internal/extract
github.com/ensera-ai/taisce/internal/extract · 2 files · 383 lines · 25 tests · source
Package extract turns a message into claims that can be cited.
The model is a port. What is in this package is everything that happens to a model's output before it is allowed to become a claim, and that is the part with a defensible correctness argument: a model asked for a quote will paraphrase it, normalise its whitespace, or invent it outright, and every one of those produces a citation that resolves to the wrong words while looking correct.
Where it sits
Imports: internal/domain
Imported by: cmd/taisce, internal/formation, internal/infra/inference
Files
| File | Lines | Declarations | What it is for |
|---|---|---|---|
| extract.go | 336 | 11 | Carries the package documentation. |
| identity.go | 47 | 3 | Hash the compiled admission implementation as well as configuration. |
25 tests hold this package's behaviour; the list is named for what each one proves.