nolo-cli 0.1.39 → 0.1.40

package/ai/agent/cliExecutor.ts

@@ -15,8 +15,8 @@
 
 import { exec, execSync, spawn } from "child_process";
 import { randomUUID } from "node:crypto";
-import { mkdtempSync, readFileSync, rmSync } from "node:fs";
-import { join } from "node:path";
+import { mkdtempSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { isAbsolute, join, resolve } from "node:path";
 import { tmpdir } from "node:os";
 import { resolveLaunchableCodexCommand } from "../../connector-experimental/codexBinary";
 import { buildCliPrompt } from "./cliPrompt";
@@ -26,6 +26,19 @@ import { buildCliPrompt } from "./cliPrompt";
 /** 已支持的 CLI 工具。新增时在此联合类型追加，并在 EXECUTORS 里注册实现 */
 export type CliProvider = "copilot" | "gemini" | "codex" | "claude" | "agy";
 
+/**
+ * CLI provider 图片输入。
+ *
+ * - `source` 是原始输入（本地路径 / data URL / http URL / file URL）
+ * - `materializedPath` 是 materialize 后写入磁盘的绝对路径
+ *   （本地路径保持不变，data URL 写入临时文件，file URL 转换为本地路径）
+ * - HTTP/HTTPS URL 目前无法原生传给 CLI，只作为 prompt 引用保留
+ */
+export interface CliImageInput {
+  source: string;
+  materializedPath?: string;
+}
+
 export interface CliExecuteOptions {
   model?: string;
   timeout?: number;   // ms，默认 120_000
@@ -41,6 +54,8 @@ export interface CliExecuteOptions {
   maxTokens?: number;
   enableThinking?: boolean;
   thinkingBudget?: number;
+  /** 图片输入（本地路径 / data URL / file URL）。materialize 由内部处理。 */
+  imageInputs?: CliImageInput[];
 }
 
 export interface CliExecuteResult {
@@ -184,6 +199,138 @@ function shouldPassClaudeModel(options: CliExecuteOptions): boolean {
   return (options.env?.NOLO_CLAUDE_CLI_ALLOW_MODEL ?? process.env.NOLO_CLAUDE_CLI_ALLOW_MODEL) === "1";
 }
 
