Skip to main content
Playground

Overview

Entry point for multi-knowledge-base aggregated Q&A experience

The Playground is an interactive debugging panel for knowledge bases. Select one or more knowledge bases, enter a question, and instantly view retrieval results and model-generated answers.
Knowledge-based Q&A consumes free quota or incurs pay-as-you-go charges.

Features

FeatureDescription
Retrieval / Q&A ToggleToggle between Knowledge Q&A (retrieval followed by LLM-generated natural language answers) and Knowledge Retrieval (returns only relevant chunks, no generation) at the top
Model SwitchingSupports Qwen-series models; enables side-by-side comparison of generation quality
Citation HighlightingAnswers include source citations (e.g., [1], [2]); click to expand and navigate to the original chunk
Knowledge Source VisualizationDisplays matched document cards and data source counts

Usage Steps

1

Enter the Playground

Log in to the console and navigate to the Playground.
2

Select a Knowledge Base

Click on a knowledge base card from the list.
Select a knowledge base
3

Choose Retrieval or Q&A Mode

Toggle at the top between Knowledge Q&A (retrieval followed by natural language answer generation) and Knowledge Retrieval (returns only a list of relevant chunks).
Toggle between retrieval and Q&A modes
4

Enter Your Question

Type your question into the input box at the bottom and press Enter to submit. The system first retrieves relevant chunks, then invokes the large language model to generate an answer.
5

View Results

The results area displays:
  • The model-generated answer text
  • Citation markers (e.g., [1], [2]); click to expand the original text
  • A list of retrieved chunks with relevance scores
Q&A results and citation annotations

Best Practices

The Playground is ideal for validating knowledge base retrieval quality and Q&A performance. Before integrating into production applications, we recommend running several test rounds in the Playground to ensure recall and answer quality meet expectations.
  • Chunk Quality Check: Examine whether returned chunks contain correct answers. If recall is low, refine your chunking strategy.
  • Model Comparison: Switch among different Qwen models to compare answer quality and select the one best suited to your use case.

Differences from the Knowledge Base Detail Page

Entry PointUse Case
PlaygroundQuickly validate retrieval and Q&A performance across one or more knowledge bases
Knowledge Base Detail Page → Knowledge RetrievalPerform recall testing on a single knowledge base
Knowledge Base Detail Page → Knowledge Q&APerform Q&A testing on a single knowledge base