chore: refactor ragas_once & split dependency

.github/workflows/image_build.yml

@@ -60,7 +60,7 @@ jobs:
         uses: docker/build-push-action@v5
         with:
           context: .
-          file: docker/Dockerfile
+          file: docker/Dockerfile.server
           platforms: linux/amd64,linux/arm64
           tags: |
             kubeagi/core-library-cli:latest

.github/workflows/image_build_test.yml

@@ -46,7 +46,7 @@ jobs:
         uses: docker/build-push-action@v5
         with:
           context: .
-          file: docker/Dockerfile
+          file: docker/Dockerfile.server
           platforms: linux/amd64,linux/arm64
           push: false
           build-args: |

.github/workflows/cli_test.yaml → .github/workflows/server_test.yaml

@@ -46,7 +46,7 @@ jobs:
           uses: docker/build-push-action@v5
           with:
             context: .
-            file: docker/Dockerfile
+            file: docker/Dockerfile.server
             load: true
             push: false
             tags: test-image
@@ -59,6 +59,7 @@ jobs:
             docker ps
         - name: Run test script
           run: |
+            sleep 10
             docker logs cli-container
             docker cp tests/example_test.sh cli-container:/tmp/example_test.sh
             docker exec cli-container bash /tmp/example_test.sh

Makefile

@@ -11,7 +11,13 @@ lint:
 .PHONY: install
 install:
 	@pip install -e libs/core/
-	@pip install -e libs/cli/
+	@pip install -e 'libs/cli/.[server, core, eval]'
+
+install-eval:
+	@pip install -e 'libs/cli/.[eval]'
+
+install-server:
+	@pip install -e 'libs/cli/.[server]'
 
 .PHONY: server
 server: install

docker/Dockerfile → docker/Dockerfile.server

@@ -17,12 +17,13 @@ RUN apt-get install -y gcc python3-dev
 # Official: https://pypi.org/simple
 ARG PYTHON_INDEX_URL=https://pypi.mirrors.ustc.edu.cn/simple/
 COPY libs /libs
+RUN python -m pip install --upgrade pip -i ${PYTHON_INDEX_URL}
 RUN python -m pip install ragas langchain sentencepiece protobuf -i ${PYTHON_INDEX_URL}
 WORKDIR /libs/core
 RUN pip install -e . -i ${PYTHON_INDEX_URL}
 
 WORKDIR /libs/cli
-RUN pip install -e . -i ${PYTHON_INDEX_URL}
+RUN pip install -e ".[server]" -i ${PYTHON_INDEX_URL}
 
 ENV RERANKING_MODEL_PATH=BAAI/bge-reranker-large
 

libs/cli/README.md

@@ -13,6 +13,12 @@ installation.
 
 - Install the Python SDK `pip install kubeagi-cli`
 
+- Install Server `pip install kubeagi-cli[server]`
+
+- Install Evaluation `pip install kubeagi-cli[eval]`
+
+- Install Core `pip install kubeagi-cli[core]`
+
 At this point, you should be able to run the following code:
 
 ```shell

libs/cli/kubeagi_cli/cli.py

@@ -16,9 +16,8 @@
 import typer
 
 from typing_extensions import Annotated
-from kubeagi_core.evaluation.ragas_eval import RagasEval
+
 from kubeagi_cli import convert
-from kubeagi_cli.server import webapp
 
 __version__ = "0.0.1"
 
@@ -52,6 +51,7 @@ def serve(
     ] = "info",
 ):
     import uvicorn
+    from kubeagi_cli.server import webapp
 
     uvicorn.run(app=webapp, host=host, port=port, log_level=log_level)
 
@@ -113,6 +113,8 @@ def evaluate(
         ),
     ] = None,
 ):
+    from kubeagi_cli.evaluation.ragas_eval import RagasEval
+
     print("evaluate RAG(Retrieval Augmented Generation)")
     # Initialize ragas_once with provided arguments
     eval = RagasEval(

libs/cli/kubeagi_cli/convert.py

@@ -14,14 +14,10 @@
 
 
 import typer
-import ujson
 
 from typing import List
 from typing_extensions import Annotated
 
-from kubeagi_core.document_transformers.pdf2csv import PDF2CSVTransform
-
-
 convert_cli = typer.Typer(no_args_is_help=True, add_completion=False)
 
 
@@ -56,6 +52,9 @@ def pdf(
         typer.Option(help="text chunk overlap"),
     ] = 50,
 ):
+    import ujson
+    from kubeagi_core.document_transformers.pdf2csv import PDF2CSVTransform
+
     """
     pdf transformer csv.
     """

libs/cli/pyproject.toml

@@ -17,16 +17,21 @@ classifiers = [
 ]
 dependencies = [
     "typer==0.9.0",
+]
+
+[project.optional-dependencies]
+dev = ["black==23.3.0", "pylint==3.1.0"]
+eval = ["langchain>=0.1.0", "ragas>=0.1.0"]
+server = [
     "fastapi==0.109.0",
     "uvicorn==0.27.0",
     "FlagEmbedding==1.2.3",
     "BCEmbedding==0.1.3",
-    "kubeagi-core==0.0.1",
-    "ujson==5.9.0",
+    "sentencepiece",
+    "protobuf",
+    "ujson==5.9.0"
 ]
-
-[project.optional-dependencies]
-dev = ["black==23.3.0", "pylint==3.1.0"]
+core = ["kubeagi-core==0.0.1"]
 
 [project.urls]
 Homepage = "https://github.com/kubeagi/core-library"

libs/core/README.md

@@ -13,6 +13,8 @@ installation.
 
 - Install the Python SDK `pip install kubeagi_core`
 
+- For experimental features (e.g. reading PDF files) use `pip install kubeagi_core[experiment]`
+
 At this point, you should be able to run the following code:
 
 ```python

libs/core/pyproject.toml

@@ -20,23 +20,23 @@ dependencies = [
     "emoji==2.2.0",
     "ftfy==6.1.1",
     "kubernetes==25.3.0",
-    "langchain>=0.1.0",
-    "ragas>=0.1.0",
     "opencc-python-reimplemented==0.1.7",
     "selectolax==0.3.17",
     "spacy==3.5.4",
     "zhipuai==1.0.7",
+]
+
+[project.optional-dependencies]
+dev = ["black==23.3.0", "pylint==3.1.0"]
+experiment = [
     "unstructured==0.12.0",
     "unstructured-inference==0.7.21",
     "unstructured.pytesseract==0.3.12",
     "pdf2image==1.17.0",
     "pdfminer.six==20231228",
-    "pikepdf==8.13.0",
+    "pikepdf==8.13.0"
 ]
 
-[project.optional-dependencies]
-dev = ["black==23.3.0", "pylint==3.1.0"]
-
 [project.urls]
 Homepage = "https://github.com/kubeagi/core-library"
 Issues = "https://github.com/kubeagi/core-library/issues"