+// ── CLI 图片 materialize ─────────────────────────────────────────────────────
+
+const DATA_URL_RE = /^data:image\/([\w.+-]+);base64,([A-Za-z0-9+/=]+)$/;
+
+/** data URL extension (without prefix) → 完整扩展名 */
+const MIME_EXT: Record<string, string> = {
+  "png": ".png",
+  "jpeg": ".jpg",
+  "jpg": ".jpg",
+  "gif": ".gif",
+  "webp": ".webp",
+  "svg+xml": ".svg",
+  "bmp": ".bmp",
+  "tiff": ".tiff",
+  "avif": ".avif",
+};
+
+function inferExtFromMime(extPart: string): string {
+  return MIME_EXT[extPart.toLowerCase().trim()] ?? ".bin";
+}
+
+const FILE_URL_RE = /^file:\/\/(.+)$/;
+
+/**
+ * 将单个 data URL 图片写入临时文件，返回绝对路径。
+ * caller 负责在执行完后清理 tempDir。
+ */
+function writeDataUrlToTempImage(dataUrl: string, tempDir: string, index: number): string {
+  const match = dataUrl.match(DATA_URL_RE);
+  if (!match) {
+    throw new Error(`Invalid data URL format for image #${index + 1}; expected data:image/<type>;base64,<data>`);
+  }
+  const [, mimeExt, b64] = match;
+  const ext = inferExtFromMime(mimeExt);
+  const filePath = join(tempDir, `image-${index + 1}${ext}`);
+  writeFileSync(filePath, Buffer.from(b64, "base64"));
+  return filePath;
+}
+
+/**
+ * Materialize image input sources into on-disk paths where possible.
+ *
+ * - 本地 path / file URL → 解析为绝对路径，无需写临时文件
+ * - data:image/...;base64,... → 写入 tempDir 内临时文件
+ * - http(s) URL → 不做下载，保留 source 引用；materializedPath 为空
+ *
+ * 返回 `{ materialized, tempDir }`，caller 需在执行完成后 `rmSync(tempDir, { recursive: true, force: true })`。
+ * 当没有任何 data URL 时 tempDir 不会被创建，返回 null。
+ */
+export function materializeCliImageInputs(
+  imageInputs: CliImageInput[],
+  cwd: string,
+): { materialized: CliImageInput[]; tempDir: string | null } {
+  let tempDir: string | null = null;
+  const materialized = imageInputs.map((input, index) => {
+    const src = input.source.trim();
+
+    if (src.startsWith("http://") || src.startsWith("https://")) {
+      return { ...input, materializedPath: undefined };
+    }
+
+    const fileUrlMatch = src.match(FILE_URL_RE);
+    if (fileUrlMatch) {
+      const localPath = fileUrlMatch[1];
+      return {
+        ...input,
+        materializedPath: isAbsolute(localPath) ? localPath : resolve(cwd, localPath),
+      };
+    }
+
+    if (src.startsWith("data:")) {
+      if (!tempDir) {
+        tempDir = mkdtempSync(join(tmpdir(), "nolo-cli-assets-"));
+      }
+      const filePath = writeDataUrlToTempImage(src, tempDir, index);
+      return { ...input, materializedPath: filePath };
+    }
+
+    // 本地路径
+    return {
+      ...input,
+      materializedPath: isAbsolute(src) ? src : resolve(cwd, src),
+    };
+  });
+
+  return { materialized, tempDir };
+}
+
+/**
+ * 根据 provider 构建图片降级 prompt 块。
+ * Codex 使用原生 `-i` 标志，无需注入 prompt。
+ * 其他 provider 使用此函数将图片路径列表注入到 prompt 前面。
+ */
+function buildImageReferencePromptBlock(
+  provider: CliProvider,
+  materialized: CliImageInput[],
+): { promptSuffix: string; warning: string | null } {
+  const label = getCliProviderLabel(provider);
+  const localPaths = materialized
+    .filter((img) => img.materializedPath)
+    .map((img) => img.materializedPath!);
+  const httpUrls = materialized
+    .filter((img) => !img.materializedPath)
+    .map((img) => img.source);
+
+  const lines: string[] = [];
+  if (localPaths.length > 0) {
+    lines.push(
+      `[User attached image(s) - local file references (accessible at these paths):]`,
+      ...localPaths.map((p) => `  - ${p}`),
+      `Please read and analyze these image files as part of the user's request.`,
+    );
+  }
+  if (httpUrls.length > 0) {
+    lines.push(
+      `[User attached image(s) - remote URLs (content not available locally):]`,
+      ...httpUrls.map((u) => `  - ${u}`),
+      `These remote URLs could not be downloaded automatically. If you cannot access them, describe what you can infer from context.`,
+    );
+  }
+
+  const warning =
+    materialized.length > 0
+      ? `${label} image input is passed as local file references; native image flags are not available in this wrapper.`
+      : null;
+
+  return {
+    promptSuffix: lines.length > 0 ? lines.join("\n") + "\n\n" : "",
+    warning,
+  };
+}
+
 function collectCliCapabilityWarnings(
   provider: CliProvider,
   options: CliExecuteOptions
@@ -226,18 +373,51 @@ function resolveCliExecution(
   provider: CliProvider,
   prompt: string,
   options: CliExecuteOptions
-): { prompt: string; options: CliExecuteOptions; warnings: string[] } {
+): { prompt: string; options: CliExecuteOptions; warnings: string[]; imageTempDir?: string } {
   const warnings = collectCliCapabilityWarnings(provider, options);
+
+  // Handle image inputs
+  const imageInputs = options.imageInputs ?? [];
+  let resolvedPrompt = prompt;
+  let resolvedOptions = options;
+  let imageTempDir: string | undefined;
+
+  if (imageInputs.length > 0) {
+    const { materialized, tempDir } = materializeCliImageInputs(
+      imageInputs,
+      options.cwd ?? process.cwd(),
+    );
+    imageTempDir = tempDir ?? undefined;
+
+    if (provider === "codex") {
+      // Codex uses native -i flags; update options with materialized imageInputs
+      resolvedOptions = {
+        ...resolvedOptions,
+        imageInputs: materialized,
+      };
+    } else {
+      // Non-codex providers: inject image references into prompt
+      const block = buildImageReferencePromptBlock(provider, materialized);
+      if (block.promptSuffix) {
+        resolvedPrompt = block.promptSuffix + resolvedPrompt;
+      }
+      if (block.warning) {
+        warnings.push(block.warning);
+      }
+    }
+  }
+
   if (supportsNativeSystemPrompt(provider) || !options.systemPrompt?.trim()) {
-    return { prompt, options, warnings };
+    return { prompt: resolvedPrompt, options: resolvedOptions, warnings, imageTempDir };
   }
   return {
-    prompt: buildCliPrompt(options.systemPrompt, prompt),
+    prompt: buildCliPrompt(options.systemPrompt, resolvedPrompt),
     options: {
-      ...options,
+      ...resolvedOptions,
       systemPrompt: undefined,
     },
     warnings,
+    imageTempDir,
   };
 }
 
