reprebot/feat: #3 add hugging face model to llm client

- use `ChatHuggingFace` to instantiate a `Hugging Face` models
- use `HuggingFaceEndpoint` to connect to the `Hugging Face` model
- add `hugging-face` to `LLMClient`
- add `TestHuggingFaceLLMClient` unit test #20
- update `langchain` to `0.1.12` version #23
- install `transformers` dependency
- install `Jinja2` dependency
- add `HUGGINGFACEHUB_API_TOKEN` to `pipeline.yml`

README.md

@@ -47,8 +47,10 @@ There are various ways to contribute to **Reprebot**:
 Reprebot relies on the following dependencies:
 
 - `chromadb`
+- `Jinja2`
 - `langchain`
 - `langchain-openai`
 - `pytest`
+- `transformers`
 
 For a comprehensive list of dependencies, including both direct and transitive dependencies, please refer to the `requirements.txt` file.

direct-requirements.txt

@@ -1,4 +1,6 @@
 chromadb==0.4.24
-langchain==0.1.11
+Jinja2==3.1.3
+langchain==0.1.12
 langchain-openai==0.0.8
-pytest==8.1.1
+pytest==8.1.1
+transformers==4.38.2

requirements.txt

@@ -22,7 +22,7 @@ fastapi==0.110.0
 filelock==3.13.1
 flatbuffers==24.3.7
 frozenlist==1.4.1
-fsspec==2024.2.0
+fsspec==2024.3.0
 google-auth==2.28.2
 googleapis-common-protos==1.63.0
 greenlet==3.0.3
@@ -35,17 +35,19 @@ huggingface-hub==0.21.4
 humanfriendly==10.0
 idna==3.6
 importlib-metadata==6.11.0
-importlib_resources==6.1.3
+importlib_resources==6.3.0
 iniconfig==2.0.0
+Jinja2==3.1.3
 jsonpatch==1.33
 jsonpointer==2.4
 kubernetes==29.0.0
-langchain==0.1.11
-langchain-community==0.0.27
-langchain-core==0.1.30
+langchain==0.1.12
+langchain-community==0.0.28
+langchain-core==0.1.32
 langchain-openai==0.0.8
 langchain-text-splitters==0.0.1
-langsmith==0.1.23
+langsmith==0.1.27
+MarkupSafe==2.1.5
 marshmallow==3.21.1
 mmh3==4.1.0
 monotonic==1.6
@@ -55,7 +57,7 @@ mypy-extensions==1.0.0
 numpy==1.26.4
 oauthlib==3.2.2
 onnxruntime==1.17.1
-openai==1.13.3
+openai==1.14.1
 opentelemetry-api==1.23.0
 opentelemetry-exporter-otlp-proto-common==1.23.0
 opentelemetry-exporter-otlp-proto-grpc==1.23.0
@@ -88,6 +90,7 @@ regex==2023.12.25
 requests==2.31.0
 requests-oauthlib==1.4.0
 rsa==4.9
+safetensors==0.4.2
 six==1.16.0
 sniffio==1.3.1
 SQLAlchemy==2.0.28
@@ -97,6 +100,7 @@ tenacity==8.2.3
 tiktoken==0.6.0
 tokenizers==0.15.2
 tqdm==4.66.2
+transformers==4.38.2
 typer==0.9.0
 typing-inspect==0.9.0
 typing_extensions==4.10.0
@@ -107,4 +111,4 @@ websocket-client==1.7.0
 websockets==12.0
 wrapt==1.16.0
 yarl==1.9.4
-zipp==3.17.0
+zipp==3.18.1

src/llm_client/__init__.py

@@ -7,6 +7,8 @@
 from langchain_community.vectorstores import Chroma
 from langchain_community.embeddings import FakeEmbeddings
 from langchain.chat_models.fake import FakeListChatModel
+from langchain_community.chat_models.huggingface import ChatHuggingFace
+from langchain_community.llms.huggingface_endpoint import HuggingFaceEndpoint
 
 class LLMClient:
     def __init__(self, model_type: str):
@@ -25,14 +27,11 @@ def setup_model(self, temperature=0):
             )
         elif self.model_type == "fake":
             model = FakeListChatModel(responses=["Hello",])
-        """
         elif self.model_type == "hugging-face":
             llm = HuggingFaceEndpoint(
                 repo_id="google/gemma-7b",
             )
             model = ChatHuggingFace(llm=llm)
-        # https://github.com/langchain-ai/langchain/issues/18639
-        """
         return model
 
     def setup_chain(self, retriever, prompt, model):

test/unit/src/test_llm_client.py

@@ -7,6 +7,7 @@
 from langchain_community.embeddings import FakeEmbeddings
 from langchain_core.runnables.base import RunnableSequence
 from langchain.chat_models.fake import FakeListChatModel
+from langchain_community.chat_models.huggingface import ChatHuggingFace
 
 
 class TestFakeLLMClient:
@@ -38,4 +39,35 @@ def test_setup_chain(self, llm_client):
     def test_query(self, llm_client):
         response = llm_client.query(user_input="")
         assert isinstance(response, str)
-        assert response == "Hello"
+        assert response == "Hello"
+
+
+class TestHuggingFaceLLMClient:
+    @pytest.fixture
+    def llm_client(self):
+        return LLMClient(model_type="hugging-face")
+
+    def test_init(self, llm_client):
+        assert llm_client.model_type == "hugging-face"
+
+    def test_setup_model(self, llm_client):
+        model = llm_client.setup_model()
+        assert isinstance(model, ChatHuggingFace)
+
+    def test_setup_chain(self, llm_client):
+        # Empty retriever for testing
+        retriever = Chroma.from_documents(
+            documents=[Document(page_content="")],
+            embedding=FakeEmbeddings(size=1),
+        ).as_retriever(search_kwargs={"k": 1})
+        # Empty prompt for testing
+        prompt = ChatPromptTemplate.from_messages([""])
+        # Default "fake" model use in this test
+        model = llm_client.setup_model()
+        # RAG chain
+        chain = llm_client.setup_chain(retriever=retriever, prompt=prompt, model=model)
+        assert isinstance(chain, RunnableSequence)
+
+    def test_query(self, llm_client):
+        response = llm_client.query(user_input="")
+        assert isinstance(response, str)