diff --git a/DESCRIPTION b/DESCRIPTION
index f07c870..dcf1723 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -1,6 +1,6 @@
 Package: stt.api
 Title: 'OpenAI' Compatible Speech-to-Text API Client
-Version: 0.2.1
+Version: 0.2.1.1
 Authors@R: c(
     person("Troy", "Hernandez", email = "troy@cornball.ai", role = c("aut", "cre"),
            comment = c(ORCID = "0009-0005-4248-604X")),
diff --git a/R/stt.R b/R/stt.R
index c4928f5..ebee9b8 100644
--- a/R/stt.R
+++ b/R/stt.R
@@ -27,6 +27,10 @@
 #'   \item{backend}{Which backend was used ("api" or "whisper").}
 #'   \item{raw}{The raw response from the backend.}
 #' }
+#' The result also carries a \code{"call_record"} attribute (cornball_sidecar
+#' v1, as in xtx.api/tts.api): the resolved request, elapsed seconds, and a
+#' timestamp -- provenance that rides with the transcription when callers
+#' serialize it.
 #'
 #' @examples
 #' \dontrun{
@@ -57,7 +61,8 @@ stt <- function(file, model = NULL, language = NULL,
     resolved_backend <- .choose_backend(backend)
 
     # Dispatch to appropriate backend
-    if (resolved_backend == "openai") {
+    started <- Sys.time()
+    res <- if (resolved_backend == "openai") {
         .via_api(
                  file = file,
                  model = model,
@@ -68,5 +73,21 @@ stt <- function(file, model = NULL, language = NULL,
     } else {
         .via_whisper(file = file, model = model, language = language)
     }
+    # stt produces an R object, not a media file, so the call record rides as
+    # an attribute (cornball_sidecar v1, as in xtx.api/tts.api); callers that
+    # serialize the result keep its provenance with it.
+    attr(res, "call_record") <- list(
+        cornball_sidecar = 1L, package = "stt.api",
+        version = as.character(utils::packageVersion("stt.api")),
+        fn = "stt",
+        request = Filter(Negate(is.null),
+                         list(file = file, model = model,
+                              language = language,
+                              response_format = response_format,
+                              backend = resolved_backend, prompt = prompt)),
+        elapsed = round(as.numeric(difftime(Sys.time(), started,
+            units = "secs")), 2),
+        created = format(Sys.time(), "%Y-%m-%dT%H:%M:%S%z"))
+    res
 }
 
diff --git a/man/stt.Rd b/man/stt.Rd
index 2f3022e..d89cfd9 100644
--- a/man/stt.Rd
+++ b/man/stt.Rd
@@ -38,6 +38,10 @@ A list with components:
   \item{backend}{Which backend was used ("api" or "whisper").}
   \item{raw}{The raw response from the backend.}
 }
+The result also carries a \code{"call_record"} attribute (cornball_sidecar
+v1, as in xtx.api/tts.api): the resolved request, elapsed seconds, and a
+timestamp -- provenance that rides with the transcription when callers
+serialize it.
 }
 \description{
 Convert an audio file to text using a local whisper backend or