Documenttaion

[Remya-Ramachandran]

Is there any documentation available for the endpoint http://localhost:8087/retrieve_chatbot?
I'm particularly interested in understanding how the "mode" parameter functions, as it supports values like "vector", "graph", etc.
Could you please provide details on what each mode does, especially the actual functionality and processing logic behind the diferrent values in "graph" mode?

[kaustubh-darekar]

Currently there is no documentation but we are coming with blog post to explain retrievers.
Here is simple explaination of retrievers.

1. Graph
   Purpose: Explore relationships between entities
   How: Converts natural language questions into Cypher queries using GraphCypherQAChain
   When to Use: For structured queries — like “What companies has Alice worked for?”

2. Vector
   Purpose: Search based on meaning
   How: Matches query vector with document vectors
   When to Use: For semantic search across unstructured text

3. Full-Text
   Purpose: Find exact keyword matches
   How: Combines vector and keyword search using hybrid mode
   When to Use: When specific terms or phrases matter

4. Graph + Vector
   Purpose: Combine structure (graph) and meaning (vector)
   How: Uses a graph query first, then refines with vector similarity
   When to Use: When both entity relationships and content matter

5.Graph + Vector + Full-Text
Purpose: Use everything — graph structure + meaning + keywords
How: Enables the most comprehensive retrieval mode
When to Use: For complex questions that require meaning plus terms search

6. Entity + Vector
   Purpose: Focus on named entities like people, organizations
   How: Extracts entities and finds contextually similar documents
   When to Use: For entity-aware applications like resumes, company profiles

7. Global Search (Community-Based)
   Purpose: Summarize or search across large documents using communities
   How: Uses Graph Data Science clustering to group related nodes
   When to Use: For summarization, overview, or long document context

you will find the logic of each mode in constant.py here