lark-remote 0.1.0

package/dist/config/codex-config.js

@@ -0,0 +1,428 @@
+/**
+ * Codex config loader
+ *
+ * Reads model_provider and model from ~/.codex/config.toml,
+ * provides dynamic dropdown options for /config card.
+ *
+ * Model filtering strategy:
+ * - catalog 模式（config.toml 声明 model_catalog_json）：模型列表/档位来自活动目录
+ * - 非 catalog 模式：openai 用 bundled 全量；anthropic 仅在用户显式配置时存在
+ */
+import { execFileSync } from 'node:child_process';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import TOML from '@iarna/toml';
+import { getLogger } from '../logger/index.js';
+/**
+ * Resolve codex home directory.
+ * Priority: explicit codexHome param > $CODEX_HOME env > ~/.codex
+ */
+export function resolveCodexHome(codexHome) {
+    return codexHome ?? process.env.CODEX_HOME ?? path.join(os.homedir(), '.codex');
+}
+/**
+ * Fallback lists
+ * Used when config.toml doesn't exist or cannot be parsed, and bundled command also fails
+ */
+/**
+ * 与 codex 内置 provider 注册表对齐（model-provider-info built_in_model_providers：
+ * openai/amazon-bedrock/ollama/lmstudio）。anthropic 不是内置 provider，必须用户显式
+ * 配置 [model_providers.anthropic] 才存在，fallback 阶段不得虚构。
+ */
+const FALLBACK_PROVIDERS = ['openai'];
+const FALLBACK_MODELS = [
+    'o3',
+    'o4-mini',
+    'gpt-4o',
+    'gpt-4.1',
+    'gpt-4.1-mini',
+    'gpt-4.1-nano',
+    'claude-sonnet-4-20250514',
+    'claude-opus-4-20250115',
+];
+/** Bundled models from codex binary, 1h TTL is enough (only change on binary upgrade) */
+const BUNDLED_CACHE_TTL_MS = 60 * 60 * 1000;
+/** 失败/空结果负缓存 TTL：避免每次卡片构建同步阻塞最长 8s（P3-5） */
+const NEGATIVE_CACHE_TTL_MS = 30 * 1000;
+let catalogCache = null;
+/**
+ * Read config.toml `model_catalog_json` path (tilde-expanded).
+ * 返回语义：
+ * - undefined：config.toml 不存在或**未声明**该键 → 非 catalog 模式；
+ * - ''（空串哨兵）：已声明但不可用（值为空串/非字符串）→ 按"catalog 声明退化"处理
+ *   （P2-1，绝不回退 bundled/FALLBACK）；
+ * - TOML_PARSE_FAILED_SENTINEL：config.toml 存在但解析失败 → 无法得知是否声明，
+ *   按非 catalog 处理（用户未声明 model_catalog_json 时的默认行为）；
+ * - 具体路径：声明有效（含文件缺失，文件缺失由调用方按 active 命令失败处理）。
+ */
+const TOML_PARSE_FAILED_SENTINEL = '\u0000toml-parse-failed';
+function readModelCatalogJsonPath(codexHome) {
+    const configFile = path.join(codexHome, 'config.toml');
+    if (!fs.existsSync(configFile))
+        return undefined;
+    try {
+        const parsed = TOML.parse(fs.readFileSync(configFile, 'utf-8'));
+        const raw = parsed.model_catalog_json;
+        if (raw === undefined)
+            return undefined; // 未声明 → 非 catalog
+        if (typeof raw !== 'string' || raw.length === 0)
+            return '';
+        if (raw === '~')
+            return os.homedir();
+        if (raw.startsWith('~/'))
+            return path.join(os.homedir(), raw.slice(2));
+        if (path.isAbsolute(raw))
+            return raw;
+        // 与 codex AbsolutePathBuf::resolve_path_against_base 一致：相对路径按
+        // config.toml 所在目录解析（而非进程 cwd）。
+        return path.resolve(path.dirname(configFile), raw);
+    }
+    catch {
+        // config.toml 解析失败：codex 对非法配置是硬错误；此处无法得知是否声明了
+        // model_catalog_json，按非 catalog 处理（P3-5），不冒充 catalog 声明退化
+        return TOML_PARSE_FAILED_SENTINEL;
+    }
+}
+/**
+ * Catalog mode = config.toml **声明了** model_catalog_json（键存在即算，不要求文件存在）。
+ * 依据：codex 运行时对"已配置但文件缺失/不可解析"的目录是硬错误
+ * （core/src/config/mod.rs load_model_catalog ?），绝不会回退内置目录；因此键存在即
+ * catalog 模式，文件缺失自然落入 `codex debug models` 失败路径 → 模型列表回退
+ * [currentModel]（P1-1）。
+ */
+export function isCodexCatalogMode(codexHome) {
+    const home = resolveCodexHome(codexHome);
+    const p = readModelCatalogJsonPath(home);
+    return p !== undefined && p !== TOML_PARSE_FAILED_SENTINEL;
+}
+/**
+ * 单一解析器：把 `codex debug models`（活动目录）或 `--bundled`（内置目录）的原始 JSON
+ * 输出解析为模型列表（P2-4 去重）。过滤/排序镜像 codex：
+ * - show_in_picker = visibility == List（openai_models.rs ModelPreset::from）；
+ * - API 模式（lark 用 API key 认证）下 codex filter_by_auth 过滤 supported_in_api=false；
+ * - 按 priority 升序（build_available_models）。
+ * 档位按目录声明**原样透传**（含 custom/none，P2-5），不按标准枚举过滤、不虚构兜底：
+ * 未声明档位 → 空列表（codex with_model 语义：len=0 时用声明 default，都没有则不发
+ * effort；P1）。空串档位/空串 default 被过滤（P3-9：codex ReasoningEffort::from_str
+ * 对 "" 硬错误）。supported_in_api===true 过滤等价于 codex 的 chatgpt_mode ||
+ * supported_in_api：本项目以 API-key 模式运行 codex，chatgpt_mode 恒 false（P3-3）。
+ */
