Show HN: DeepSQL – A self-hostable DBA agent for Postgres and MySQL
DeepSQL launches as an AI-powered, self-hostable DBA agent for Postgres and MySQL, promising significant cost cuts and query optimization. While the promise of an autonomous database "brain" excites many, questions quickly arise regarding its closed-source nature and the true extent of its "self-hostable" claims, especially concerning LLM usage. The discussion highlights the tension between AI's potential for automation and concerns over data privacy and vendor lock-in.
The Lowdown
DeepSQL, developed by a former Oracle Query Engine team member, presents itself as an AI agent designed to act as a 24/7 DBA. Originating as an internal tool, it aims to prevent databases from becoming bottlenecks by leveraging AI to optimize performance and reduce costs. The product targets common database pain points, positioning itself as an intelligent, autonomous solution.
- AI-Driven Optimization: DeepSQL learns database schema, relationships, rules, and query logs to function as an autonomous agent, proactively identifying and solving performance issues.
- Cost Reduction: Claims to fix slow queries and database bloat, leading to substantial cuts in database spend. The creator cited personal experience reducing costs by over 3x.
- BI Dashboards: DeepSQL can generate BI dashboards, potentially replacing dedicated tools like Tableau, Retool, and Appsmith, further contributing to cost savings.
- Security Features: Offers PII redaction and role-based access to allow wider organizational interaction with data while maintaining security.
- Self-Hostable & Private: Designed for self-hosting within a user's VPC, ensuring data remains on-premises and read-only credentials are used.
- Integrated Workflow: Provides CLI, MCP, Web UI, and Slack integrations, along with a daily database health report.
In essence, DeepSQL seeks to give existing DBAs 'superpowers' or enable senior engineers to manage complex database tasks autonomously, focusing on efficiency and cost-effectiveness through continuous AI-driven monitoring and optimization.
The Gossip
Self-Hostable & Secretive Software
A primary concern among commenters was the product's 'self-hostable' claim juxtaposed with its closed-source nature and lack of clear pricing. Users noted the absence of a GitHub link and discovered that the `install.sh` script bundles DeepSQL's managed Azure OpenAI key, meaning user data and queries are processed through their shared LLM resource, despite assurances of data staying on-premises. The founder clarified that it's not open-source 'yet' but will likely move in that direction, and that 'Bring Your Own Key' (BYOK) will be an option for enterprise users, while acknowledging the current setup for consistency.
Skeptical Savings Statements
The claimed '40% cut in database costs' drew significant scrutiny. Commenters found the exact percentage arbitrary and questioned its applicability across different user scenarios. The founder elaborated on their own 3x savings and savings from other YC companies, attributing them to fixing issues like over-indexing, schema bloat, and replacing BI tools. However, critics pointed out that including savings from non-DB tools like Retool/Appsmith was misleading and that such high savings often imply existing 'horrendously bad' database practices, which might not resonate with more optimized setups. The founder ultimately agreed and removed the specific metric from the landing page.
BI Beyond Belief & Benchmarking
Users sought clarification on DeepSQL's BI dashboard capabilities, asking if they were for business analytics or DB ops, and how they are generated. The founder confirmed DeepSQL builds and shares BI dashboards from connected data sources. Discussion also revolved around how DeepSQL compares to existing DBA tools for query optimization, with some suggesting many tools already address slow queries based on rules and statistics. The founder emphasized DeepSQL's autonomous, context-aware approach, contrasting it with tools that analyze workloads in isolation and require expert DBAs, highlighting its continuous evaluation and holistic decision-making based on deployment type and data characteristics.