iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Building a database tool for multiple clouds took Abd Alrhman Alloush longer than he expected—not because the connectors were the hardest code, but because identity, permissions, schema discovery, and customer expectations were part of the product too. His retrospective on building Query1AI is a useful reminder that a working connection is only one piece of a multi-cloud tool.
What took longer than the connector code
Alloush expected connector work to take “a few weeks.” In his account, it took “a full quarter longer” than he had estimated. Those are his own estimates, not a general benchmark for cloud integrations. He identifies the surrounding work—credential setup, access scope, documentation, and user education—as the part he had underestimated. He summed up the mismatch this way: “The connectors themselves were the easy part.”
The product had to account for different identity models: AWS IAM roles and access keys, Google Cloud service accounts and permissions scoped per project, and MongoDB Atlas API keys and project scoping. Explaining how to grant access appropriately, and helping users understand those choices, took time beyond implementing connector code. These are examples from Alloush’s build, not current provider-specific setup instructions; consult each provider’s documentation for present-day configuration details.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteFor a multi-cloud product, onboarding is an integration surface in its own right. A connector can work in a developer’s environment while the real user still cannot confidently grant it access. Planning for that gap means budgeting for clear permission guidance and testing whether the instructions make sense to the people responsible for the accounts.
#1 Best Overall
Why schema inventory had to come before natural-language queries
Query1AI aimed to let people ask database questions in plain English. Alloush says he initially treated natural-language querying as a feature to build, then recognized that it needed reliable information about the underlying data first: table names, column names, and relationships.
Without that inventory, a language model has too little grounding to turn a question into a useful database query. A question may sound clear to a person but still be ambiguous without knowing which tables contain the relevant data and how they connect. Alloush reordered the roadmap around inventory and schema discovery, describing the dependency plainly: “Without the inventory and schema-discovery layer built and working first, there’s nothing for the natural-language layer to ground itself in.”
Rank #2
The practical lesson is to map the data before promising that users can ask arbitrary questions. Schema discovery is not merely preparatory plumbing; it supplies context the query experience depends on. For teams evaluating a similar product, useful questions include:
- Does it discover the tables, columns, and relationships needed for the intended questions?
- Can users inspect or correct that metadata when the discovered structure is incomplete or misleading?
- Are the questions people want to ask answerable from the available schema and permissions?
How read-only and agentless became a positioning challenge
Alloush describes Query1AI as read-only and API-based, with no agent installed on database nodes. He found that this constraint made feature comparisons with agent-based alternatives difficult. His response was to explain the operational trade-off instead of trying to win on feature checklists: customers would have nothing to deploy or patch on their database infrastructure.
Rank #3
That is a positioning lesson from his experience, not proof that an agentless design is inherently safer or preferable in every environment. The useful question is what the constraint means for a particular team. An agentless approach may reduce infrastructure components the customer has to install and maintain; other designs may offer capabilities that matter more for a given use case. The choice should be evaluated against deployment requirements, permissions, operational ownership, and the work the tool must perform.
Alloush reflected: “The lesson was to stop trying to win that feature comparison and instead lead with what the constraint actually buys: nothing to deploy, nothing to patch, nothing running on infrastructure we don’t own the risk for.” The wording expresses his product’s rationale, rather than a universal security conclusion.
Why it matters to say what is not ready
Alloush says Azure interest arrived before support was ready. Rather than imply the provider was already supported, he chose to describe it as in development. That status refers to the time of his article; it does not establish Query1AI’s current Azure availability.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →The broader product lesson is to distinguish between supported, being developed, and merely requested. Those labels help prospective users decide whether a tool fits their environment without relying on an implied promise. For builders, demand is useful evidence about priorities, but it is not evidence that an integration is complete.
Best Value
Who the product is for can change the roadmap
Alloush initially focused on DBAs and platform engineers. He later recognized that product managers and support leads could also benefit from asking questions in plain English rather than waiting for a technical intermediary. As he framed the discovery: “can a PM or a support lead just ask this in English and get a real answer”.
That shift affects more than marketing language. A tool designed only around technical operators may assume users understand schemas, query terminology, and permission boundaries. Serving non-technical requesters raises questions about how answers are grounded, what they are allowed to see, and how uncertainty is communicated. Identifying those users early can change which workflows deserve priority.
The central lesson from Alloush’s retrospective
Alloush’s account is a first-person product-building retrospective, not an independently verified case study or a universal estimate for multi-cloud projects. Its most transferable point is that the difficult decisions were not necessarily the visible feature decisions: they were the supporting systems and expectations around them. Identity and onboarding shaped whether connections could be used; schema inventory shaped whether natural-language questions could be grounded; and honest readiness labels shaped whether users could trust what was promised.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Looking back, Alloush wrote: “The common thread is that the actual engineering judgment calls — read-only by default, no agents, say what’s not ready yet — all held up fine.” The distinction he draws is between those design judgments and the underestimated work of making them understandable and useful to customers. Read Alloush’s retrospective on DEV Community.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
