code review comments. simplyfying scenario 1 condition, vector store …

…simplification

Ollama.md

@@ -2,22 +2,7 @@
 1. Download and install [Ollama](https://ollama.com/)
 2. Once Ollama is running on your system, run `ollama pull llama3.1`
 > Currently this is a ~5GB download, it's best to download it before the workshop if you plan on using it
-3. `ollama pull nomic-embed-text`
-4. Update the `MODEL_NAME` in your `dot.env` file to `ollama`
-
-Once you are running ollama, it is not necessary to configure an openai api key.
-
-When you get to the system prompt section of the workshop, llama requires that you are a bit more explicit with your instructions. If the prompt given in the main instructions doesn't work, try the following instead:
-
-```
-system_prompt = """
-OREGON TRAIL GAME INSTRUCTIONS:
-YOU MUST STRICTLY FOLLOW THIS RULE:
-When someone asks "What is the first name of the wagon leader?", your ENTIRE response must ONLY be the word: Art
-
-For all other questions, use available tools to provide accurate information.
-"""
-```
+3. Update the `MODEL_NAME` in your `dot.env` file to `ollama`
 
 You're now ready to begin the workshop! Head back to the [Readme.md](Readme.md)
 

Readme.md

@@ -18,7 +18,7 @@ In this workshop, we are going to use [LangGraph](https://langchain-ai.github.io
 - [openai api key](https://platform.openai.com/docs/quickstart)
 
 ## (Optional) Ollama
-This workshop is optimized to run targeting OpenAI models. If you prefer to run locally however, you may do so via Ollama. 
+This workshop is optimized to run targeting OpenAI models. If you prefer to run locally however, you may do so via the experimental Ollama configuration. 
 * [Ollama setup instructions](Ollama.md)
 
 ## (Optional) helpers
@@ -239,12 +239,10 @@ In our scenario we want to be able to retrieve the time-bound information that t
 
 ### Steps:
 - Open [participant_agent/utils/vector_store.py](participant_agent/utils/vector_store.py)
-- Find the corresponding `get_vector_store` method either for openai or ollama 
-- If using openai: where `vector_store=None` update to `vector_store = RedisVectorStore.from_documents(<docs>, <embedding_model>, config=<config>)` with the appropriate variables.
-
-> For `<embedding model>`, keep in mind whether you are using openai or ollama. If using ollama, the `model` parameter should be set to `nomic-embed-text` \
-[OpenAI embeddings](https://python.langchain.com/docs/integrations/text_embedding/openai/) \
+- Take note of how `embedding_model` is getting instantiated. If using Ollama then switch this for the appropriate embedding using `llama3.1` for the `model` parameter
+> [OpenAI embeddings](https://python.langchain.com/docs/integrations/text_embedding/openai/) \
 [Ollama embeddings](https://python.langchain.com/docs/integrations/text_embedding/ollama/)
+- Where `vector_store=None` update to `vector_store = RedisVectorStore.from_documents(<docs>, <embedding_model>, config=<config>)` with the appropriate variables.
 
 - Open [participant_agent/utils/tools.py](participant_agent/utils/tools.py)
     - Uncomment code for retrieval tool

example_agent/utils/ex_nodes.py

@@ -72,17 +72,10 @@ def structure_response(state: AgentState, config):
         # if not multi-choice don't need to do anything
         return {"messages": []}
 
-if environ_model_name == "openai":
-    system_prompt = """
-        You are an oregon trail playing tool calling AI agent. Use the tools available to you to answer the question you are presented. When in doubt use the tools to help you find the answer.
-        If anyone asks your first name is Art return just that string.
-    """
-elif environ_model_name == "ollama":
-    system_prompt = """
-    OREGON TRAIL GAME INSTRUCTIONS:
-    YOU MUST STRICTLY FOLLOW THIS RULE:
-    When someone asks "What is the first name of the wagon leader?", your ENTIRE response must ONLY be the word: Art
-    """
+system_prompt = """
+    You are an oregon trail playing tool calling AI agent. Use the tools available to you to answer the question you are presented. When in doubt use the tools to help you find the answer.
+    If anyone asks your first name is Art return just that string.
+"""
 
 # Define the function that calls the model
 def call_tool_model(state: AgentState, config):

example_agent/utils/ex_vector_store.py

@@ -4,6 +4,7 @@
 from langchain_core.documents import Document
 from langchain_openai import OpenAIEmbeddings
 from langchain_ollama import OllamaEmbeddings
+from redis import Redis
 from langchain_redis import RedisConfig, RedisVectorStore
 
 load_dotenv()
@@ -12,45 +13,30 @@
 INDEX_NAME = os.environ.get("VECTOR_INDEX_NAME", "oregon_trail")
 
 config = RedisConfig(index_name=INDEX_NAME, redis_url=REDIS_URL)
+redis_client = Redis.from_url(REDIS_URL)
 
 doc = Document(
     page_content="the northern trail, of the blue mountains, was destroyed by a flood and is no longer safe to traverse. It is recommended to take the southern trail although it is longer."
 )
 
