Verify extraction provenance and prune semantic staging before inference

crates/searchi-core/Cargo.toml

@@ -7,6 +7,7 @@ license.workspace = true
 [dependencies]
 anyhow.workspace = true
 serde.workspace = true
+sha2 = "0.10"
 rusqlite = { version = "0.32", features = ["bundled"] }
 walkdir = "2"
 lopdf = { version = "0.45", default-features = false }