review classifier: the certainty column's header is just certainty, so the column stays narrow

The review body's table is now: result | check | certainty / threshold |
reason. The result cell is a symbol for what the row did to the review,
after polarity (✅ counted and passed, ❌ counted and failed, ❔ not
counted or no answer), then the model's verdict. The overall line carries
✅ / ❌. Polarity, threshold and counted are folded into those cells.

The prompt asks for a reason under 350 characters; the enforced cap
stays 500, so an overshoot is truncated less often.

docs/swarm/review-classifier.md

@@ -300,8 +300,8 @@ The model answers each classification in JSON. The answer holds:
 
 - a **verdict**, one of _match_, _no match_ or _not applicable to this PR_
 - a **certainty** from 0 to 100, how sure the model is of that verdict
-- a **reason**; the classifier truncates one over 500 characters to 499 plus
-  `…`, and keeps the verdict and certainty
+- a **reason**; the prompt asks for under 350 characters, and the classifier
+  truncates one over 500 to 499 plus `…`, keeping the verdict and certainty
 
 An answer:
 
@@ -339,14 +339,20 @@ all.
 The service account posts the verdict as a forge review, approve or request
 changes, and the table is the review's body. An illustrative table:
 
-| classification | polarity | verdict        | certainty | threshold | counted | reason                                       |
-| -------------- | -------- | -------------- | --------- | --------- | ------- | -------------------------------------------- |
-| datetime       | positive | no match       | 92        | 80        | yes ❌  | Two new timestamps use naive datetimes.      |
-| no-ads         | negative | no match       | 99        | 90        | yes ✅  | No commit message carries an AI attribution. |
-| bao-first      | positive | not applicable | 95        | 70        | no      | The PR adds no secret.                       |
-| docs-updated   | positive | match          | 41        | 60        | no      | Below threshold.                             |
+| result      | check        | certainty | reason                                       |
+| ----------- | ------------ | --------- | -------------------------------------------- |
+| ❌ no match | datetime     | 92 / 80   | Two new timestamps use naive datetimes.      |
+| ✅ no match | no-ads       | 99 / 90   | No commit message carries an AI attribution. |
+| ❔ unsure   | bao-first    | 95 / 70   | The PR adds no secret.                       |
+| ❔ match    | docs-updated | 41 / 60   | Below threshold.                             |
 
-**Overall: request changes.**
+**Overall: ❌ request changes**
+
+The result cell is the symbol for what the row did to the review, then the
+model's verdict. ✅ is a counted row that passed its polarity and ❌ a counted
+row that failed it, so _match_ is ✅ on a positive rule and ❌ on a negative one.
+❔ is a row that doesn't count: below its threshold, not applicable (shown as
+_unsure_), or without an answer. The certainty cell reads _certainty / threshold_.
 
 The approval is a real one: it counts toward a repo's required approvals
 like any other ([`pr-review-gate.md`](../process/pr-review-gate.md)). The

docs/swarm/ui.md

@@ -115,9 +115,9 @@ doesn't parse, or it holds more than ten rules, the page shows the reason and
 leaves your classifications as they were.
 
 A run takes a minute or more, and up to half an hour for ten classifications.
-The page says it's running and shows how long it has been, then shows the review's table: for each classification the verdict,
-the model's certainty, the threshold, whether the verdict counted and the
-reason. A classification that couldn't run shows its error where the reason
+The page says it's running and shows how long it has been, then lists each classification with its verdict, the model's certainty, the
+threshold, whether the verdict counted and the reason; the posted review folds
+those fields into four columns ([`review-classifier.md`](review-classifier.md)). A classification that couldn't run shows its error where the reason
 would be. Under the table are the overall verdict, approve or request changes,
 and the head commit the run read. A failed run shows its error, for example when the swarm
 has no API key yet.

