin/open-text-embeddings"
+      <a href="https://pypi.org/project/open-text-embeddings/"
         >open-text-embeddings</a
       >
       package, we are excited to introduce the text embeddings model hosted in

open/text/embeddings/server/__main__.py

@@ -1,37 +0,0 @@
-"""FastAPI server for open-text-embeddings.
-
-To run this example:
-
-```bash
-pip install -r --no-cache-dir server-requirements.txt
-```
-
-Then run:
-```
-MODEL=intfloat/e5-large-v2 python -m open.text.embeddings.server
-```
-
-Then visit http://localhost:8000/docs to see the interactive API docs.
-
-"""
-import uvicorn
-from fastapi.responses import HTMLResponse
-from open.text.embeddings.server.app import create_app
-import os
-
-app = create_app()
-
-# Read the content of index.html once and store it in memory
-with open("index.html", "r") as f:
-    content = f.read()
-
-
-@app.get("/", response_class=HTMLResponse)
-async def read_items():
-    return content
-
-if __name__ == "__main__":
-    uvicorn.run(app,
-                host=os.environ["HOST"],
-                port=int(os.environ["PORT"])
-                )

open/text/embeddings/server/app.py

@@ -1,116 +0,0 @@
-
-from typing import List, Optional, Union
-from starlette.concurrency import run_in_threadpool
-from fastapi import FastAPI, APIRouter
-from fastapi.middleware.cors import CORSMiddleware
-from pydantic import BaseModel, Field
-from langchain.embeddings import HuggingFaceEmbeddings
-from langchain.embeddings import HuggingFaceInstructEmbeddings
-from langchain.embeddings import HuggingFaceBgeEmbeddings
-import os
-
-router = APIRouter()
-
-DEFAULT_MODEL_NAME = "intfloat/e5-large-v2"
-E5_EMBED_INSTRUCTION = "passage: "
-E5_QUERY_INSTRUCTION = "query: "
-BGE_EN_QUERY_INSTRUCTION = "Represent this sentence for searching relevant passages: "
-BGE_ZH_QUERY_INSTRUCTION = "为这个句子生成表示以用于检索相关文章："
-
-
-def create_app():
-    app = FastAPI(
-        title="Open Text Embeddings API",
-        version="0.0.2",
-    )
-    app.add_middleware(
-        CORSMiddleware,
-        allow_origins=["*"],
-        allow_credentials=True,
-        allow_methods=["*"],
-        allow_headers=["*"],
-    )
-    app.include_router(router)
-
-    return app
-
-
-class CreateEmbeddingRequest(BaseModel):
-    model: Optional[str] = Field(
-        description="The model to use for generating embeddings.", default=None)
-    input: Union[str, List[str]] = Field(description="The input to embed.")
-    user: Optional[str] = Field(default=None)
-
-    model_config = {
-        "json_schema_extra": {
-            "examples": [
-                {
-                    "input": "The food was delicious and the waiter...",
-                }
-            ]
-        }
-    }
-
-
-class Embedding(BaseModel):
-    embedding: List[float]
-
-
-class CreateEmbeddingResponse(BaseModel):
-    data: List[Embedding]
-
-
-embeddings = None
-
-
-def _create_embedding(
-    model: Optional[str],
-    input: Union[str, List[str]]
-):
-    global embeddings
-
-    if embeddings is None:
-        if model and model != "text-embedding-ada-002":
-            model_name = model
-        else:
-            model_name = os.environ["MODEL"]
-        print("Loading model:", model_name)
-        encode_kwargs = {
-            "normalize_embeddings": bool(os.environ.get("NORMALIZE_EMBEDDINGS", ""))
-        }
-        print("encode_kwargs", encode_kwargs)
-        if "e5" in model_name:
-            embeddings = HuggingFaceInstructEmbeddings(model_name=model_name,
-                                                       embed_instruction=E5_EMBED_INSTRUCTION,
-                                                       query_instruction=E5_QUERY_INSTRUCTION,
-                                                       encode_kwargs=encode_kwargs)
-        elif model_name.startswith("BAAI/bge-") and model_name.endswith("-en"):
-            embeddings = HuggingFaceBgeEmbeddings(model_name=model_name,
-                                                  query_instruction=BGE_EN_QUERY_INSTRUCTION,
-                                                  encode_kwargs=encode_kwargs)
-        elif model_name.startswith("BAAI/bge-") and model_name.endswith("-zh"):
-            embeddings = HuggingFaceBgeEmbeddings(model_name=model_name,
-                                                  query_instruction=BGE_ZH_QUERY_INSTRUCTION,
-                                                  encode_kwargs=encode_kwargs)
-        else:
-            embeddings = HuggingFaceEmbeddings(
-                model_name=model_name, encode_kwargs=encode_kwargs)
-
-    if isinstance(input, str):
-        return CreateEmbeddingResponse(data=[Embedding(embedding=embeddings.embed_query(input))])
-    else:
-        data = [Embedding(embedding=embedding)
-                for embedding in embeddings.embed_documents(input)]
-        return CreateEmbeddingResponse(data=data)
-
-
-@router.post(
-    "/v1/embeddings",
-    response_model=CreateEmbeddingResponse,
-)
-async def create_embedding(
-    request: CreateEmbeddingRequest
-):
-    return await run_in_threadpool(
-        _create_embedding, **request.dict(exclude={"user"})
-    )

server-requirements.txt

@@ -1,5 +0,0 @@
-fastapi
-sse-starlette
-sentence_transformers
-langchain
-uvicorn