236d1ddbd8
Full-stack application combining LlamaIndex vector search with Neo4j knowledge graph (GraphRAG) for answering queries about Netflix marketing materials. Flask/Hypercorn backend with custom ReAct agent, React frontend. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
116 lines
No EOL
4.2 KiB
Markdown
116 lines
No EOL
4.2 KiB
Markdown
```mermaid
|
|
graph TD
|
|
%% Main system components
|
|
Client[Chat Interface Frontend]
|
|
Server[Flask Server]
|
|
MongoDB[(MongoDB)]
|
|
Neo4j[(Neo4j)]
|
|
LLM[OpenAI LLM API]
|
|
|
|
%% Initialization flow
|
|
Init[System Initialization] --> |1. Start server| Server
|
|
Server --> |2. Connect to MongoDB| MongoDB
|
|
Server --> |3. Load or create index| IdxInit
|
|
Server --> |4. Connect to Neo4j| Neo4j
|
|
|
|
%% Subgraphs for clarity
|
|
subgraph "Initialization"
|
|
IdxInit[Initialize Global Index]
|
|
IdxInit --> |Load existing index| LoadIdx[Load from storage]
|
|
IdxInit --> |Build new index| BuildIdx[Process documents]
|
|
|
|
BuildIdx --> LlamaParser[LlamaParse Text & Images]
|
|
LlamaParser --> VectorIdx[Create Vector Index]
|
|
LlamaParser --> ExtractImages[Extract Images]
|
|
|
|
LoadIdx --> LoadGraph[Load GraphRAG components]
|
|
VectorIdx --> CreateGraph[Create GraphRAG components]
|
|
|
|
CreateGraph --> KGExtract[Extract entities & relations]
|
|
KGExtract --> CommDetect[Community detection]
|
|
CommDetect --> CommSummary[Generate community summaries]
|
|
|
|
LoadGraph -.-> |Check cache| CommCache[Load community cache]
|
|
end
|
|
|
|
%% Chat processing flow
|
|
Client --> |5. Send message with sessionId| Server
|
|
Server --> |6. Get session state| SessionManager
|
|
|
|
subgraph "Session Management"
|
|
SessionManager[Session Manager]
|
|
SessionManager --> |Check cache| MemCache[In-memory cache]
|
|
SessionManager --> |Check DB| SessionDB[Session in MongoDB]
|
|
SessionManager --> |Create if needed| NewSession[Create new session]
|
|
NewSession --> NewConv[Create conversation]
|
|
end
|
|
|
|
Server --> |7. Store user message| MongoDB
|
|
Server --> |8. Process query| Agent
|
|
|
|
subgraph "Agent Workflow"
|
|
Agent[ReAct Agent]
|
|
Agent --> |Select tool| Tools
|
|
|
|
subgraph "Query Tools"
|
|
Tools[Query Tools]
|
|
VectorTool[Vector Query Tool]
|
|
GraphTool[GraphRAG Tool]
|
|
|
|
Tools --> VectorTool
|
|
Tools --> GraphTool
|
|
|
|
VectorTool --> VectorRetriever[Vector Retriever]
|
|
VectorRetriever --> VectorRank[Top-K Similarity]
|
|
|
|
GraphTool --> DualRetrieval[Dual Retrieval]
|
|
DualRetrieval --> VectorRetriever
|
|
DualRetrieval --> GraphRetriever[Graph Community Retriever]
|
|
|
|
GraphRetriever --> ExtractEntities[Extract entities]
|
|
ExtractEntities --> MapCommunities[Map to communities]
|
|
MapCommunities --> GetSummaries[Get community summaries]
|
|
end
|
|
|
|
VectorRank --> |Context| GenResponse[Generate Response]
|
|
GetSummaries --> |Context| GenResponse
|
|
GenResponse --> |Use LLM| LLM
|
|
end
|
|
|
|
Agent --> |9. Format response with sources & images| Response
|
|
Response --> |10. Store assistant message| MongoDB
|
|
Response --> |11. Send response to client| Client
|
|
|
|
%% Image handling flow
|
|
subgraph "Image Processing"
|
|
ImgProcess[Image Handling]
|
|
ExtractImages --> SaveImages[Save to images directory]
|
|
SaveImages --> LinkToNodes[Link images to nodes]
|
|
Client --> |12. Request images| ImgEndpoint[Image endpoint]
|
|
ImgEndpoint --> ServeImage[Serve image files]
|
|
end
|
|
|
|
%% Conversation management
|
|
subgraph "Conversation Management"
|
|
ConvMgmt[Conversation Management]
|
|
Client --> |List conversations| GetConvs[Get conversations]
|
|
GetConvs --> MongoDB
|
|
Client --> |Get messages| GetMsgs[Get messages]
|
|
GetMsgs --> MongoDB
|
|
Client --> |Create new conversation| CreateConv[Create conversation]
|
|
CreateConv --> MongoDB
|
|
Client --> |Delete conversation| DeleteConv[Delete conversation]
|
|
DeleteConv --> MongoDB
|
|
end
|
|
|
|
%% Styling
|
|
classDef primary fill:#3498db,stroke:#2980b9,color:white;
|
|
classDef secondary fill:#2ecc71,stroke:#27ae60,color:white;
|
|
classDef storage fill:#9b59b6,stroke:#8e44ad,color:white;
|
|
classDef apiService fill:#e74c3c,stroke:#c0392b,color:white;
|
|
|
|
class Client,Server,Agent,Response primary;
|
|
class SessionManager,Tools,ImgProcess,ConvMgmt secondary;
|
|
class MongoDB,Neo4j storage;
|
|
class LLM apiService;
|
|
``` |