frontend/packages/swarm-ui/src/pages/review-tryout/TryOutResult.tsx

@@ -1,6 +1,7 @@
 // <TryOutResult> — a finished try-out: one row per classification, the
-// overall verdict under them, and the head commit the run read. The columns
-// are the review's own table (`docs/swarm/review-classifier.md`); no token
+// overall verdict under them, and the head commit the run read. One column per
+// field of a row, where the posted review folds them into four
+// (`docs/swarm/review-classifier.md`); no token
 // counts, cache counts or cost, here or anywhere in the UI.
 import { Badge } from "@hive/shared/badge.js";
 import { Table, type TableColumn } from "../../ui/table/Table.js";

swarm-controller/src/review_classifier/review.rs

@@ -681,13 +681,13 @@ fn review_for(outcome: &Outcome) -> Ruling {
                 );
                 Ruling::no_verdict(reason, rows)
             } else {
-                let (event, said) = match overall {
-                    Overall::Approve => (Event::Approve, "approve"),
-                    Overall::RequestChanges => (Event::RequestChanges, "request changes"),
+                let event = match overall {
+                    Overall::Approve => Event::Approve,
+                    Overall::RequestChanges => Event::RequestChanges,
                 };
                 Ruling {
                     event,
-                    body: format!("{}\n**Overall: {said}.**\n", table(rows)),
+                    body: format!("{}\n{}", table(rows), overall_line(*overall)),
                     reason: None,
                 }
             }
@@ -711,32 +711,18 @@ fn no_verdict(reason: &str, rows: &[Row]) -> String {
 /// The result table, one row per classification.
 fn table(rows: &[Row]) -> String {
     let mut out = String::from(
-        "| classification | polarity | verdict | certainty | threshold | counted | reason |\n\
-         | --- | --- | --- | --- | --- | --- | --- |\n",
+        "| result | check | certainty | reason |\n\
+         | --- | --- | --- | --- |\n",
     );
     for r in rows {
-        let polarity = match r.polarity {
-            swarm_queue_client::review_classifier::Polarity::Positive => "positive",
-            swarm_queue_client::review_classifier::Polarity::Negative => "negative",
-        };
-        let (verdict, certainty, counted, reason) = if let Some(error) = &r.error {
-            ("no answer", String::new(), "no", error.as_str())
-        } else {
-            let verdict = match r.verdict {
-                Verdict::Match => "match",
-                Verdict::NoMatch => "no match",
-                Verdict::NotApplicable => "not applicable",
-            };
-            let counted = match (r.counted, r.passed) {
-                (false, _) => "no",
-                (true, true) => "yes ✅",
-                (true, false) => "yes ❌",
-            };
-            (verdict, r.certainty.to_string(), counted, r.reason.as_str())
+        let (certainty, reason) = match &r.error {
+            Some(error) => ("—".to_owned(), error.as_str()),
+            None => (r.certainty.to_string(), r.reason.as_str()),
         };
         let _ = writeln!(
             out,
-            "| {} | {polarity} | {verdict} | {certainty} | {} | {counted} | {} |",
+            "| {} | {} | {certainty} / {} | {} |",
+            result(r),
             cell(&r.name),
             r.threshold,
             cell(reason)
@@ -745,6 +731,32 @@ fn table(rows: &[Row]) -> String {
     out
 }
 
+/// A row's result cell: the symbol for what the row did to the review, then
+/// the model's verdict. ✅ is a counted row that passed its polarity, ❌ a
+/// counted row that failed it, ❔ a row that doesn't count: below its
+/// threshold, not applicable, or without an answer. So `match` is ✅ under a
+/// positive rule and ❌ under a negative one.
+fn result(r: &Row) -> &'static str {
+    match (r.error.is_some(), r.counted, r.passed, r.verdict) {
+        (true, ..) => "❔ no answer",
+        (_, true, true, Verdict::Match) => "✅ match",
+        (_, true, true, Verdict::NoMatch) => "✅ no match",
+        (_, true, false, Verdict::Match) => "❌ match",
+        (_, true, false, Verdict::NoMatch) => "❌ no match",
+        (_, _, _, Verdict::Match) => "❔ match",
+        (_, _, _, Verdict::NoMatch) => "❔ no match",
+        (_, _, _, Verdict::NotApplicable) => "❔ unsure",
+    }
+}
+
+/// The closing line of a review: the overall verdict, with its symbol.
+fn overall_line(overall: Overall) -> &'static str {
+    match overall {
+        Overall::Approve => "**Overall: ✅ approve**\n",
+        Overall::RequestChanges => "**Overall: ❌ request changes**\n",
+    }
+}
+
 /// `text` as one table cell's inline text: on one line, with every character
 /// that could end the cell, start markup, math (`$`) or raw HTML escaped, and a zero-width
 /// space after every `@`, `#`, `!`, `://` and `www.`, so it carries no live
@@ -834,6 +846,97 @@ mod tests {
         }
     }
 