@@ -410,6 +590,11 @@ function executeCodex(
     yolo = true,
   } = options;
 
+  // Collect materialized image paths from imageInputs
+  const imagePaths = (options.imageInputs ?? [])
+    .map((img) => img.materializedPath)
+    .filter((p): p is string => Boolean(p));
+
   return new Promise((resolve, reject) => {
     const tempDir = mkdtempSync(join(tmpdir(), "codex-exec-"));
     const outputFile = join(tempDir, "last-message.txt");
@@ -429,6 +614,10 @@ function executeCodex(
       args.push("--sandbox");
       args.push("danger-full-access");
     }
+    // Add image flags for Codex CLI native image support
+    for (const imgPath of imagePaths) {
+      args.push("-i", imgPath);
+    }
     const start = Date.now();
     const proc = spawn(executable, [...args, prompt], {
       cwd,
@@ -699,11 +888,17 @@ export async function executeCli(
     throw new Error(`Unknown CLI provider: "${provider}". Supported: ${Object.keys(EXECUTORS).join(", ")}`);
   }
   const resolved = resolveCliExecution(provider, prompt, options);
-  const result = await executor(resolved.prompt, resolved.options);
-  return {
-    ...result,
-    warnings: [...resolved.warnings, ...(result.warnings ?? [])],
-  };
+  try {
+    const result = await executor(resolved.prompt, resolved.options);
+    return {
+      ...result,
+      warnings: [...resolved.warnings, ...(result.warnings ?? [])],
+    };
+  } finally {
+    if (resolved.imageTempDir) {
+      rmSync(resolved.imageTempDir, { recursive: true, force: true });
+    }
+  }
 }
 
