@iinm/plain-agent 1.10.1 → 1.10.3

package/src/{cliInteractive.mjs → cli/interactive.mjs}

@@ -1,25 +1,28 @@
 /**
- * @import { UserEventEmitter, AgentEventEmitter, AgentCommands } from "./agent"
- * @import { ClaudeCodePlugin } from "./claudeCodePlugin.mjs"
- * @import { VoiceInputConfig, VoiceSession } from "./voiceInput.mjs"
+ * @import { UserEventEmitter, AgentEventEmitter, AgentCommands } from "../agent"
+ * @import { ClaudeCodePlugin } from "../claudeCodePlugin.mjs"
+ * @import { VoiceInputConfig } from "../voice/input.mjs"
+ * @import { VoiceSession } from "../voice/session.mjs"
  */
 
 import readline from "node:readline";
 import { styleText } from "node:util";
-import { createCommandHandler } from "./cliCommands.mjs";
-import { createCompleter, SLASH_COMMANDS } from "./cliCompleter.mjs";
+import { appendUsageRecord, buildUsageRecord } from "../usageStore.mjs";
+import { createSequentialExecutor } from "../utils/createSequentialExecutor.mjs";
+import { notify } from "../utils/notify.mjs";
+import { startVoiceSession } from "../voice/input.mjs";
+import { parseVoiceToggleKey } from "../voice/toggleKey.mjs";
+import { createCommandHandler } from "./commands.mjs";
+import { createCompleter, SLASH_COMMANDS } from "./completer.mjs";
 import {
   formatCostSummary,
   formatProviderTokenUsage,
   printMessage,
-} from "./cliFormatter.mjs";
-import { createInterruptTransform } from "./cliInterruptTransform.mjs";
-import { createMuteTransform } from "./cliMuteTransform.mjs";
-import { createPasteHandler } from "./cliPasteTransform.mjs";
-import { appendUsageRecord, buildUsageRecord } from "./usageStore.mjs";
-import { createSequentialExecutor } from "./utils/createSequentialExecutor.mjs";
-import { notify } from "./utils/notify.mjs";
-import { parseVoiceToggleKey, startVoiceSession } from "./voiceInput.mjs";
+} from "./formatter.mjs";
+import { createInterruptTransform } from "./interruptTransform.mjs";
+import { createMuteTransform } from "./muteTransform.mjs";
+import { createPasteHandler } from "./pasteTransform.mjs";
+import { createTableDetector } from "./tableDetector.mjs";
 
 const HELP_MESSAGE = [
   "Commands:",
@@ -70,7 +73,7 @@ const HELP_MESSAGE = [
  * Persist the session's cost summary to the usage log.
  * Failures are logged but never thrown so exit is not blocked.
  *
- * @param {import("./costTracker.mjs").CostSummary} summary
+ * @param {import("../costTracker.mjs").CostSummary} summary
  * @param {{ sessionId: string, modelName: string, startTime: Date }} meta
  */
 async function persistUsage(summary, { sessionId, modelName, startTime }) {
@@ -122,6 +125,9 @@ export function startInteractiveSession({
    */
   let voice = null;
 
+  // Create the table buffer instance for this session
+  const tableBuffer = createTableBuffer();
+
   // Parse the voice toggle key once at startup so misconfiguration fails
   // loudly instead of silently falling back.
   const voiceToggle = parseVoiceToggleKey(voiceInput?.toggleKey);
@@ -281,7 +287,7 @@ export function startInteractiveSession({
     // Agent turn: pause auto-approve; do not clear input.
     if (!state.turn) {
       agentCommands.pauseAutoApprove();
-      console.log(
+      console.error(
         styleText(
           "yellow",
           "\n\n⚠️ Ctrl-C: Auto-approve paused. Finishing current tool...\nPress Ctrl-D twice to exit.\n",
@@ -328,7 +334,7 @@ export function startInteractiveSession({
       );
       cli.prompt();
     } else {
-      console.log(styleText("yellow", "\n\n⚠️ Press Ctrl-D again to exit.\n"));
+      console.error(styleText("yellow", "\n\n⚠️ Press Ctrl-D again to exit.\n"));
     }
   };
 
