Bump LangChain to 0.0.350 and add Gemini tests (#153)

* Tests: add Google Gemini LLM

* Bump LC

* Fix

* fix

* fix

* fix format

* fo

Author: nicoloboschi

pyproject.toml

@@ -15,9 +15,9 @@ astrapy = "~0.6.2"
 cassio = "~0.1.3"
 unstructured = "^0.10"
 llama-index = { version = "0.9.14", extras = ["langchain"] }
-langchain = { version = "0.0.349", extras = ["openai"] }
-langchain-core = "0.0.13"
-langchain-community = "0.0.1"
+langchain = { version = "0.0.350", extras = ["openai"] }
+langchain-core = "0.1.1"
+langchain-community = "0.0.3"
 
 [tool.poetry.group.test.dependencies]
 pytest = "*"

ragstack-e2e-tests/e2e_tests/conftest.py

@@ -146,9 +146,11 @@ def dump_report():
         + str(tests_stats["skipped"])
         + "\n"
     )
+    all_report_lines.sort()
     with open("all-tests-report.txt", "w") as f:
         f.write(stats_str + "\n")
         f.write("\n".join(all_report_lines))
+    failed_report_lines.sort()
     with open("failed-tests-report.txt", "w") as f:
         f.write(stats_str + "\n")
         f.write("\n".join(failed_report_lines))

ragstack-e2e-tests/e2e_tests/langchain/chat_application.py renamed to ragstack-e2e-tests/e2e_tests/langchain/rag_application.py

@@ -0,0 +1,30 @@
+import pytest
+from e2e_tests.conftest import (
+    set_current_test_info,
+)
+from langchain_community.chat_models import ChatVertexAI
+from langchain_core.prompts import ChatPromptTemplate
+
+
+def set_test_info(chat: str):
+    set_current_test_info("langchain::chat", chat)
+
+
+@pytest.fixture
+def vertex_gemini():
+    return ChatVertexAI(model_name="gemini-pro")
+
+
+@pytest.mark.parametrize(
+    "chat",
+    ["vertex_gemini"],
+)
+def test_chat(chat, request):
+    set_test_info(chat)
+    chat_model = request.getfixturevalue(chat)
+    prompt = ChatPromptTemplate.from_messages(
+        [("human", "Hello! Where Archimede was born?")]
+    )
+    chain = prompt | chat_model
+    response = chain.invoke({})
+    assert "Syracuse" in response.content

ragstack-e2e-tests/e2e_tests/langchain/test_chat.py

@@ -11,7 +11,7 @@
     delete_all_astra_collections_with_client,
     delete_astra_collection,
 )
-from e2e_tests.langchain.chat_application import (
+from e2e_tests.langchain.rag_application import (
     run_rag_custom_chain,
     run_conversational_rag,
 )
@@ -228,7 +228,7 @@ def _run_test(test_case: str, vector_store, embedding, llm):
     vector_store = vector_store(embedding)
     llm_name, llm = llm
     set_current_test_info(
-        test_case,
+        "langchain::" + test_case,
         f"{llm_name},{embedding_name},{vector_store_name}",
     )
     if test_case == "rag_custom_chain":

ragstack-e2e-tests/e2e_tests/langchain/test_compatibility_matrix.py renamed to ragstack-e2e-tests/e2e_tests/langchain/test_compatibility_rag.py

@@ -8,7 +8,7 @@
 
 
 def set_current_test_info_document_loader(doc_loader: str):
-    set_current_test_info("document_loader", doc_loader)
+    set_current_test_info("langchain::document_loader", doc_loader)
 
 
 def test_csv_loader():

ragstack-e2e-tests/e2e_tests/langchain/test_document_loaders.py

@@ -0,0 +1,28 @@
+import pytest
+from e2e_tests.conftest import (
+    set_current_test_info,
+)
+
+# from llama_index.llms import Gemini
+
+
+def set_test_info(chat: str):
+    set_current_test_info("llama_index::chat", chat)
+
+
+# @pytest.fixture
+# def gemini():
+#     return Gemini(model="models/gemini-pro")
+
+
+@pytest.mark.parametrize(
+    "chat",
+    [
+        # "gemini",
+    ],
+)
+def test_chat(chat, request):
+    set_test_info(chat)
+    chat_model = request.getfixturevalue(chat)
+    response = chat_model.complete("Hello! Where Archimede was born?")
+    assert "Syracuse" in response.content

ragstack-e2e-tests/e2e_tests/llama_index/test_chat.py

@@ -204,7 +204,7 @@ def _run_test(vector_store, embedding, llm):
     vector_store = vector_store(embedding_dimensions)
     llm_name, llm = llm
     set_current_test_info(
-        "llama_index_retrieve",
+        "llama_index::rag",
         f"{llm_name},{embedding_name},{vector_store_name}",
     )
     storage_context = StorageContext.from_defaults(vector_store=vector_store)

ragstack-e2e-tests/e2e_tests/llama_index/test_llama_index.py renamed to ragstack-e2e-tests/e2e_tests/llama_index/test_compatibility_rag.py

@@ -13,6 +13,7 @@ pytest = "*"
 black = "*"
 ruff = "*"
 google-cloud-aiplatform = "^1.36.4"
+langchain-google-genai = "^0.0.4"
 boto3 = "^1.29.6"
 huggingface-hub = "^0.19.4"
 

ragstack-e2e-tests/pyproject.langchain.toml

@@ -13,6 +13,7 @@ pytest = "*"
 black = "*"
 ruff = "*"
 google-cloud-aiplatform = "^1.36.4"
+langchain-google-genai = "^0.0.4"
 boto3 = "^1.29.6"
 huggingface-hub = "^0.19.4"
 
@@ -21,9 +22,9 @@ beautifulsoup4 = "^4"
 
 llama-index = { git = "https://github.com/run-llama/llama_index.git", branch = "main" }
 
-langchain = { version = "0.0.349", extras = ["openai"] }
-langchain-core = "0.0.13"
-langchain-community = "0.0.1"
+langchain = { version = "0.0.350", extras = ["openai"] }
+langchain-core = "0.1.1"
+langchain-community = "0.0.3"
 astrapy = "^0.6"
 cassio = "^0.1"