 export { buildCliPrompt };
@@ -831,36 +1026,56 @@ export function executeCliStreaming(
 ): Promise<CliExecuteResult> {
   const resolved = resolveCliExecution(provider, prompt, options);
   const executor = EXECUTORS[provider];
+
+  const cleanupImageTempDir = () => {
+    if (resolved.imageTempDir) {
+      rmSync(resolved.imageTempDir, { recursive: true, force: true });
+    }
+  };
+
+  const withCleanup = <T>(promise: Promise<T>): Promise<T> =>
+    promise.finally(cleanupImageTempDir);
+
   if (provider === "gemini") {
-    return executeGeminiStreaming(resolved.prompt, resolved.options as CliExecuteOptions & { onChunk: (chunk: string) => void })
-      .then((result) => ({ ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] }));
+    return withCleanup(
+      executeGeminiStreaming(resolved.prompt, resolved.options as CliExecuteOptions & { onChunk: (chunk: string) => void })
+        .then((result) => ({ ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] }))
+    );
   }
   if (provider === "codex") {
-    return executor(resolved.prompt, resolved.options).then((result) => {
-      const merged = { ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] };
-      if (merged.text) options.onChunk(merged.text);
-      return merged;
-    });
+    return withCleanup(
+      executor(resolved.prompt, resolved.options).then((result) => {
+        const merged = { ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] };
+        if (merged.text) options.onChunk(merged.text);
+        return merged;
+      })
+    );
   }
   if (provider === "claude") {
-    return executor(resolved.prompt, resolved.options).then((result) => {
-      const merged = { ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] };
-      if (merged.text) options.onChunk(merged.text);
-      return merged;
-    });
+    return withCleanup(
+      executor(resolved.prompt, resolved.options).then((result) => {
+        const merged = { ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] };
+        if (merged.text) options.onChunk(merged.text);
+        return merged;
+      })
+    );
   }
   if (provider === "agy") {
-    return executor(resolved.prompt, resolved.options).then((result) => {
-      const merged = { ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] };
-      if (merged.text) options.onChunk(merged.text);
-      return merged;
-    });
+    return withCleanup(
+      executor(resolved.prompt, resolved.options).then((result) => {
+        const merged = { ...result, warnings: [...resolved.warnings, ...(result.warnings ?? [])] };
+        if (merged.text) options.onChunk(merged.text);
+        return merged;
+      })
+    );
   }
   if (provider !== "copilot") {
-    return executor(resolved.prompt, resolved.options).then((result) => ({
-      ...result,
-      warnings: [...resolved.warnings, ...(result.warnings ?? [])],
-    }));
+    return withCleanup(
+      executor(resolved.prompt, resolved.options).then((result) => ({
+        ...result,
+        warnings: [...resolved.warnings, ...(result.warnings ?? [])],
+      }))
+    );
   }
 
   const {
@@ -871,7 +1086,7 @@ export function executeCliStreaming(
     onChunk,
   } = resolved.options as CliExecuteOptions & { onChunk: (chunk: string) => void };
 
-  return new Promise((resolve, reject) => {
+  return withCleanup(new Promise<CliExecuteResult>((resolve, reject) => {
     const args = [
       "copilot", "--",
       "-p", resolved.prompt,
@@ -931,7 +1146,7 @@ export function executeCliStreaming(
       if (timer) clearTimeout(timer);
       reject(err);
     });
-  });
+  }));
 }
 
 function executeGeminiStreaming(

package/cli/agentAliases.ts

@@ -1,4 +1,5 @@
 export const PLATFORM_DEMO_USER_ID = "b2e06f801f";
+export const CURRENT_OPERATOR_USER_ID = "0e95801d90";
 export const NOLO_DEFAULT_AGENT_ID = "01NOLOAPPBLD000000019KCKT0";
 export const NOLO_DEFAULT_AGENT_KEY = `agent-pub-${NOLO_DEFAULT_AGENT_ID}`;
 export const NOLO_PROJECT_MANAGER_AGENT_ID = "01NOLOPROJMGR00000000MSVGG";
@@ -9,7 +10,7 @@ export const NOLO_SENIOR_FULLSTACK_AGENT_KEY =
   `agent-${PLATFORM_DEMO_USER_ID}-${NOLO_SENIOR_FULLSTACK_AGENT_ID}`;
 export const NOLO_FRONTEND_AGENT_ID = "01FRONTENDAG0000000115N4E1";
 export const NOLO_FRONTEND_AGENT_KEY =
-  `agent-${PLATFORM_DEMO_USER_ID}-${NOLO_FRONTEND_AGENT_ID}`;
+  `agent-${CURRENT_OPERATOR_USER_ID}-${NOLO_FRONTEND_AGENT_ID}`;
 export const WIN_CODEX_AGENT_ID = "WINCODEX00000000FQ0LK0";
 export const WIN_CODEX_AGENT_KEY =
   `agent-0e95801d90-${WIN_CODEX_AGENT_ID}`;

