35 lines (28 loc) · 1.61 KB

ROADMAP

Suggestions for Future Extensions

Security & Access Control

Authentication: Add JWT or OAuth2 authentication for API access
Multi-tenancy: Support multiple users/organizations with isolated document stores
Audit logging: Track document uploads, queries, and user actions

Scalability

Async processing: Queue document processing for large files
Caching: Add Redis caching for frequent queries
Horizontal scaling: Support multiple worker instances with shared vector store

Enhanced RAG Capabilities

Hybrid search: Combine vector similarity with keyword search (BM25)
Re-ranking: Add cross-encoder re-ranking for improved retrieval
Query routing: Route queries to specialized agents based on intent
Structured extraction: Extract structured data (rates, terms, fees) from documents

User Experience

Streaming responses: Stream LLM responses for better perceived latency
Feedback loop: Allow users to rate responses for continuous improvement
Export functionality: Export comparisons and checklists as PDF
Webhook notifications: Notify when document processing completes

Observability

Metrics: Add Prometheus metrics for request latency, token usage
Tracing: Integrate OpenTelemetry for distributed tracing
Logfire integration: Use pydantic-ai's Logfire for LLM observability

Data Management

Document versioning: Track document versions and changes over time
Batch operations: Support bulk document upload and processing
Data retention policies: Auto-delete documents after configurable period