@@ -418,7 +424,7 @@ export function startInteractiveSession({
 
   cli.on("line", async (lineInput) => {
     if (!state.turn) {
-      console.warn(
+      console.error(
         styleText(
           "yellow",
           `\nAgent is working. Ignore input: ${lineInput.trim()}`,
@@ -465,6 +471,8 @@ export function startInteractiveSession({
     if (partialContent.content) {
       if (partialContent.type === "tool_use") {
         process.stdout.write(styleText("gray", partialContent.content));
+      } else if (partialContent.type === "text") {
+        tableBuffer.feed(partialContent.content);
       } else {
         process.stdout.write(partialContent.content);
       }
@@ -511,7 +519,7 @@ export function startInteractiveSession({
   });
 
   agentEventEmitter.on("error", (error) => {
-    console.log(
+    console.error(
       styleText(
         "red",
         `\nError: message=${error.message}, stack=${error.stack}`,
@@ -520,6 +528,9 @@ export function startInteractiveSession({
   });
 
   agentEventEmitter.on("turnEnd", async () => {
+    // Flush any remaining table buffer content
+    tableBuffer.forceFlush();
+
     const err = notify(notifyCmd);
     if (err) {
       console.error(
@@ -543,3 +554,26 @@ export function startInteractiveSession({
   process.on("exit", cleanup);
   process.on("SIGTERM", cleanup);
 }
+
+/**
+ * Creates a table buffer for detecting and formatting markdown tables
+ * in streaming text output.
+ * Thin shell: delegates pure logic to createTableDetector and handles I/O.
+ */
+function createTableBuffer() {
+  const detector = createTableDetector();
+
+  function feed(/** @type {string} */ chunk) {
+    const { output, warnings } = detector.feed(chunk);
+    for (const s of output) process.stdout.write(s);
+    for (const w of warnings) console.error(styleText("yellow", w));
+  }
+
+  function forceFlush() {
+    const { output, warnings } = detector.forceFlush();
+    for (const s of output) process.stdout.write(s);
+    for (const w of warnings) console.error(styleText("yellow", w));
+  }
+
+  return { feed, forceFlush };
+}