package/cli/agentRunCommand.ts

@@ -50,11 +50,10 @@ type ParsedAgentRunArgs = {
   parentDialogId?: string;
   background: boolean;
   noStream: boolean;
-  noDefaultTestRoot: boolean;
   cwd?: string;
-  maxToolRounds?: number;
   timeoutMs?: number;
   traceTools: boolean;
+  eventsMode?: "jsonl";
   injectFeatureWorktreeInstruction: boolean;
   taskRunContext?: TaskRunPromptContext;
   workflowRef?: string;
@@ -66,10 +65,6 @@ type ResolvedWorkflowReference = {
   config?: Partial<WorkflowReferenceConfig>;
 };
 
-const DEFAULT_TEST_ROOT_DIALOG_ID = "01KN6V3MVEQA4Q1PNKVM2W7YY8";
-const DEFAULT_TEST_ROOT_DIALOG_KEY = `dialog-b2e06f801f-${DEFAULT_TEST_ROOT_DIALOG_ID}`;
-const DEFAULT_TEST_DIALOG_CATEGORY_ID = "test-dialogs";
-
 function readFlagValue(args: string[], flag: string) {
   const index = args.indexOf(flag);
   if (index < 0) return undefined;
@@ -109,13 +104,6 @@ function isMonthlyMimoAgentRef(raw: string | undefined, resolved: string) {
   );
 }
 
-function parseNonNegativeInteger(value: string | undefined) {
-  if (!value) return undefined;
-  const parsed = Number(value);
-  if (!Number.isFinite(parsed) || parsed < 0) return undefined;
-  return Math.floor(parsed);
-}
-
 function parsePositiveInteger(value: string | undefined) {
   if (!value) return undefined;
   const parsed = Number(value);
@@ -134,7 +122,6 @@ function positionalArgs(args: string[]) {
     "--bg",
     "--no-stream",
     "--debug",
-    "--no-default-test-root",
   ]);
   for (let index = 0; index < args.length; index += 1) {
     const arg = args[index];
@@ -167,20 +154,13 @@ export function parseAgentRunArgs(
   const spaceId = readFlagValue(args, "--space");
   const category = readFlagValue(args, "--category");
   const inheritedFromDialog = readFlagValue(args, "--inherit-from-dialog");
-  const noDefaultTestRoot = args.includes("--no-default-test-root");
   const cwd = readFlagValue(args, "--cwd");
-  const maxToolRounds = parseNonNegativeInteger(
-    readFlagValue(args, "--max-tool-rounds") ?? readFlagValue(args, "--rounds")
-  );
   const timeoutMs = parsePositiveInteger(readFlagValue(args, "--timeout-ms"));
+  const rawEventsMode = readFlagValue(args, "--events");
+  const eventsMode = rawEventsMode === "jsonl" ? "jsonl" : undefined;
   const inheritedRef = inheritedFromDialog
     ? parseDialogReference(inheritedFromDialog)
-    : !continueDialogId && !noDefaultTestRoot
-      ? {
-          dialogKey: DEFAULT_TEST_ROOT_DIALOG_KEY,
-          dialogId: DEFAULT_TEST_ROOT_DIALOG_ID,
-        }
-      : undefined;
+    : undefined;
   const imageUrls = [
     ...readRepeatedFlagValues(args, "--image"),
     ...readRepeatedFlagValues(args, "--image-url"),
@@ -199,6 +179,7 @@ export function parseAgentRunArgs(
     imageUrls,
     allowShell: args.includes("--dangerously-allow-shell") || mimoLocalWorkspaceDefault,
     traceTools: args.includes("--trace-tools"),
+    ...(eventsMode ? { eventsMode } : {}),
     injectFeatureWorktreeInstruction: mimoLocalWorkspaceDefault,
     ...(workflowRef ? { workflowRef } : {}),
     ...(taskRowDbKey
@@ -218,14 +199,12 @@ export function parseAgentRunArgs(
         : {}),
     background: args.includes("--bg"),
     noStream: args.includes("--no-stream"),
-    noDefaultTestRoot,
     ...(continueDialogId ? { continueDialogId } : {}),
     ...(!continueDialogId && spaceId ? { spaceId } : {}),
-    ...(category ? { category } : !continueDialogId && !noDefaultTestRoot ? { category: DEFAULT_TEST_DIALOG_CATEGORY_ID } : {}),
+    ...(category ? { category } : {}),
     ...(inheritedRef?.dialogKey ? { inheritedFromDialogKey: inheritedRef.dialogKey } : {}),
     ...(inheritedRef?.dialogId ? { parentDialogId: inheritedRef.dialogId } : {}),
     ...(cwd ? { cwd } : {}),
-    ...(typeof maxToolRounds === "number" ? { maxToolRounds } : {}),
     ...(typeof timeoutMs === "number" ? { timeoutMs } : {}),
   };
 }
@@ -411,16 +390,16 @@ function buildLocalRunEnv(args: {
   env: EnvLike;
   allowShell: boolean;
 }) {
+  void args.allowShell;
   return {
     ...args.env,
-    ...(args.allowShell ? { NOLO_LOCAL_SHELL_MODE: "worktree" } : {}),
   };
 }
 
