feat: transcribe audio and surface transcript to prompts

docs/audio.md

@@ -0,0 +1,47 @@
+# Audio / Voice Notes — 2025-11-25
+
+## What works
+- **Optional transcription**: If `inbound.transcribeAudio.command` is set in `~/.warelay/warelay.json`, warelay will:
+  1) Download inbound audio (Web or Twilio) to a temp path if only a URL is present.
+  2) Run the configured CLI (templated with `{{MediaPath}}`), expecting transcript on stdout.
+  3) Replace `Body` with the transcript, set `{{Transcript}}`, and prepend the original media path plus a `Transcript:` section in the command prompt so models see both.
+  4) Continue through the normal auto-reply pipeline (templating, sessions, Claude/command).
+- **Verbose logging**: In `--verbose`, we log when transcription runs and when the transcript replaces the body.
+
+## Config example (OpenAI Whisper CLI)
+Requires `OPENAI_API_KEY` in env and `openai` CLI installed:
+```json5
+{
+  inbound: {
+    transcribeAudio: {
+      command: [
+        "openai",
+        "api",
+        "audio.transcriptions.create",
+        "-m",
+        "whisper-1",
+        "-f",
+        "{{MediaPath}}",
+        "--response-format",
+        "text"
+      ],
+      timeoutSeconds: 45
+    },
+    reply: {
+      mode: "command",
+      command: ["claude", "{{Body}}"]
+    }
+  }
+}
+```
+
+## Notes & limits
+- We don’t ship a transcriber; you opt in with any CLI that prints text to stdout (Whisper cloud, whisper.cpp, vosk, Deepgram, etc.).
+- Size guard: inbound audio must be ≤5 MB (same as other media).
+- If transcription fails, we fall back to the original body/media note; replies still go through.
+- Transcript is available to templates as `{{Transcript}}`; models get both the media path and a `Transcript:` block in the prompt when using command mode.
+
+## Gotchas
+- Ensure your CLI exits 0 and prints plain text; JSON needs to be massaged via `jq -r .text`.
+- Keep timeouts reasonable (`timeoutSeconds`, default 45s) to avoid blocking the reply queue.
+- Twilio paths are hosted URLs; Web paths are local. The temp download uses HTTPS for Twilio and a temp file for Web-only media.

docs/images.md

@@ -57,7 +57,7 @@ This document defines how `warelay` should handle sending and replying with imag
   - `{{MediaUrl}}` original URL (Twilio) or pseudo-URL (web).
   - `{{MediaPath}}` local temp path written before running the command.
 - Size guard: only download if ≤5 MB; else skip and log.
-- Audio/voice notes: if you set `inbound.transcribeAudio.command`, warelay will run that CLI (templated with `{{MediaPath}}`) and replace `Body` with the transcript before continuing the reply flow; verbose logs indicate when transcription runs.
+- Audio/voice notes: if you set `inbound.transcribeAudio.command`, warelay will run that CLI (templated with `{{MediaPath}}`) and replace `Body` with the transcript before continuing the reply flow; verbose logs indicate when transcription runs. The command prompt includes the original media path plus a `Transcript:` section so the model sees both.
 
 ## Errors & Messaging
 - Local path with twilio + Funnel disabled → error: “Twilio media needs a public URL; start `warelay webhook --ingress tailscale` or pass an https:// URL.”

src/auto-reply/reply.ts

@@ -110,12 +110,15 @@ export async function getReplyFromConfig(
 		started = true;
 		await opts?.onReplyStart?.();
 	};
+	let transcribedText: string | undefined;
 