+    /// A row as the aggregation produces it for `polarity`, `verdict`,
+    /// `certainty` against `threshold`.
+    fn answered(polarity: Polarity, verdict: Verdict, certainty: u8, threshold: u8) -> Row {
+        let counted = verdict != Verdict::NotApplicable && certainty >= threshold;
+        let wanted = match polarity {
+            Polarity::Positive => Verdict::Match,
+            Polarity::Negative => Verdict::NoMatch,
+        };
+        Row {
+            name: "no-ads".to_owned(),
+            polarity,
+            verdict,
+            certainty,
+            threshold,
+            counted,
+            passed: counted && verdict == wanted,
+            reason: "Because.".to_owned(),
+            error: None,
+        }
+    }
+
+    #[test]
+    fn a_row_shows_whether_it_passed_not_the_raw_verdict() {
+        use Polarity::{Negative, Positive};
+        use Verdict::{Match, NoMatch, NotApplicable};
+        for (case, r, expected) in [
+            (
+                "positive match passes",
+                answered(Positive, Match, 85, 65),
+                "| ✅ match | no-ads | 85 / 65 | Because. |",
+            ),
+            (
+                "negative match fails",
+                answered(Negative, Match, 85, 65),
+                "| ❌ match | no-ads | 85 / 65 | Because. |",
+            ),
+            (
+                "negative no match passes",
+                answered(Negative, NoMatch, 97, 80),
+                "| ✅ no match | no-ads | 97 / 80 | Because. |",
+            ),
+            (
+                "positive no match fails",
+                answered(Positive, NoMatch, 97, 80),
+                "| ❌ no match | no-ads | 97 / 80 | Because. |",
+            ),
+            (
+                "not applicable is unsure",
+                answered(Negative, NotApplicable, 98, 70),
+                "| ❔ unsure | no-ads | 98 / 70 | Because. |",
+            ),
+            (
+                "below the threshold does not count",
+                answered(Positive, Match, 41, 60),
+                "| ❔ match | no-ads | 41 / 60 | Because. |",
+            ),
+            (
+                "no answer shows its error",
+                Row {
+                    error: Some("timed out".to_owned()),
+                    verdict: NotApplicable,
+                    certainty: 0,
+                    counted: false,
+                    passed: false,
+                    reason: String::new(),
+                    ..answered(Positive, Match, 0, 65)
+                },
+                "| ❔ no answer | no-ads | — / 65 | timed out |",
+            ),
+        ] {
+            let body = table(&[r]);
+            assert_eq!(
+                body,
+                format!(
+                    "| result | check | certainty | reason |\n\
+                     | --- | --- | --- | --- |\n{expected}\n"
+                ),
+                "{case}"
+            );
+        }
+    }
+
+    #[test]
+    fn the_overall_line_carries_its_emoji() {
+        assert_eq!(overall_line(Overall::Approve), "**Overall: ✅ approve**\n");
+        assert_eq!(
+            overall_line(Overall::RequestChanges),
+            "**Overall: ❌ request changes**\n"
+        );
+    }
+
     fn request(name: &str) -> Requested {
         Requested {
             owner: "o".to_owned(),
@@ -1437,7 +1540,7 @@ mod tests {
         assert_eq!(commit.as_deref(), Some("abc123"));
         assert_eq!(review.event, "APPROVED");
         assert!(
-            review.body.ends_with("**Overall: approve.**\n"),
+            review.body.ends_with("**Overall: ✅ approve**\n"),
             "{}",
             review.body
         );
@@ -1451,7 +1554,11 @@ mod tests {
         )
         .await;
         assert_eq!(review.event, "REQUEST_CHANGES");
-        assert!(review.body.contains("| yes ❌ |"), "{}", review.body);
+        assert!(
+            review.body.contains("| ❌ match | datetime | 90 / 80 |"),
+            "{}",
+            review.body
+        );
     }
 