package/src/cli/tableDetector.mjs

@@ -0,0 +1,228 @@
+import { formatMarkdownTable } from "./formatter.mjs";
+
+/**
+ * @typedef {{ output: string[], warnings: string[] }} DetectorResult
+ */
+
+/**
+ * Creates a table detector for detecting and formatting markdown tables
+ * in streaming text output. This is a pure logic module with no I/O side effects.
+ *
+ * @param {(lines: string[], maxWidth?: number) => string} [formatTable=formatMarkdownTable] - Table formatting function (injectable for testing)
+ * @param {number} [maxWidth] - Maximum terminal display width (defaults to process.stdout.columns - 4 or 80)
+ * @returns {{ feed: (chunk: string) => DetectorResult, forceFlush: () => DetectorResult }}
+ */
+export function createTableDetector(
+  formatTable = formatMarkdownTable,
+  maxWidth = process.stdout.columns ? process.stdout.columns - 4 : 80,
+) {
+  /** @type {string} - Accumulated incomplete line */
+  let pendingLine = "";
+  /** @type {string[]} - Lines of the current table being detected */
+  const tableLines = [];
+  /** @type {boolean} - Inside a code block (```) */
+  let inCodeBlock = false;
+  const MAX_TABLE_LINES = 200;
+
+  /**
+   * Feed a text chunk to the detector.
+   * @param {string} chunk
+   * @returns {DetectorResult}
+   */
+  function feed(chunk) {
+    if (chunk.length === 0) return { output: [], warnings: [] };
+    pendingLine += chunk;
+
+    /** @type {string[]} */
+    const output = [];
+    /** @type {string[]} */
+    const warnings = [];
+
+    // Process complete lines (those containing newlines)
+    while (pendingLine.includes("\n")) {
+      const idx = pendingLine.indexOf("\n");
+      const line = pendingLine.slice(0, idx); // Exclude the newline
+      pendingLine = pendingLine.slice(idx + 1);
+      const result = processLine(`${line}\n`); // Add newline back for output
+      output.push(...result.output);
+      warnings.push(...result.warnings);
+    }
+
+    // If not buffering a table and pendingLine has no pipe, output immediately
+    // This ensures non-table text is streamed without delay
+    if (tableLines.length === 0 && !pendingLine.includes("|")) {
+      output.push(pendingLine);
+      pendingLine = "";
+    }
+
+    return { output, warnings };
+  }
+
+  /**
+   * Force flush any pending content (call on turn end).
+   * @returns {DetectorResult}
+   */
+  function forceFlush() {
+    /** @type {string[]} */
+    const output = [];
+    /** @type {string[]} */
+    const warnings = [];
+
+    // Process any remaining pending line
+    if (pendingLine.length > 0) {
+      // If we have a table buffer, add pending line to it or output directly
+      if (tableLines.length > 0) {
+        tableLines.push(`${pendingLine}\n`);
+      } else {
+        output.push(pendingLine);
+      }
+      pendingLine = "";
+    }
+    const flushResult = flushTable();
+    output.push(...flushResult.output);
+    warnings.push(...flushResult.warnings);
+
+    return { output, warnings };
+  }
+
+  /**
+   * Process a complete line.
+   * @param {string} line - Line including trailing newline
+   * @returns {DetectorResult}
+   */
+  function processLine(line) {
+    /** @type {string[]} */
+    const output = [];
+    /** @type {string[]} */
+    const warnings = [];
+
+    // Code block detection
+    if (line.trimStart().startsWith("```")) {
+      inCodeBlock = !inCodeBlock;
+      const flushResult = flushTable(); // Code block terminates any ongoing table
+      output.push(...flushResult.output);
+      warnings.push(...flushResult.warnings);
+      output.push(line);
+      return { output, warnings };
+    }
+
+    if (inCodeBlock) {
+      output.push(line);
+      return { output, warnings };
+    }
+
+    // Table start: line begins with pipe
+    if (isTableStart(line)) {
+      tableLines.push(line);
+
+      // Buffer limit check
+      if (tableLines.length > MAX_TABLE_LINES) {
+        const flushResult = flushTableAsIs();
+        output.push(...flushResult.output);
+        warnings.push(...flushResult.warnings);
+      }
+      return { output, warnings };
+    }
+
+    // Table continuation: line contains pipe (for rows without leading pipe)
+    if (tableLines.length > 0 && isTableContinuation(line)) {
+      tableLines.push(line);
+      if (tableLines.length > MAX_TABLE_LINES) {
+        const flushResult = flushTableAsIs();
+        output.push(...flushResult.output);
+        warnings.push(...flushResult.warnings);
+      }
+      return { output, warnings };
+    }
+
+    // Table ended: format and flush buffer, then output current line
+    const flushResult = flushTable();
+    output.push(...flushResult.output);
+    warnings.push(...flushResult.warnings);
+    output.push(line);
+    return { output, warnings };
+  }
+
+  /**
+   * Flush table buffer with formatting.
+   * @returns {DetectorResult}
+   */
+  function flushTable() {
+    if (tableLines.length === 0) return { output: [], warnings: [] };
+
+    /** @type {string[]} */
+    const output = [];
+    /** @type {string[]} */
+    const warnings = [];
+
+    // Separate trailing empty lines (preserve spacing after table)
+    /** @type {string[]} */
+    const trailingEmpty = [];
+    while (tableLines.length > 0 && tableLines.at(-1)?.trim() === "") {
+      const line = tableLines.pop();
+      if (line !== undefined) trailingEmpty.unshift(line);
+    }
+
+    if (tableLines.length > 0) {
+      // Remove trailing newlines for formatting, then add them back
+      const rawLines = tableLines.map((l) =>
+        l.endsWith("\n") ? l.slice(0, -1) : l,
+      );
+      try {
+        const formatted = formatTable(rawLines, maxWidth);
+        output.push(`${formatted}\n`);
+      } catch (err) {
+        // Fallback: output raw lines if formatting fails
+        const message = err instanceof Error ? err.message : String(err);
+        warnings.push(`Warning: Table formatting failed: ${message}`);
+        for (const line of tableLines) {
+          output.push(line);
+        }
+      }
+    }
+
+    tableLines.length = 0;
+
+    // Output trailing empty lines
+    for (const empty of trailingEmpty) {
+      output.push(empty);
+    }
+
+    return { output, warnings };
+  }
+
+  /**
+   * Flush table buffer without formatting (for oversized tables).
+   * @returns {DetectorResult}
+   */
+  function flushTableAsIs() {
+    if (tableLines.length === 0) return { output: [], warnings: [] };
+    const output = [...tableLines];
+    tableLines.length = 0;
+    return { output, warnings: [] };
+  }
+
+  /**
+   * Check if a line starts a table.
+   * @param {string} line
+   * @returns {boolean}
+   */
+  function isTableStart(line) {
+    const trimmed = line.trimStart();
+    return trimmed.startsWith("|");
+  }
+
+  /**
+   * Check if a line continues a table.
+   * This is a heuristic: any line containing a pipe character is considered
+   * a potential table row. This may produce false positives for non-table
+   * content with pipes (e.g., "Choose A | B | C").
+   * @param {string} line
+   * @returns {boolean}
+   */
+  function isTableContinuation(line) {
+    return line.includes("|");
+  }
+
+  return { feed, forceFlush };
+}

package/src/config.d.ts

@@ -10,7 +10,7 @@ import {
   WebSearchToolGeminiOptions,
   WebSearchToolGeminiVertexAIOptions,
 } from "./tools/webSearch.mjs";
-import { VoiceInputConfig } from "./voiceInput.mjs";
+import { VoiceInputConfig } from "./voice/input.mjs";
 
