clinical-biomarkers / biomarker-issue-repo

Issues repo for the biomarker team.
0 stars 0 forks source link

Neo4j implementing LLM query search: Troubleshooting #13

Closed MiguelMazumder closed 19 hours ago

MiguelMazumder commented 2 months ago

Ticket contains information regarding adding LLM to Data Distillery Neo4j Knowledge Graph to upscale query process. Materials as well as steps taken so far trouble shooting will be included here.

LLM implementation instructions currently being followed (CypherQAChain process): LangChain Neo4j Integration

DataDistillery17Jul2024 build_conainer file and config file currently in use: build_container.sh

Container.cfg

your code here Current Troubleshooting Steps:

  1. Neo4j User Interface exists at "http://aws.glygen.org:7474/browser/" once container is running
  2. Neo4j/bolt Interface exists at http://aws.glygen.org:7678 once container is running
  3. Tried google colab notebook but due to permissions and location, cannot utilize Neo4jGraph function from langchain_community.graphs package
  4. Running in jupyter notebook in VScode attempting to run graph = Neo4jGraph(url=NEO4J_URI, username=NEO4J_USERNAME, password=NEO4J_PASSWORD) generated error: ClientError: {code: Neo.ClientError.Procedure.ProcedureRegistrationFailed} {message: apoc.meta.data is unavailable because it is sandboxed and has dependencies outside of the sandbox. Sandboxing is controlled by the dbms.security.procedures.unrestricted setting. Only unrestrict procedures you can trust with access to database internals.}
  5. Attempted to alter dbms.security.procedures.unrestricted setting in container.cfg file by adding the line: dbms.security.procedures.unrestricted=jwt.security.*,apoc.*,n10s.*. This resulted in the same error

*Currently troubleshooting APOC plug-in to Data Distillery Knowledge graph

MiguelMazumder commented 3 weeks ago

Getting access to biomarkerkb dev and prod server. Move Neo4j Data Distillery Knowledge Graph to there. Currently using a rag retrieval model on smaller non-medical datasets to test operability (https://neo4j.com/developer-blog/advanced-rag-strategies-neo4j/). A obstacle once I get server access is setting up the knowledge graph port listeners to find the bolt/neo4j ports on the server side. Will most likely need Sean's help with that