     #[tokio::test]
@@ -1499,12 +1606,12 @@ mod tests {
         assert_eq!(body.lines().count(), 3, "{body}");
         assert!(!line.contains("@mara"), "{line}");
         assert!(line.contains("@\u{200b}mara"), "{line}");
-        // Seven cells: eight unescaped pipes.
+        // Four cells: five unescaped pipes.
         let unescaped = line
             .char_indices()
             .filter(|(i, c)| *c == '|' && (*i == 0 || line.as_bytes()[i - 1] != b'\\'))
             .count();
-        assert_eq!(unescaped, 8, "{line}");
+        assert_eq!(unescaped, 5, "{line}");
         assert!(
             line.contains("\\!\u{200b}\\[\\]") && line.contains("\\<b\\>"),
             "{line}"
@@ -1531,7 +1638,7 @@ mod tests {
             .char_indices()
             .filter(|(i, c)| *c == '|' && (*i == 0 || line.as_bytes()[i - 1] != b'\\'))
             .count();
-        assert_eq!(unescaped, 8, "{line}");
+        assert_eq!(unescaped, 5, "{line}");
     }
 
     #[test]

swarm-review-classifier/README.md

@@ -20,7 +20,7 @@ forge and holds no forge credential.
   (`--tools ""`), no MCP server, its own empty directory and
   `ANTHROPIC_API_KEY` in its environment only.
 - An answer is a JSON object with a verdict, a certainty and a reason; the
-  classifier truncates a reason over 500 characters to 499 plus `…`. A failed run, a
+  prompt asks for a reason under 350 characters, and the classifier truncates one over the 500 cap to 499 plus `…`. A failed run, a
   timeout or an answer that doesn't validate leaves that row without a verdict, and it doesn't count.
 - A run with no API key, or with no classification answered, fails as a
   whole: the caller sees an error, never a verdict.

swarm-review-classifier/src/answer.rs

@@ -12,6 +12,11 @@ use swarm_queue_client::review_classifier::Verdict;
 /// truncated to this, ending in `…`.
 pub const REASON_MAX_CHARS: usize = 500;
 
+/// The length the prompt asks a reason to stay under. Below `REASON_MAX_CHARS`
+/// so a model that overshoots the ask still lands inside the cap, uncut.
+pub const REASON_ASKED_CHARS: usize = 350;
+const _: () = assert!(REASON_ASKED_CHARS < REASON_MAX_CHARS);
+
 /// One validated answer.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct Answer {

swarm-review-classifier/src/prompt.rs

@@ -9,7 +9,7 @@ use std::fmt::Write as _;
 
 use swarm_queue_client::review_classifier::{Classification, Polarity, TryOut};
 
-use crate::answer::REASON_MAX_CHARS;
+use crate::answer::REASON_ASKED_CHARS;
 
 /// The system prompt, the same for every run.
 pub const SYSTEM: &str = "\
@@ -89,7 +89,7 @@ pub fn classification_prompt<'a>(shared: &'a str, c: &Classification) -> Prompt<
     };
     let question = format!(
         "## The question\n\nClassification `{}` ({wanted}):\n\n{}\n\n\
-         Answer with the JSON object only. Keep the reason under {REASON_MAX_CHARS} characters.\n",
+         Answer with the JSON object only. Keep the reason under {REASON_ASKED_CHARS} characters.\n",
         c.name,
         c.description.trim()
     );
@@ -296,6 +296,16 @@ diff --git a/src/main.rs b/src/main.rs
         assert_eq!(shared, shared_context(&run()));
     }
 
+    #[test]
+    fn the_question_asks_for_a_reason_under_the_asked_length_not_the_cap() {
+        let c = classification("datetime", "d", Polarity::Positive);
+        let question = classification_prompt("", &c).question;
+        assert!(
+            question.contains("Keep the reason under 350 characters."),
+            "{question}"
+        );
+    }
+
     #[test]
     fn the_shared_context_has_the_full_commit_message_and_a_collapsed_lockfile() {
         let shared = shared_context(&run());