labring/fastgpt: technical due diligence, read from its repository

The median source file here is 95 lines. 21 files exceed 1,000 lines, the largest being 1,761. Size is not a defect on its own — some problems genuinely live in one place — but these are where merge conflicts, review fatigue and single-owner knowledge concentrate, and they are the first thing that slows an inheriting engineer down.

Of five ordinary markers — CI workflows, CODEOWNERS, a pull-request template, pre-commit hooks, contribution guidance — 1 is present: CI workflows. Branch protection itself is a GitHub setting rather than a file, so this is a reading of proxies and not of the rules themselves. What it does establish is that the process is not written down anywhere a new engineer would find it.

The code is pinned to Go 1.22 (end of life 2025-02-11). None of these receives security fixes any more; the first stopped on 2025-02-11. Moving off an unsupported runtime is rarely a one-line change, so a buyer should ask whether the upgrade is planned and what it touches.

8 third-party CI actions are referenced by a tag or branch rather than a commit, so they run whatever that tag points to on the day; 3 steps pipe a download straight into a shell. Tags have been moved to malicious commits in real supply-chain attacks, and a piped script is whatever the server returns that minute. Pinning to a commit is the fix, and it is small.

2 places in product code build SQL by formatting or joining strings rather than passing parameters. In each, what the code shows going in is a table or column name, a constant, or something the line does not show, not a request's value. Ask whether any value reaches these strings, rather than a parameter.

5 places set HTML from a value that is neither a literal nor passed through a sanitiser. Where the value can come from a user, that is stored cross-site scripting, and it is the first thing a buyer's security review looks for in a frontend.

Third-party code is committed into this repository without any statement of the terms it arrived under. Unknown terms are harder to deal with than inconvenient ones: an awkward licence can be complied with, an absent one cannot. Each of these needs its origin established before the code changes hands.

LICENSE grants Apache-2.0 with terms on offering it as a service: open source; offering it over a network counts as distributing it, so a hosted product's changes must be published. The reading comes from the licence's own title and clauses. Whether the terms fit the buyer's intended use, and whether any paid licence transfers with the code, is a question for counsel and the licensor.

The root licence changed 1 time(s). The latest terms restrict what users, hosts or a buyer may do in ways the earlier ones did not. Code taken under the old terms stays under them, so both matter in diligence.

The newest was published on 2026-10-12. A project that publishes advisories is telling its users about its fixes, and the count and pace say how much security work it carries. Whether 6 of them are fixed in the code you ship could not be told from this history: its version is only known to be at least 4.17.2. Answer lives in: the release you run, set against each fixed version.

1 images in the deployment configuration are named without a version or as `latest`. A redeploy can change the software without a commit, a rollback has nothing to roll back to, and nobody can say from the repository what is running.

1 deployments set `replicas: 1`, and nothing (an autoscaler, a disruption budget) adds more. Every deploy and every node drain is downtime, and the availability the buyer inherits has a ceiling.

13 places outside configuration test whether they are running in production or staging. Staging then does not exercise production's code, and each deploy is the first run of those branches. Differences belong in configuration the code reads, not in the code.