 function writeUsage(output: OutputLike) {
   output.write(
-    "Usage: nolo agent run <agent> <message> [--local|--server|--auto] [--continue <dialogId>] [--cwd <path>] [--max-tool-rounds <n>]\n" +
-      "       nolo agent run --agent <agent> (--msg <message>|--msg-file <path>) [--image <url-or-path>] [--space <spaceId>] [--category <name>] [--inherit-from-dialog <dialog>] [--task-row-dbkey <key>] [--bg] [--timeout-ms <n>] [--no-stream]\n"
+    "Usage: nolo agent run <agent> <message> [--local|--server|--auto] [--continue <dialogId>] [--cwd <path>]\n" +
+      "       nolo agent run --agent <agent> (--msg <message>|--msg-file <path>) [--image <url-or-path>] [--space <spaceId>] [--category <name>] [--inherit-from-dialog <dialog>] [--task-row-dbkey <key>] [--bg] [--timeout-ms <n>] [--events jsonl] [--no-stream]\n"
   );
 }
 
@@ -480,10 +459,9 @@ export async function runAgentRunCommand(args: string[], deps: AgentRunCommandDe
     ...(parsed.parentDialogId ? { parentDialogId: parsed.parentDialogId } : {}),
     background: parsed.background,
     noStream: parsed.noStream,
-    noDefaultTestRoot: parsed.noDefaultTestRoot,
-    ...(typeof parsed.maxToolRounds === "number" ? { maxToolRounds: parsed.maxToolRounds } : {}),
     ...(typeof parsed.timeoutMs === "number" ? { timeoutMs: parsed.timeoutMs } : {}),
     traceTools: parsed.traceTools,
+    ...(parsed.eventsMode ? { eventsMode: parsed.eventsMode } : {}),
     ...(parsed.taskRunContext ? { taskRunContext: parsed.taskRunContext } : {}),
   });
 

package/cli/agentRuntimeLocal.ts

@@ -20,6 +20,7 @@ export {
   resolveLocalToolPolicy,
   resolveOpenAiCompatibleProviderConfig,
   resolveRequestedRuntimeToolNames,
+  resolveRuntimeToolSurfaceForAgent,
   evaluateShellCommandPolicy,
   isDestructiveShellCommand,
   shouldUsePlatformChatProvider,