 	// Optional audio transcription before templating/session handling.
 	if (cfg.inbound?.transcribeAudio && isAudio(ctx.MediaType)) {
 		const transcribed = await transcribeInboundAudio(cfg, ctx, defaultRuntime);
 		if (transcribed?.text) {
+			transcribedText = transcribed.text;
 			ctx.Body = transcribed.text;
+			ctx.Transcript = transcribed.text;
 			logVerbose("Replaced Body with audio transcript for reply flow");
 		}
 	}
@@ -193,9 +196,15 @@ export async function getReplyFromConfig(
 	const bodyPrefix = reply?.bodyPrefix
 		? applyTemplate(reply.bodyPrefix, sessionCtx)
 		: "";
-	const prefixedBody = bodyPrefix
+	const prefixedBodyBase = bodyPrefix
 		? `${bodyPrefix}${sessionCtx.BodyStripped ?? sessionCtx.Body ?? ""}`
 		: (sessionCtx.BodyStripped ?? sessionCtx.Body);
+	const prefixedBody =
+		transcribedText && reply?.mode === "command"
+			? [prefixedBodyBase, `Transcript:\n${transcribedText}`]
+					.filter(Boolean)
+					.join("\n\n")
+			: prefixedBodyBase;
 	const mediaNote = ctx.MediaPath?.length
 		? `[media attached: ${ctx.MediaPath}${ctx.MediaType ? ` (${ctx.MediaType})` : ""}${ctx.MediaUrl ? ` | ${ctx.MediaUrl}` : ""}]`
 		: undefined;
@@ -543,7 +552,7 @@ export async function autoReplyIfConfigured(
 }
 
 function isAudio(mediaType?: string | null) {
-	return Boolean(mediaType && mediaType.startsWith("audio"));
+	return Boolean(mediaType?.startsWith("audio"));
 }
 
 async function transcribeInboundAudio(
@@ -596,9 +605,7 @@ async function transcribeInboundAudio(
 		return undefined;
 	} finally {
 		if (tmpPath) {
-			void fs
+			void fs.unlink(tmpPath).catch(() => {});
-				.unlink(tmpPath)
-				.catch(() => {});
 		}
 	}
 }

src/auto-reply/templating.ts

@@ -6,6 +6,7 @@ export type MsgContext = {
 	MediaPath?: string;
 	MediaUrl?: string;
 	MediaType?: string;
+	Transcript?: string;
 };
 
 export type TemplateContext = MsgContext & {

src/index.core.test.ts

@@ -5,8 +5,8 @@ import path from "node:path";
 import type { MessageInstance } from "twilio/lib/rest/api/v2010/account/message.js";
 import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 import { createMockTwilio } from "../test/mocks/twilio.js";
-import { withWhatsAppPrefix } from "./utils.js";
 import * as exec from "./process/exec.js";
+import { withWhatsAppPrefix } from "./utils.js";
 
 // Twilio mock factory shared across tests
 vi.mock("twilio", () => {
@@ -125,8 +125,8 @@ describe("config and templating", () => {
 					command: ["echo", "voice transcript"],
 				},
 				reply: {
-					mode: "text" as const,
+					mode: "command" as const,
-					text: "{{Body}}",
+					command: ["echo", "{{Body}}"],
 				},
 			},
 		};
@@ -135,6 +135,13 @@ describe("config and templating", () => {
 			stdout: "voice transcript\n",
 			stderr: "",
 		});
+		const commandRunner = vi.fn().mockResolvedValue({
+			stdout: "ok",
+			stderr: "",
+			code: 0,
+			signal: null,
+			killed: false,
+		});
 
 		const result = await index.getReplyFromConfig(
 			{
@@ -146,11 +153,17 @@ describe("config and templating", () => {
 			},
 			undefined,
 			cfg,
+			commandRunner,
 		);
 
 		expect(runExec).toHaveBeenCalled();
-		expect(result?.text).toContain("voice transcript");
+		expect(commandRunner).toHaveBeenCalled();
-		expect(result?.text).toContain("/tmp/voice.ogg");
+		const argv = commandRunner.mock.calls[0][0];
+		const prompt = argv[argv.length - 1] as string;
+		expect(prompt).toContain("/tmp/voice.ogg");
+		expect(prompt).toContain("Transcript:");
+		expect(prompt).toContain("voice transcript");
+		expect(result?.text).toBeUndefined();
 	});
 
 	it("getReplyFromConfig skips transcription when not configured", async () => {

src/provider-web.test.ts

@@ -602,7 +602,8 @@ describe("provider-web", () => {
 
 				fetchMock.mockRestore();
 			}
-	});
+		},
+	);
 
 	it("honors mediaMaxMb from config", async () => {
 		loadConfigMock = () => ({ inbound: { reply: { mediaMaxMb: 1 } } });