@getdial/cli 0.42.0 → 0.43.0

package/dist/commands/call/get.js

@@ -1,6 +1,48 @@
 import { getCall } from "../../lib/ops/calls.js";
 import { isDialError } from "../../lib/ops/errors.js";
 import { printDialError } from "../../lib/cli-error.js";
+/**
+ * A silence long enough to call out. Three seconds is past a natural beat between
+ * speakers but short enough to catch an agent thinking too long, which is what
+ * anyone reading a transcript for pacing is looking for.
+ */
+const NOTABLE_PAUSE_MS = 3000;
+/**
+ * Plain words for who spoke. "person" rather than "caller" because on an outbound
+ * call the human is the callee; "transferred" for the third party a call was
+ * handed off to, who is a different human from the one who was on the line first.
+ */
+function speakerLabel(speaker) {
+    if (speaker === "agent")
+        return "agent";
+    return speaker === "transfer_target" ? "transferred" : "person";
+}
+/** ms into the call's audio → "m:ss.t", the stamp each transcript line carries. */
+function formatOffset(ms) {
+    const minutes = Math.floor(ms / 60000);
+    const seconds = Math.floor((ms % 60000) / 1000);
+    const tenths = Math.floor((ms % 1000) / 100);
+    return `${minutes}:${String(seconds).padStart(2, "0")}.${tenths}`;
+}
+/**
+ * Print the transcript with each turn stamped, and call out the long silences
+ * between them. The flat string cannot show a pause, which is the whole reason
+ * the timed turns exist.
+ */
+function printTimedTranscript(turns) {
+    console.log(`transcript:`);
+    turns.forEach((turn, i) => {
+        const previous = turns[i - 1];
+        if (previous) {
+            const pauseMs = turn.startMs - previous.endMs;
+            if (pauseMs >= NOTABLE_PAUSE_MS) {
+                console.log(`            ... ${(pauseMs / 1000).toFixed(1)}s pause`);
+            }
+        }
+        const speaker = speakerLabel(turn.speaker).padEnd(12);
+        console.log(`  ${formatOffset(turn.startMs).padStart(7)}  ${speaker}${turn.text}`);
+    });
+}
 export async function runCallGet(opts) {
     try {
         const c = await getCall(opts.callId);
@@ -19,7 +61,10 @@ export async function runCallGet(opts) {
             console.log(`instruction:`);
             console.log(c.instruction);
         }
-        if (c.transcript) {
+        if (c.transcriptTurns?.length) {
+            printTimedTranscript(c.transcriptTurns);
+        }
+        else if (c.transcript) {
             console.log(`transcript:`);
             console.log(c.transcript);
         }

package/dist/mcp/schemas.js

@@ -147,6 +147,17 @@ export const messageSchema = z
     createdAt: z.string().optional(),
 })
     .passthrough();
+export const transcriptTurnSchema = z
+    .object({
+    speaker: z
+        .enum(["agent", "user", "transfer_target"])
+        .describe("`agent` is Dial's AI voice agent, `user` the human on the other end, and " +
+        "`transfer_target` the human the call was cold-transferred to"),
+    text: z.string().describe("What was said during the turn"),
+    startMs: z.number().describe("Approximate ms into the call's audio at which the turn began"),
+    endMs: z.number().describe("Approximate ms into the call's audio at which the turn ended"),
+})
+    .describe("One uninterrupted stretch of speech by one party, placed in time");
 export const callSchema = z
     .object({
     id: z.string(),
@@ -156,6 +167,13 @@ export const callSchema = z
     status: statusSchema,
     duration: z.number().nullish(),
     transcript: z.string().nullish(),
+    transcriptTurns: transcriptTurnSchema
+        .array()
+        .nullish()
+        .describe("The same conversation as `transcript`, split into timed turns and ordered by " +
+        "startMs. Use it to measure pacing: the pause before a turn is its startMs minus " +
+        "the previous turn's endMs. Null when the call has no transcript, or when the " +
+        "call's turn timing was not recorded."),
     instruction: z.string().nullable().optional(),
     createdAt: z.string().optional(),
 })

package/dist/mcp/tools/get-call.js

@@ -9,7 +9,9 @@ export const getCallTool = {
     name: "get_call",
     config: {
         title: "Get Call",
-        description: "Fetch a single call by id — status, duration, and transcript when available.",
+        description: "Fetch a single call by id — status, duration, and transcript when available. " +
+            "The transcript comes back twice: `transcript` as flat text, and `transcriptTurns` " +
+            "as timed turns for analysing pacing and spotting long pauses.",
         inputSchema,
         outputSchema: { call: callSchema },
         annotations: { readOnlyHint: true, openWorldHint: true },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@getdial/cli",
-  "version": "0.42.0",
+  "version": "0.43.0",
   "description": "Dial CLI — install, sign up, and run the local listen service.",
   "license": "MIT",
   "repository": {