Engineering consultancy for private AI systems and secure on-premise RAG implementations for business customers.
The source repository for the demo architecture is private; system architecture and assessments are available on request.
Chat interface with document upload and cited answers.
FileGPT.dev is positioned as a B2B engineering agency ('The Compliance Honeypot') for the DACH region. Rather than operating as a standardized SaaS product, the platform demonstrates expertise in implementing private, physically or logically isolated RAG systems. The focus is on 'Sovereignty by Design': companies retain full control over their data, models, and infrastructure.
Verifiable build status: Private on-premise deployment, dedicated retrieval, distributed access control, and operational documentation are implemented in the reference architecture.
Architecture
Loading diagram…
Features
Citation-first chat — streaming answers with explicit source badges like [Source: file.pdf, Page N] rendered as source chips.
Guarded hybrid retrieval — Gemini embeddings with vector search in Supabase (pgvector), plus keyword-style fallback when needed for exact-term recall.
Conversation memory — condenses recent messages into a standalone query before retrieval so follow-ups keep context.
Ingestion with lifecycle tracking — documents move PROCESSING → COMPLETE/FAILED with chunk counts; supports PDF/TXT/MD/DOCX/XLSX extraction.
Operational boundaries — authenticated API routes, user-scoped data access, distributed rate limiting, structured logs, and readiness checks.
Implemented trust boundaries
Data minimization: only retrieved top-k chunks are sent to the model for an answer, never entire documents.
Source attribution: every answer includes the originating document (and page when available).
Authentication and access: Supabase sessions plus RLS scope documents, chunks, chats, and usage data to the current user.
Operability: health/readiness endpoints, structured logging, and a production runbook support diagnosis and deployment.
Integration: upload and query flows run through authenticated API routes; the source repository remains private.
Security & operations
Access isolation: RLS and a private object-storage bucket prevent public or cross-user file access.
Abuse controls: distributed Upstash rate limiting is required for production endpoints; query-embedding caching reduces repeat work.
Deployment boundaries: production startup can fail closed when required runtime configuration or distributed rate limiting is missing.
