Add query caching to QAChain

[anyxling]

Description

from langchain_arangodb import ArangoGraphQAChain
from arango import ArangoClient

# Connect to ArangoDB
db = ArangoClient().db(...)

# Initialize components
graph = ArangoGraph()
emb_model = OpenAIEmbeddings(model="text-embedding-3-large")  # Replace with your embedding model
llm = ChatOpenAI(model="gpt-4o", temperature=0)  # Replace with your LLM

# Create the chain
chain = ArangoGraphChain.from_llm(
    llm=llm,
    graph=graph,
    embedding=emb_model,
    query_cache_collection_name="Queries",
)

# Ask a question
query = "How old is Cersei Lannister?"  # Replace with your query
chain.invoke({
    "query": query, 
    "use_query_cache": True  # Optional: remove to force LLM-based AQL generation,
   "query_cache_similarity_threshold": 0.80  # default is 0.80 or replace with your desired value
}) 

As shown in the screenshot below, when the exact query exists in the cache, it is reused directly with a similarity score of 1.0.

If we ask a semantically similar question like:

query = "What is the age of Cersei Lannister"
chain.invoke({"query": query, "use_query_cache": True})

The system will perform vector search and fetch the AQL query with the highest similarity score (0.92 in this case).

If no cached query is greater than the threshold(0.80), LLM-based AQL generation will be invoked.

Type of Change

• New feature
• Bug fix
• Breaking change
• Project configuration change

Complexity

Note

Please provide an estimated complexity of this PR of either Low, Medium or High

How Has This Been Tested?

• Unit tests
• Integration tests
• Manual tests

Checklist

• Unit tests updated
• Integration tests updated
• CHANGELOG.md updated

[aMahanna]

looks good! here's the 1st review round

[aMahanna]

LGTM! thank you, feel free to merge