rust(feat): expose measurement received time in MCP get_data

rust/crates/sift_cli/CHANGELOG.md

@@ -7,6 +7,10 @@ This project adheres to [Semantic Versioning](http://semver.org/).
 
 ### What's New
 
+- `get_data` accepts `include_received_at` (alias `includeReceivedAt`). When
+  set, and the data service returns `sift_received_at`, the Parquet file
+  includes a receipt-time column per channel in unix nanos next to the
+  generated `timestamp_unix_nanos`. `upload_dataset` skips those columns.
 - `create_artifact` and the Sift skill now say which link relation to use:
   `source` for the run or asset an artifact was built from, and `derived_from`
   only for another artifact. Agents no longer send `derived_from` for a run and

rust/crates/sift_cli/assets/skills/sift/SKILL.md

@@ -109,6 +109,13 @@ tool.
   to enable them, then restart the MCP client.
 - **Data:** `get_data` writes channel data to a Parquet file. `sql` queries
   Parquet files. `upload_dataset` streams a Parquet dataset into Sift.
+- **Compare generation time with receipt time.** Call `get_data` with
+  `include_received_at: true` and `sample_ms: 0`. When Sift has a receipt time,
+  the file includes an Int64 column whose name contains `received_at="true"`,
+  in unix nanos, aligned with that channel's samples. `timestamp_unix_nanos` is
+  still when the measurement was generated. The receipt column is absent when
+  the service has no receipt time, including enum and bit-field channels.
+  `upload_dataset` skips receipt columns.
 - **Links:** `explore_url`.
 - **Docs:** `search_docs`.
 - **Writes:** `create_rule`, `update_rule`, `archive_rule`, `unarchive_rule`,

rust/crates/sift_cli/assets/skills/sift/references/cli.md

@@ -129,7 +129,9 @@ If the CLI does not support the user's file type:
 Do not reach for the MCP `upload_dataset` tool to import a user's file. It
 accepts only Sift's canonical Parquet schema — every column named
 `<channel_name> {channel_id="<uuid>"}` — so it round-trips the output of
-`get_data`, and rejects a foreign file with `missing attribute block`.
+`get_data`, and rejects a foreign file with `missing attribute block`. A
+`received_at="true"` column from `get_data` is a receipt time, not a channel,
+and upload skips it.
 
 ### Stream data
 

rust/crates/sift_mcp/src/service/common/mod.rs

@@ -74,16 +74,18 @@ pub fn secs_and_subsec_nanos_to_unix_nanos(sec: i64, subsec_nanos: i32) -> i64 {
 
 /// A fully-qualified Arrow column name with the channel-specific fields the
 /// data pipeline cares about. `Display` emits the canonical Parquet column-name
-/// string: `<name> {channel_id="...", bit_field_element="...", run="...", units="..."}`.
-/// Empty optional fields are omitted. Construct via [`ColumnName::builder`] or
-/// `TryFrom<String>` (when re-hydrating a string from a Parquet schema).
+/// string: `<name> {channel_id="...", bit_field_element="...", run="...", units="...", received_at="true"}`.
+/// Empty optional fields are omitted. `received_at="true"` marks a receipt-time
+/// column rather than the channel's samples. Construct via [`ColumnName::builder`] or
+/// `TryFrom<&str>` (when re-hydrating a string from a Parquet schema).
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
 pub struct ColumnName {
     name: String,
     channel_id: String,
     bit_field_element: Option<String>,
     run: Option<String>,
     units: Option<String>,
+    received_at: bool,
 }
 
 impl ColumnName {
@@ -94,6 +96,20 @@ impl ColumnName {
             bit_field_element: None,
             run: None,
             units: None,
+            received_at: false,
+        }
+    }
+
+    /// The same channel identity, marked as that channel's receipt-time column.
+    /// Units are dropped: the column holds unix nanos, not the measurement unit.
+    pub fn with_received_at(&self) -> Self {
+        Self {
+            name: self.name.clone(),
+            channel_id: self.channel_id.clone(),
+            bit_field_element: self.bit_field_element.clone(),
+            run: self.run.clone(),
+            units: None,
+            received_at: true,
         }
     }
 
