The Dev Tools section in Search AI provides advanced configuration options and developer utilities to optimize search performance, extend functionality, and integrate custom solutions.
Navigation: Dev Tools menu in Search AI
Advanced Configuration
Advanced Configurations allow you to fine-tune retrieval and answer optimization settings for specific requirements.
Accessing Advanced Configuration
- Navigate to Dev Tools > Advanced Configurations
- Search for the configuration you want to modify
- Select or provide the appropriate values
Available Configurations
| Configuration | Description | Use Case |
|---|
| Re-Rank Chunks | Select the reranking feature and Re-Ranker model | Improve result relevance by reordering chunks based on semantic similarity |
| Re-Rank Chunk Fields | Select fields used to rerank chunks | Customize which chunk attributes influence reranking |
| Maximum Re-Rank Chunks | Set maximum chunks sent for reranking | Balance performance vs. quality by limiting reranking scope |
| Enable Exact KNN Matching | Enable Exact K-Nearest Neighbors matching | Improve precision for vector similarity searches |
| Single-Use URLs for Uploaded Documents | Enable secure, temporary access to uploaded documents | Enhance security for sensitive document access |
Configuration Details
Re-Ranking
Re-ranking improves search quality by applying a secondary model to reorder initially retrieved chunks based on deeper semantic analysis. Search AI supports the following re-rankers.
- Cross Encoder Re-Ranker - Uses the cross-encoder/ms-marco-MiniLM-L-6-v2 model. It’s lightweight, fast, and most suitable for English Language.
- BGE Re-Ranker - Uses BAAI/bge-reranker-v2-m3 model. It’s a lightweight re-ranking model and possesses multilingual capabilities.
- MixedBread Re-Ranker - Uses the mixedbread-ai/mxbai-rerank-large-v1 model, which is resource intensive to run and has a higher latency but delivers the highest accuracy and performance.
This feature doesn’t require training of the application.
| Setting | Purpose |
|---|
| Re-Rank Chunks | Enable/disable reranking and select model |
| Re-Rank Chunk Fields | Define which fields (title, content, metadata) to use for reranking. By default, Chunk Title, Chunk Text, and Record Title are used. Note that selecting different fields impacts the results generated by the re-ranker. |
| Maximum Re-Rank Chunks | Maximum number of chunks to be sent for reranking (Values: 5 - 20, default: 20) |
Increasing the number of fields or chunks for reranking may lead to higher latency due to the added complexity in computation, data retrieval, and processing load. If you are using the generative model and re-ranking is also enabled, the overall latency includes the latency induced by the re-ranker and the LLM.
- Exact KNN - Finds the truly exact neighbors by comparing the query with every vector of the indexed content. This type of matching guarantees higher accuracy, and precision but can be computationally expensive and can affect performance.
- Approximate KNN - Uses different techniques to find the nearest neighbors quickly rather than comparing with every vector. This method may provide a close match rather than the best one, but it’s faster and more scalable for large datasets.
Exact KNN (K-Nearest Neighbors) matching provides more precise vector similarity searches compared to approximate methods.
| Consideration | Impact |
|---|
| Accuracy | Higher precision in finding similar vectors |
| Performance | May increase query latency for large datasets |
| Use Case | Recommended when accuracy is prioritized over speed |
Enabling Exact KNN Match can introduce some latency and thereby increase the average Response Time.
This applies only to uploaded documents. References for data from connectors or web pages include a direct link to the corresponding web page or third-party application.
| Tool | Purpose | Key Features |
|---|
| RAG Evaluator | Evaluate RAG system performance | Measures search quality using RAGAS and CEQA frameworks; API integration; flexible results storage |
| Tool | Purpose | Key Features |
|---|
| Custom Connector SDK | Build custom data source integrations | Standardized data ingestion; metadata integrity; optimized for enterprise RAG applications |
| Tool | Purpose | Key Features |
|---|
| HTML to Structured Data Extractor | Extract content from HTML sources | Identifies tables of contents; preserves heading-content relationships; outputs JSON |
| Adobe Extraction Utility | Extract content from PDFs | Preserves original layout and structure; intelligent document parsing |
| Azure Extraction Utility | Extract from Azure-hosted documents | Uses Azure AI Document Intelligence; automatic content structuring |
| Google Document AI | Batch process documents from cloud storage | Automates extraction from unstructured/semi-structured documents |
| Salesforce Custom Extraction Utility | Extract from Salesforce Knowledge Base | Retains hierarchy and relationship structure |
| Utility | GitHub Link |
|---|
| HTML to Structured Data | View Repository |
| Adobe Extraction | View Repository |
| Azure Extraction | View Repository |
| Google Document AI | View Repository |
| Salesforce Extraction | View Repository |
Model Optimization
| Tool | Purpose | Key Features |
|---|
| Fine-Tune Embedding Utility | Fine-tune embedding models | Uses domain-specific documents; compares pre/post fine-tuning performance |
Quick Reference
Advanced Configuration Summary
| Category | Configurations |
|---|
| Retrieval Optimization | Re-Rank Chunks, Re-Rank Chunk Fields, Maximum Re-Rank Chunks |
| Vector Search | Enable Exact KNN Matching |
| Security | Single-Use URLs for Uploaded Documents |
| Category | Tools |
|---|
| Evaluation | RAG Evaluator |
| Integration | Custom Connector SDK |
| Extraction | HTML, Adobe, Azure, Google Document AI, Salesforce utilities |
| Optimization | Fine-Tune Embedding Utility |
| Scenario | Recommended Tool |
|---|
| Measure answer quality | RAG Evaluator |
| Connect custom data source | Custom Connector SDK |
| Ingest HTML documentation | HTML to Structured Data Extractor |
| Process PDF documents | Adobe or Azure Extraction Utility |
| Batch process cloud documents | Google Document AI |
| Extract Salesforce knowledge | Salesforce Custom Extraction Utility |
| Improve domain-specific search | Fine-Tune Embedding Utility |