 /**
  * JSON-serializable webFetch configuration.

package/src/config.mjs

@@ -128,7 +128,7 @@ export async function loadConfigFile(filePath, skipTrustCheck = false) {
 
   if (!isTrusted) {
     if (!process.stdout.isTTY) {
-      console.warn(
+      console.error(
         styleText(
           "yellow",
           `WARNING: Config file found at '${filePath}' but cannot ask for approval without a TTY. Skipping.`,

package/src/context/loadAgentRoles.mjs

@@ -54,7 +54,7 @@ export async function loadAgentRoles(claudeCodePlugins) {
       agentDirs.map(async ({ dir, idPrefix, only }) => {
         const files = await getMarkdownFiles(dir).catch((err) => {
           if (err.code !== "ENOENT") {
-            console.warn(`Failed to list agent roles in ${dir}:`, err);
+            console.error(`Failed to list agent roles in ${dir}:`, err);
           }
           return /** @type {string[]} */ ([]);
         });
@@ -72,7 +72,7 @@ export async function loadAgentRoles(claudeCodePlugins) {
         files.map(async ({ dir, file, idPrefix }) => {
           const fullPath = path.join(dir, file);
           const content = await fs.readFile(fullPath, "utf-8").catch((err) => {
-            console.warn(`Failed to read agent role file ${fullPath}:`, err);
+            console.error(`Failed to read agent role file ${fullPath}:`, err);
             return null;
           });
 

package/src/context/loadPrompts.mjs

@@ -69,7 +69,7 @@ export async function loadPrompts(claudeCodePlugins) {
       promptDirs.map(async ({ dir, idPrefix, only }) => {
         const files = await getMarkdownFiles(dir).catch((err) => {
           if (err.code !== "ENOENT") {
-            console.warn(`Failed to list prompts in ${dir}:`, err);
+            console.error(`Failed to list prompts in ${dir}:`, err);
           }
           return /** @type {string[]} */ ([]);
         });
@@ -95,7 +95,7 @@ export async function loadPrompts(claudeCodePlugins) {
         files.map(async ({ dir, file, idPrefix }) => {
           const fullPath = path.join(dir, file);
           const content = await fs.readFile(fullPath, "utf-8").catch((err) => {
-            console.warn(`Failed to read prompt file ${fullPath}:`, err);
+            console.error(`Failed to read prompt file ${fullPath}:`, err);
             return null;
           });
 

package/src/context/loadUserMessageContext.mjs

@@ -143,5 +143,5 @@ function inferMimeType(filePath) {
  * @returns {void}
  */
 function warn(message) {
-  console.warn(styleText("yellow", message));
+  console.error(styleText("yellow", message));
 }

package/src/main.mjs

@@ -10,15 +10,15 @@ import {
   installClaudeCodePlugins,
   resolvePluginPaths,
 } from "./claudeCodePlugin.mjs";
-import { parseCliArgs, printHelp } from "./cliArgs.mjs";
-import { startBatchSession } from "./cliBatch.mjs";
-import { runCostCommand } from "./cliCost.mjs";
-import { startInteractiveSession } from "./cliInteractive.mjs";
+import { parseCliArgs, printHelp } from "./cli/args.mjs";
+import { startBatchSession } from "./cli/batch.mjs";
+import { runCostCommand } from "./cli/cost.mjs";
+import { startInteractiveSession } from "./cli/interactive.mjs";
 import { loadAppConfig } from "./config.mjs";
 import { loadAgentRoles } from "./context/loadAgentRoles.mjs";
 import { loadPrompts } from "./context/loadPrompts.mjs";
 import { AGENT_PROJECT_METADATA_DIR, USER_NAME } from "./env.mjs";
-import { setupMCPServer } from "./mcpIntegration.mjs";
+import { setupMCPServer } from "./mcp/integration.mjs";
 import { createModelCaller } from "./modelCaller.mjs";
 import { createPrompt } from "./prompt.mjs";
 import { listSessions, loadSession } from "./sessionStore.mjs";