@@ -116,6 +132,10 @@ impl ColumnName {
     pub fn units(&self) -> Option<&str> {
         self.units.as_deref()
     }
+
+    pub fn received_at(&self) -> bool {
+        self.received_at
+    }
 }
 
 impl fmt::Display for ColumnName {
@@ -130,6 +150,9 @@ impl fmt::Display for ColumnName {
         if let Some(v) = &self.units {
             write!(f, ", units=\"{v}\"")?;
         }
+        if self.received_at {
+            write!(f, ", received_at=\"true\"")?;
+        }
         write!(f, "}}")
     }
 }
@@ -157,6 +180,7 @@ impl TryFrom<&str> for ColumnName {
         let mut bit_field_element: Option<String> = None;
         let mut run: Option<String> = None;
         let mut units: Option<String> = None;
+        let mut received_at = false;
 
         for segment in attrs.split(',') {
             let segment = segment.trim();
@@ -183,6 +207,12 @@ impl TryFrom<&str> for ColumnName {
                 "bit_field_element" => bit_field_element = Some(val),
                 "run" => run = Some(val),
                 "units" => units = Some(val),
+                "received_at" => {
+                    if val != "true" {
+                        bail!("`received_at` must be \"true\"");
+                    }
+                    received_at = true;
+                }
                 other => bail!("unknown attribute key `{other}`"),
             }
         }
@@ -196,6 +226,7 @@ impl TryFrom<&str> for ColumnName {
             bit_field_element,
             run,
             units,
+            received_at,
         })
     }
 }
@@ -210,6 +241,7 @@ pub struct ColumnNameBuilder<'a> {
     bit_field_element: Option<String>,
     run: Option<String>,
     units: Option<String>,
+    received_at: bool,
 }
 
 impl<'a> ColumnNameBuilder<'a> {
@@ -228,13 +260,19 @@ impl<'a> ColumnNameBuilder<'a> {
         self
     }
 
+    pub fn received_at(mut self, value: bool) -> Self {
+        self.received_at = value;
+        self
+    }
+
     pub fn build(self) -> ColumnName {
         ColumnName {
             name: self.name.to_string(),
             channel_id: self.channel_id.to_string(),
             bit_field_element: self.bit_field_element,
             run: self.run,
             units: self.units,
+            received_at: self.received_at,
         }
     }
 }

rust/crates/sift_mcp/src/service/common/test.rs

@@ -137,6 +137,33 @@ fn column_name_try_from_missing_channel_id_errors() {
     );
 }
 
+#[test]
+fn column_name_received_at_round_trips_and_drops_units() {
+    let original = ColumnName::builder("temp", "c1")
+        .run(Some("r42"))
+        .units(Some("C"))
+        .build()
+        .with_received_at();
+    assert!(original.received_at());
+    assert_eq!(original.units(), None);
+    assert_eq!(
+        original.to_string(),
+        "temp {channel_id=\"c1\", run=\"r42\", received_at=\"true\"}"
+    );
+    let parsed = ColumnName::try_from(original.to_string().as_str()).expect("should parse");
+    assert_eq!(parsed, original);
+}
+
+#[test]
+fn column_name_try_from_rejects_received_at_other_than_true() {
+    let err = ColumnName::try_from("temp {channel_id=\"c1\", received_at=\"false\"}")
+        .expect_err("only true is a receipt column");
+    assert!(
+        err.to_string().contains("`received_at` must be \"true\""),
+        "unexpected error: {err}"
+    );
+}
+
 #[test]
 fn column_name_try_from_unknown_key_errors() {
     let err = ColumnName::try_from("temp {channel_id=\"c1\", flavor=\"strawberry\"}")