Features
Our Process and Forget RAG system combines powerful features with privacy-first design, ensuring sensitive data never persists beyond its use.
Process & Forget RAG Pipeline
Extract knowledge from documents without storing originals - data is processed, vectorized, and immediately discarded for maximum privacy.
- Documents are converted to embeddings and then immediately deleted
- Original content never persists beyond processing
- Only semantic vectors remain, preserving privacy while enabling retrieval
- API designed around this security-first approach
Privacy-First Knowledge Bases
Access insights from multiple sources without persistent storage of sensitive source content, protecting confidential information.
- Only semantic vectors are stored, not the actual document contents
- Support for chunking and embedding various document types
- Metadata extraction system maintains necessary context without sensitive info
- Native support for any knowledge source without code changes
Bring Your Own Credentials
User-provided API keys for OpenAI, Pinecone and other services with each request - no server-side credential storage.
- Pass your own OpenAI API key with each request
- Use your Pinecone credentials without sharing them with our server
- Optional workspace-based credential management system
- Maximum control and security for your API access
Flexible Data Sources
Query custom indexes without mandatory credentials for unused services, making the API more adaptable to your needs.
- Connect to REST APIs like Jira and other documentation systems
- Web page scraping capabilities for documentation
- Extensible system for any data source
- Custom index support with flexible metadata extraction
Secure Ingestion Pipeline
Process documents with immediate deletion after vector storage to enhance privacy with our 'process and forget' approach.
- Automated ingestion of documents from various sources
- Immediate document deletion after processing
- Configurable chunking and embedding strategies
- Supports PDFs, text files, HTML, and more
Modular Authentication
Workspace-based credential system with JWT authentication for customizable security and flexible integration.
- JWT-based authentication with configurable expiration
- Organize credentials by workspace
- Securely manage API access
- Credentials are never exposed in responses
Streamlined API
Clean, consistent API design with proper error handling and clear responses.
- 15+ carefully designed endpoints
- Consistent request and response formats
- Detailed error messages and validation
- Comprehensive documentation
Flexible Metadata Registry
Dynamic metadata extraction system that can handle any custom index without hardcoded index types.
- Registry of extractor functions mapped to index types
- Default extractor for unknown index types
- Runtime registration of custom extractors
- Backward compatibility with existing extractors
Custom Index Combinations
Query multiple indexes with a single request, with control over result count from each source.
- Specify custom indexes via comma-separated list
- Control number of results from each index
- Consistent metadata structure across all sources
- Mix and match sources for comprehensive results
Core Benefits
Our approach provides significant advantages for organizations that need to maintain privacy while leveraging the power of AI.
Enhanced Privacy
Original documents are immediately deleted after processing, ensuring sensitive information is never stored longer than necessary.
Greater Flexibility
Query custom indexes with your own credentials, only providing what's needed for each request without mandatory service dependencies.
Reduced Liability
By not storing sensitive data, your organization reduces the risk associated with data breaches while still enabling AI-enhanced experiences.