Skip to content

[VectorRAG] Implement Re-ranking Step in RAG Pipeline #10

New issue
New issue
Open
#133
Open
[VectorRAG] Implement Re-ranking Step in RAG Pipeline#10
#133
Assignees
fishonamos
Labels
component:vector-ragonlydust-waveContribute to awesome OSS repos during OnlyDust's open source week
@dineshpinto

Description

@dineshpinto
dineshpinto
opened on May 7, 2025
  • Description:
    Enhance the RAG pipeline by adding a re-ranking step after the initial retrieval phase. This step aims to improve the relevance of the documents retrieved by the vector search, potentially using more computationally intensive models like cross-encoders or an LLM to re-evaluate the top N results.
  • Acceptance Criteria:
    • A re-ranking mechanism is integrated into the retrieval pipeline, processing the initial set of documents returned by the vector search.
    • Support for at least one re-ranking method (e.g., using a cross-encoder model, or a simple LLM-based relevance check) is implemented.
    • The re-ranking step demonstrably improves the relevance of search results based on defined metrics or qualitative assessment.
  • Related issues: [VectorRAG] Implement Data Ingestion & Retrieval Pipeline for Vector Databases using Qdrant #8
    • Key Files/Modules Involved (Tentative):
    • flare_ai_kit/rag/vector/retriever.py (New file to be created)

Metadata

Metadata

Assignees

Labels

component:vector-ragonlydust-waveContribute to awesome OSS repos during OnlyDust's open source week

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions