umputun / tg-spam / 29611674358

84%

Build Type

push

github

Committed by web-flow

Commit Message

feat: add optional --meta.image-text-len for image-only check (#430)

the image-only meta check reused `--min-msg-len` as its caption-length threshold, coupling two unrelated knobs. A short caption exactly at the `--min-msg-len` boundary slips through, because the check compares with strict `<`. Seen in production: an image ad with a 15-rune caption against `MIN_MSG_LEN=15` passed every content check, since the spam pitch was baked into the image pixels and only the short caption is visible to the classifier and LLM.

this adds `--meta.image-text-len` / `META_IMAGE_TEXT_LEN` (config `Meta.ImageTextLen`) to set a separate minimum caption length for image messages. `0` (the default) falls back to `--min-msg-len`, so existing behavior is unchanged unless you set it.

wired through every config layer: CLI/env, the CLI-to-Settings mapping, DB-backed config (with a `zeroAwarePaths` entry so an explicit `0` survives merges), and the web settings UI (form input plus read-only row, `min="0"` and a `Validate` bound that rejects negatives). Covered by unit tests for the mapping, the web form parse, and the `makeDetector` override-vs-fallback selection, plus an e2e UI round-trip, and documented in both README sections.

Coverage Stats

13 of 15 new or added lines in 4 files covered. (86.67%)

1 existing line in 1 file now uncovered.

8921 of 10653 relevant lines covered (83.74%)

254.44 hits per line