Myriade

Product Introduction

1. Myriade is an agentic AI data analyst designed for self-hosting, enabling users to connect directly to databases like PostgreSQL, MySQL, Snowflake, and BigQuery using natural language queries. It generates SQL queries and visualizations from plain English inputs while allowing users to review and edit outputs before execution. The system operates with strict privacy controls, including read-only database access, zero data retention, and encrypted processing.
2. The core value of Myriade lies in its ability to democratize data analysis by eliminating the need for SQL expertise while maintaining enterprise-grade security. It bridges the gap between business teams and technical stakeholders by providing transparent, auditable workflows and enforcing governance policies.

Main Features

1. Myriade uses agentic AI to autonomously reason through data questions, iteratively refining SQL logic, flagging uncertainties, and self-correcting errors before execution. The system provides full visibility into its thought process, including schema analysis, hypothesis testing, and query optimization steps.
2. Privacy-first architecture ensures read-only database access with zero data retention, encrypting sensitive information using zero-knowledge protocols and anonymizing outputs. All queries are processed without storing or training on user data, with automatic deletion of temporary artifacts post-execution.
3. Self-hosted deployment via a single-line Docker command enables full control over infrastructure, allowing organizations to run Myriade on-premises or in private clouds. The system integrates with existing monitoring tools through detailed audit logs tracking user activity, accessed tables, and query timestamps.

Problems Solved

1. Myriade addresses the bottleneck of relying on data teams for SQL queries by enabling business users to directly interrogate databases using natural language. It eliminates the risk of data exposure common in cloud-based AI tools through its self-hostable, zero-retention architecture.
2. The product serves regulated industries (finance, healthcare, EU-based enterprises) requiring GDPR/CNIL compliance, as well as non-technical business analysts in SMEs needing rapid insights without engineering support.
3. Typical use cases include diagnosing sudden metric changes ("Why did sales drop last week?"), auditing data quality for missing values/duplicates, generating custom reporting views, and identifying overlooked KPIs through AI-driven exploration.

Unique Advantages

1. Unlike basic NL2SQL tools, Myriade implements multi-agent debate systems that simulate human analyst reasoning, combining chain-of-thought prompting with schema-aware constraint validation. This reduces hallucination risks by 63% compared to GPT-4 fine-tuned models.
2. Patent-pending "Contextual Amnesia" ensures the AI resets its knowledge state between queries, preventing accidental data leakage across sessions. The system enforces role-based access controls that align with existing database permissions.
3. Competitive differentiation comes from combining Snowflake/BigQuery-native performance optimizations with EU-hosted infrastructure, offering <200ms query latency while meeting Article 35 GDPR requirements for data protection impact assessments.

Frequently Asked Questions (FAQ)

1. How does Myriade ensure query security? All database connections use read-only credentials with SSL encryption, while sensitive data fields are masked using format-preserving encryption (FPE) before processing. Query results are never written to disk and purged from memory within 5 minutes.
2. Can we customize the AI's logic? Users can define business rules through YAML configuration files that constrain query patterns, enforce naming conventions, and whitelist approved database schemas. The agent validates all generated SQL against these guardrails before execution.
3. What databases are supported? Myriade currently integrates with PostgreSQL 12+, MySQL 8+, Snowflake, and BigQuery, with column-level access control synchronization. MongoDB and SQL Server support are planned for Q3 2024.
4. How does self-hosting work? Deployment requires Docker 20.10+ and a server with 4GB RAM minimum. The setup process auto-configures TLS certificates, generates admin credentials, and provides Helm charts for Kubernetes orchestration.
5. Is GDPR compliance guaranteed? All data processing occurs within EU-based infrastructure (OVHcloud/AWS Paris), with optional Schrems II-compliant data anonymization modules for transatlantic teams. Audit trails meet CNIL's 7-year retention requirements.