package/src/{mcpIntegration.mjs → mcp/integration.mjs}

@@ -1,16 +1,16 @@
 /**
- * @import { StructuredToolResultContent, Tool, ToolImplementation } from "./tool";
- * @import { MCPServerConfig } from "./config";
+ * @import { StructuredToolResultContent, Tool, ToolImplementation } from "../tool";
+ * @import { MCPServerConfig } from "../config";
  */
 
 import { mkdir } from "node:fs/promises";
 import path from "node:path";
-import { AGENT_PROJECT_METADATA_DIR } from "./env.mjs";
-import { createMCPClient } from "./mcpClient.mjs";
-import { writeTmpFile } from "./tmpfile.mjs";
-import { noThrow } from "./utils/noThrow.mjs";
+import { AGENT_PROJECT_METADATA_DIR } from "../env.mjs";
+import { writeTmpFile } from "../tmpfile.mjs";
+import { noThrow } from "../utils/noThrow.mjs";
+import { createMCPClient } from "./client.mjs";
 
-/** @typedef {import("./mcpClient.mjs").MCPClient} MCPClient */
+/** @typedef {import("./client.mjs").MCPClient} MCPClient */
 
 const OUTPUT_MAX_LENGTH = 1024 * 8;
 

package/src/tools/patchFile.mjs

@@ -27,18 +27,16 @@ export function createPatchFileTool(
           },
           patch: {
             description: `
-Format:
-@@@ ${nonce} {start}:{startHash}-{end}:{endHash}
+Format — a single patch string may contain multiple blocks:
+>>> ${nonce} {start}:{startHash}-{end}:{endHash}
 new content
-@@@ ${nonce}
-
-@@@ ${nonce} {N}:{afterHash}+
+<<< ${nonce}
+>>> ${nonce} {N}:{afterHash}+
 inserted content
-@@@ ${nonce}
-
-@@@ ${nonce} 0+
+<<< ${nonce}
+>>> ${nonce} 0+
 prepended content
-@@@ ${nonce}
+<<< ${nonce}
 
 - The nonce "${nonce}" is constant; always use the exact value shown above.
 - Line numbers are 1-indexed and refer to the original file; "{start}-{end}" is inclusive.
@@ -63,7 +61,7 @@ prepended content
         const blocks = parseBlocks(patch, nonce);
         if (blocks.length === 0) {
           throw new Error(
-            `No patch blocks found. Each block must start with "@@@ ${nonce} ..." and end with "@@@ ${nonce}".`,
+            `No patch blocks found. Each block must start with ">>> ${nonce} ..." and end with "<<< ${nonce}".`,
           );
         }
 
@@ -93,8 +91,8 @@ prepended content
  * @returns {PatchBlock[]}
  */
 export function parseBlocks(patch, nonce) {
-  const openPrefix = `@@@ ${nonce} `;
-  const closeMarker = `@@@ ${nonce}`;
+  const openPrefix = `>>> ${nonce} `;
+  const closeMarker = `<<< ${nonce}`;
   const lines = patch.split("\n");
 
   /** @type {PatchBlock[]} */
@@ -124,6 +122,14 @@ export function parseBlocks(patch, nonce) {
       );
     }
     const body = lines.slice(i + 1, closeIdx);
+    const nestedOpen = body.findIndex((l) => l.startsWith(openPrefix));
+    if (nestedOpen !== -1) {
+      throw new Error(
+        `Unclosed block "${openPrefix}${headerArgs}": found another open marker "${body[nestedOpen]}" ` +
+          `at line ${i + 1 + nestedOpen + 1} of patch before the close marker. ` +
+          `Did you forget "${closeMarker}" to close the previous block?`,
+      );
+    }
     if (header.op === "insert" && body.length === 0) {
       throw new Error(
         `Insert block "${openPrefix}${headerArgs}" has empty body. Use a replace block to delete content.`,

package/src/{voiceInputGemini.mjs → voice/gemini.mjs}

@@ -1,10 +1,7 @@
-import {
-  isObjectLike,
-  startWebSocketVoiceSession,
-} from "./voiceInputSession.mjs";
+import { isObjectLike, startWebSocketVoiceSession } from "./session.mjs";
 
 /**
- * @import { VoiceProviderHooks, VoiceRecorderConfig, VoiceSession, VoiceSessionCallbacks } from "./voiceInputSession.mjs"
+ * @import { VoiceProviderHooks, VoiceRecorderConfig, VoiceSession, VoiceSessionCallbacks } from "./session.mjs"
  */
 
