Merge feat/vocabulary-hint: names-and-terms decoder prompt (spec 001 workstream 2)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

# Conflicts:
#	docs/Architecture.md
#	docs/Decisions.md
#	docs/WorkBreakdown.md
#	server-py/sidecar.py
#	src/App.tsx
#	src/server/live.ts
#	src/server/transcribe.ts
#	tests/fixtures/fake_sidecar.py
#	tests/transcribe.test.ts

src/worker/inference.worker.ts

@@ -17,6 +17,13 @@ import {
   type ProgressInfo,
 } from "@huggingface/transformers";
 import { resolveDefaultDtype } from "../lib/models";
+import {
+  buildPromptPrefix,
+  installPromptSupport,
+  PROMPT_PREFIX_KWARG,
+  type PromptableModel,
+  type PromptTokenizer,
+} from "./prompt";
 import {
   deviceAttempts,
   dtypeRequiresF16,
@@ -409,6 +416,17 @@ async function transcribe(message: TranscribeMessage): Promise<void> {
   if (message.task !== undefined) {
     options.task = message.task;
   }
+  // "Names and terms" hint (D-WS2a): one prefix per pass, applied by the
+  // generate() wrapper to every 30 s window (Transformers.js ignores
+  // `prompt_ids`, see src/worker/prompt.ts). The wrapper is idempotent.
+  const promptPrefix = buildPromptPrefix(
+    asr.tokenizer as unknown as PromptTokenizer,
+    message.prompt ?? "",
+  );
+  if (promptPrefix.length > 0) {
+    installPromptSupport(asr.model as unknown as PromptableModel);
+    options[PROMPT_PREFIX_KWARG] = promptPrefix;
+  }
 
   const output = await asr(message.audio, options);