FRIS
Features

From scattered sources to one grounded answer

Every claim FRIS makes is traceable back to a specific source record.

Connect & Fuse
πŸ”Œ

20 Native Connectors

REST APIs, PostgreSQL, MySQL, CSV/Excel, Kafka, SAP (OData), Slack, Google Sheets, Airtable, Elasticsearch, and more, each self-documented in-app.

πŸ—ΊοΈ

Schema Auto-Detection

Every source's schema is auto-mapped, then shown to you to confirm or edit before the first sync.

🧬

Entity Fusion Engine

Fuzzy name matching, exact-identifier short-circuiting (domain/email/SKU), and source-trust + recency-based conflict resolution, written straight to Neo4j.

Search & Ask
πŸ”

Semantic Search

Automatic embeddings over the fused graph, with change-detection so re-embedding never happens twice, plus a backfill endpoint.

πŸ’¬

Natural-Language Ask

A retrieval-augmented Q&A layer that answers only from your grounded entity graph, and asks a clarifying question instead of guessing on low-confidence retrieval.

πŸ•ΈοΈ

Graph Explorer

An interactive Cytoscape-based visualization to click into any fused entity and trace its relationships across sources.

Trust & Control
πŸ“œ

Full Provenance Tracking

Every fused field keeps a record of which source it came from, so nothing is a black box.

πŸ”‘

Bring Your Own LLM

Use the hosted default, or bring your own OpenAI, Anthropic, Google, or Azure API key, encrypted credential storage either way.

πŸ“Š

Usage-Metered Billing

Every query, sync, and GB of storage logged with cost, on a live Usage & Costs dashboard.

Technical Specifications
Backend
Python / FastAPI, provider-agnostic LLM adapters, Docker Compose deployment.
Data stores
Neo4j for the fused entity graph, PostgreSQL for the control plane.
Embeddings
Local, CPU-based via sentence-transformers, no data leaves your deployment for embedding.
Frontend
React + Vite, Cytoscape.js for the Graph Explorer.
APIs
REST and GraphQL, plus an MCP server for agentic tooling.
Auth
Accounts with JWT and per-account API keys, used for billing attribution.
Auditing
Full audit-trail logging with filter/search and 30-day purge.
Not included today
No Kubernetes/Terraform packaging, no Flink for production-scale streaming, no RBAC or multi-tenant data isolation, accounts exist for billing attribution, not isolation. CSV/Excel uploads currently need a server-side file path. Self-described pre-seed / early-access build.

Stop reconciling data by hand

Connect your sources once, FRIS keeps the fused graph current.

Request early access