 /**

package/src/voice/input.mjs

@@ -0,0 +1,29 @@
+import { startGeminiVoiceSession } from "./gemini.mjs";
+import { startOpenAIVoiceSession } from "./openai.mjs";
+import { failVoiceSessionAsync } from "./session.mjs";
+
+/**
+ * @typedef {import("./openai.mjs").VoiceInputOpenAIConfig | import("./gemini.mjs").VoiceInputGeminiConfig} VoiceInputConfig
+ */
+/**
+ * Start a voice input session. Dispatches to the provider-specific
+ * implementation based on `config.provider`.
+ *
+ * @param {object} options
+ * @param {VoiceInputConfig} options.config
+ * @param {import("./session.mjs").VoiceSessionCallbacks} options.callbacks
+ * @returns {import("./session.mjs").VoiceSession}
+ */
+export function startVoiceSession({ config, callbacks }) {
+  if (config.provider === "openai") {
+    return startOpenAIVoiceSession({ config, callbacks });
+  }
+  if (config.provider === "gemini") {
+    return startGeminiVoiceSession({ config, callbacks });
+  }
+  const provider = /** @type {{ provider: string }} */ (config).provider;
+  return failVoiceSessionAsync(
+    callbacks,
+    new Error(`Unsupported voiceInput.provider: ${provider}`),
+  );
+}

package/src/{voiceInputOpenAI.mjs → voice/openai.mjs}

@@ -1,24 +1,21 @@
-import {
-  isObjectLike,
-  startWebSocketVoiceSession,
-} from "./voiceInputSession.mjs";
+import { isObjectLike, startWebSocketVoiceSession } from "./session.mjs";
 
 /**
- * @import { VoiceProviderHooks, VoiceRecorderConfig, VoiceSession, VoiceSessionCallbacks } from "./voiceInputSession.mjs"
+ * @import { VoiceProviderHooks, VoiceRecorderConfig, VoiceSession, VoiceSessionCallbacks } from "./session.mjs"
  */
 
 /**
  * @typedef {Object} VoiceInputOpenAIConfig
  * @property {"openai"} provider
  * @property {string} apiKey
- * @property {string} [model] - Defaults to "gpt-4o-transcribe".
+ * @property {string} [model] - Transcription model. Defaults to "gpt-realtime-whisper".
  * @property {string} [language] - ISO-639-1 code (e.g. "ja", "en"). Improves accuracy and latency when set.
  * @property {string} [baseURL]
  * @property {VoiceRecorderConfig} [recorder]
  * @property {string} [toggleKey] - "ctrl-<char>". Defaults to "ctrl-o".
  */
 
-const OPENAI_DEFAULT_MODEL = "gpt-4o-transcribe";
+const OPENAI_DEFAULT_TRANSCRIPTION_MODEL = "gpt-realtime-whisper";
 const OPENAI_DEFAULT_WS = "wss://api.openai.com/v1/realtime";
 const OPENAI_SAMPLE_RATE = 24000;
 const OPENAI_LABEL = "OpenAI Realtime";
@@ -46,31 +43,32 @@ export function startOpenAIVoiceSession({ config, callbacks }) {
       return {
         headers: {
           Authorization: `Bearer ${config.apiKey}`,
-          "OpenAI-Beta": "realtime=v1",
         },
       };
     },
     buildSetupMessage(config) {
-      const model = config.model ?? OPENAI_DEFAULT_MODEL;
+      const model = config.model ?? OPENAI_DEFAULT_TRANSCRIPTION_MODEL;
       /** @type {{ model: string, language?: string }} */
       const transcription = { model };
       if (config.language) transcription.language = config.language;
-      // The `?intent=transcription` endpoint uses the flat transcription-session
-      // schema, not the nested `session.audio.input.*` realtime schema.
       return {
-        type: "transcription_session.update",
+        type: "session.update",
         session: {
-          input_audio_format: "pcm16",
-          input_audio_transcription: transcription,
-          turn_detection: { type: "server_vad" },
+          type: "transcription",
+          audio: {
+            input: {
+              format: { type: "audio/pcm", rate: OPENAI_SAMPLE_RATE },
+              transcription,
+            },
+          },
         },
       };
     },
     isReadyMessage(message) {
       return (
         isObjectLike(message) &&
-        (message.type === "transcription_session.created" ||
-          message.type === "transcription_session.updated")
+        (message.type === "session.created" ||
+          message.type === "session.updated")
       );
     },
     extractError(message) {