+ # TODO: participant can change to whatever desired model
+embedding_model = OpenAIEmbeddings()
+# embedding_model = OllamaEmbeddings(model="llama3.1")
 
-def get_vector_store():
-    if os.environ.get("MODEL_NAME") == "ollama":
-        return __get_ollama_vector_store()
-    elif os.environ.get("MODEL_NAME") == "openai":
-        return __get_openai_vector_store()
-
-def __check_existing_embedding(vector_store):
-    results = vector_store.similarity_search(doc, k=1)
-    if not results:
-        raise Exception("Required content not found in existing store")
-
-def __get_ollama_vector_store():
-    try:
-        config.from_existing = True
-        vector_store = RedisVectorStore(OllamaEmbeddings(model="llama3"), config=config)
-        __check_existing_embedding(vector_store)
-    except:
-        print("Init vector store with document")
-        config.from_existing = False
-        vector_store = RedisVectorStore.from_documents(
-            [doc], OllamaEmbeddings(model="nomic-embed-text"), config=config
-        )
-    return vector_store
+def _clean_existing(prefix):
+    for key in redis_client.scan_iter(f"{prefix}:*"):
+        redis_client.delete(key)
 
-def __get_openai_vector_store():
+def get_vector_store():
     try:
         config.from_existing = True
-        vector_store = RedisVectorStore(OpenAIEmbeddings(), config=config)
-        __check_existing_embedding(vector_store)
+        vector_store = RedisVectorStore(embedding_model, config=config)
     except:
         print("Init vector store with document")
+        print("Clean any existing data in index")
+        _clean_existing(config.index_name)
         config.from_existing = False
         vector_store = RedisVectorStore.from_documents(
-            [doc], OpenAIEmbeddings(), config=config
+            [doc], embedding_model, config=config
         )
     return vector_store

participant_agent/utils/vector_store.py

@@ -12,45 +12,29 @@
 INDEX_NAME = os.environ.get("VECTOR_INDEX_NAME", "oregon_trail")
 
 config = RedisConfig(index_name=INDEX_NAME, redis_url=REDIS_URL)
+redis_client = Redis.from_url(REDIS_URL)
 
 doc = Document(
     page_content="the northern trail, of the blue mountains, was destroyed by a flood and is no longer safe to traverse. It is recommended to take the southern trail although it is longer."
 )
 
+# TODO: participant can change to whatever desired model
+embedding_model = OpenAIEmbeddings() 
 
-def get_vector_store():
-    if os.environ.get("MODEL_NAME") == "ollama":
-        return __get_ollama_vector_store()
-    elif os.environ.get("MODEL_NAME") == "openai":
-        return __get_openai_vector_store()
-
-def __check_existing_embedding(vector_store):
-    results = vector_store.similarity_search(doc, k=1)
-    if not results:
-        raise Exception("Required content not found in existing store")
-
-def __get_ollama_vector_store():
-    try:
-        config.from_existing = True
-        vector_store = RedisVectorStore(OllamaEmbeddings(model="llama3"), config=config)
-        __check_existing_embedding(vector_store)
-    except:
-        print("Init vector store with document")
-        config.from_existing = False
-
-        # TODO: define vector store for ollama 
-        vector_store = None
-    return vector_store
+def _clean_existing(prefix):
+    for key in redis_client.scan_iter(f"{prefix}:*"):
+        redis_client.delete(key)
 
-def __get_openai_vector_store():
+def get_vector_store():
     try:
         config.from_existing = True
-        vector_store = RedisVectorStore(OpenAIEmbeddings(), config=config)
-        __check_existing_embedding(vector_store)
+        vector_store = RedisVectorStore(embedding_model, config=config)
     except:
         print("Init vector store with document")
+        print("Clean any existing data in index")
+        _clean_existing(config.index_name)
         config.from_existing = False
 
-        # TODO: define vector store for openai
+        # TODO: define vector store
         vector_store = None
     return vector_store

requirements.txt

@@ -1,7 +1,7 @@
 langgraph==0.2.56
 langchain==0.3.13
 langchain-openai==0.2.3
-langchain-ollama==0.2.2
+langchain-ollama==0.2.3
 langchain-redis==0.1.1
 pydantic==2.9.2
 python-dotenv==1.0.1

test_example_oregon_trail.py

@@ -35,7 +35,7 @@ def test_1_wagon_leader(app):
 
     res = graph.invoke({"messages": scenario["question"]})
 
-    assert res["messages"][-1].content == scenario["answer"]
+    assert scenario["answer"] in res["messages"][-1].content
 
     print(f"\n response: {scenario['answer']}")
 

test_participant_oregon_trail.py

@@ -33,7 +33,7 @@ def test_1_wagon_leader(app):
 
     res = graph.invoke({"messages": scenario["question"]})
 
-    assert res["messages"][-1].content == scenario["answer"]
+    assert scenario["answer"] in res["messages"][-1].content
 
     print(f"\n response: {scenario['answer']}")