A successful build leaves the database checks to you
A diagram can help explain a schema, but our drawDB build did not establish that exported SQL recreates it correctly. That distinction matters when choosing an editor for migration work.
The drawDB README describes a browser ERD editor that can build diagrams, import and export SQL scripts, and generate migrations without creating an account. Those are documented capabilities. Our recorded run covered dependency installation and a production build, rather than the browser workflow.
For developers considering drawDB, the useful next step is to import a representative schema, inspect the diagram, export the SQL and apply it to a disposable database. Compare the resulting database metadata with the intended schema before using the output in a migration. This is a recommended acceptance check, not work completed in our run.
What our October 4 run established
Our stored run began on October 4, 2026 at 03:39:08.843 UTC and used commit ee550b4117a6e1b5366205faa5cee713ee1d4e3b, abbreviated ee550b4. Dependency installation succeeded in 21,960 ms, rounded to 22 seconds. The build command succeeded in 16,806 ms, rounded to 17 seconds. Inside that command, Vite v8.1.0 reported its own build duration of 11.59 seconds. These timings describe different scopes.
The recorded configuration was 3 CPUs, 8 GB of memory and image lab-node:22, with Node major 22 recorded by detection. We have no exact runtime patch version in this evidence. One successful build does not establish compatibility across Node versions or deployment environments.
The dependency summary recorded count 1,890 and installation size 618 MB. Treat that count as a stored dependency metric, not a count of packages downloaded or audited. The repository summary recorded 295 files, about 52,534 source lines and a 3.6 MB checkout.
npm audit recorded five dependency advisories: three high, one moderate, one low and zero critical. Audit coverage is unknown. These findings do not establish exploitability in the deployed editor, and zero critical findings do not establish safety. Review the affected dependency paths against your release policy.
The build emitted a direct-eval warning involving lottie-web and a chunk-size warning. They did not fail the build and do not independently establish a security defect or measured browser slowdown.
The harness skipped tests because it found no test script. Detection listed dev, build, lint and preview scripts, no tests directory, and two CI workflows. That is narrower than proving that no tests exist. We did not test browser controls, diagram persistence, SQL imports and exports, sharing, Docker deployment or an installed binary.
Sharing adds a server
The README documents local npm development, a production build and Docker commands. It says the server is optional unless you need to share files. For Docker, it documents passing VITE_BACKEND_URL when starting the container without rebuilding the image.
Evaluate that server configuration alongside the frontend if shared files are part of your intended workflow. Our frontend build provides no evidence about the sharing service's operation or reliability.
PostgreSQL defaults need a focused acceptance check
Issue 1144 contains a report from a user running the v1.8.1 Docker image in Firefox on Linux. The report describes warnings for defaults on TEXT and VARCHAR without a length, plus difficulty representing an empty-string default. These are reported observations, not results from our sandbox.
A separate issue commenter describes checks on main at f418c29 using PostgreSQL 16.11. The commenter explicitly says the reported v1.8.1/Firefox environment remained untested. The supplied thread also records a referencing commit, 6bd5af6, on October 21, 2026. That reference alone does not establish that a fix was merged, released or resolves every default-handling case.
For your own acceptance check, distinguish an absent default, SQL NULL, an empty string, literal quotes and whitespace. Include the types and constraints your application actually uses. Our October 4 build cannot answer whether those cases work in your chosen version.
Choose alternatives by the starting task
ChartDB's README describes running a query to retrieve schema JSON and importing that result into its diagram editor. Evaluate it when an existing database is your starting point. The project describes itself as Public Beta; this review did not test its import or export behavior.
Azimutt's README describes exploring schemas through search, displaying selected tables and columns, following relationships, and adding documentation. Evaluate it when navigating and documenting an existing schema matters more than starting a new diagram. Its description does not establish a speed advantage over drawDB.
SchemaSpy is omitted from the recommendations because no primary-source extract supporting its role was supplied.
How to read the scores
These are editorial evidence ratings, not benchmark results or an overall product ranking. The scale is 1 for substantial demonstrated obstacles, 2 for material limitations, 3 for a documented or observed baseline with important gaps, 4 for strong validation across the relevant workflow, and 5 for extensive corroborated validation. Missing evidence limits a rating; it does not prove poor product quality.
Setup receives 3/5 because installation and building succeeded, while deployment and sharing remain untested. Documentation receives 3/5 because the supplied README gives setup commands and the sharing configuration boundary, but does not establish SQL round-trip correctness. Community receives 3/5 because the supplied issue discussions show maintainer and contributor participation, without establishing response-time or support guarantees. Maturity receives 3/5 because a successful build coexists with unresolved validation questions and reported behavior that needs version-specific checking.
The recorded license is AGPL-3.0. Check its terms against your deployment policy rather than treating all network use as prohibited. If the license fits, start with one representative schema and verify the exported result before expanding adoption.