+export function parseCodexModelsOutput(stdout) {
+    const parsed = JSON.parse(stdout);
+    const modelsRaw = Array.isArray(parsed?.models) ? parsed.models : [];
+    return (modelsRaw
+        .filter((m) => m && typeof m === 'object')
+        // 严格镜像 codex API 模式 filter_by_auth：show_in_picker=list 且 supported_in_api=true
+        .filter((m) => m.visibility === 'list' && m.supported_in_api === true)
+        .sort((a, b) => {
+        const pa = typeof a.priority === 'number' ? a.priority : Infinity;
+        const pb = typeof b.priority === 'number' ? b.priority : Infinity;
+        return pa - pb;
+    })
+        .map((m) => {
+        const supportedReasoningLevels = [];
+        if (Array.isArray(m.supported_reasoning_levels)) {
+            const declared = m.supported_reasoning_levels
+                .filter((r) => r !== null && typeof r === 'object' && 'effort' in r)
+                .map((r) => String(r.effort))
+                // P3-9：codex 拒绝空串档位（openai_models.rs from_str "" → Err）
+                .filter((effort) => effort.length > 0);
+            supportedReasoningLevels.push(...declared);
+        }
+        const defaultReasoningLevel = typeof m.default_reasoning_level === 'string' && m.default_reasoning_level.length > 0
+            ? m.default_reasoning_level
+            : undefined;
+        return {
+            slug: String(m.slug ?? ''),
+            displayName: String(m.display_name ?? m.slug ?? ''),
+            priority: typeof m.priority === 'number' ? m.priority : Infinity,
+            supportedReasoningLevels,
+            defaultReasoningLevel,
+        };
+    })
+        .filter((m) => m.slug.length > 0));
+}
+/**
+ * 统一目录入口：catalog 模式（model_catalog_json 已声明，含声明退化 ''）跑
+ * `debug models`（无 --bundled），否则（未声明或 config.toml 解析失败）跑
+ * `debug models --bundled`。注（P3-4）：codex 无参 `debug models` 在
+ * uses_codex_backend()/has_command_auth() 时会请求远端 /models（models-manager
+ * manager.rs / models_endpoint.rs）；本项目以 API-key 模式运行（无 ChatGPT/命令
+ * 认证），结果与 bundled 一致，故非 catalog 用 --bundled（已文档化）。
+ * 缓存：键含 binary/home/mode/models.json mtime:size + config.toml mtime:size
+ * （P3-4/P3-15/P3-5）；成功结果 TTL 1h，失败/空结果短 TTL 负缓存；stat 异常时
+ * 指纹为空（P1-1/P3-10）。
+ */
+export function getCodexCatalogModels(binary, codexHome) {
+    const home = resolveCodexHome(codexHome);
+    const catalogPath = readModelCatalogJsonPath(home);
+    const bundled = catalogPath === undefined || catalogPath === TOML_PARSE_FAILED_SENTINEL;
+    let configFingerprint = '';
+    let catalogFingerprint = '';
+    const configFile = path.join(home, 'config.toml');
+    try {
+        const st = fs.statSync(configFile);
+        configFingerprint = `${st.mtimeMs}:${st.size}`;
+    }
+    catch {
+        configFingerprint = '';
+    }
+    if (!bundled) {
+        try {
+            const st = fs.statSync(catalogPath);
+            catalogFingerprint = `${st.mtimeMs}:${st.size}`;
+        }
+        catch {
+            catalogFingerprint = '';
+        }
+    }
+    const key = `${binary}\u0000${home}\u0000${bundled ? 'bundled' : 'active'}\u0000${catalogFingerprint}\u0000${configFingerprint}`;
+    const now = Date.now();
+    if (catalogCache && catalogCache.key === key) {
+        const ttl = catalogCache.failed ? NEGATIVE_CACHE_TTL_MS : BUNDLED_CACHE_TTL_MS;
+        if (now - catalogCache.ts < ttl) {
+            return catalogCache.models;
+        }
+    }
+    try {
+        const args = bundled ? ['debug', 'models', '--bundled'] : ['debug', 'models'];
+        const stdout = execFileSync(binary, args, {
+            encoding: 'utf-8',
+            timeout: 8_000,
+            maxBuffer: 4 * 1024 * 1024,
+            env: process.env,
+            stdio: ['ignore', 'pipe', 'ignore'],
+        });
+        const models = parseCodexModelsOutput(stdout);
+        catalogCache = { key, models, ts: now, failed: models.length === 0 };
+        return models;
+    }
+    catch (err) {
+        getLogger().warn(`[codex-config] models unavailable for binary "${binary}": ${err.message}`);
+        catalogCache = { key, models: [], ts: now, failed: true };
+        return [];
+    }
+}
+/**
+ * Test utility: clear the catalog cache. Tests that exercise the bundled
+ * command path (`--bundled`) directly use `invalidateCodexBundledTestCache`
+ * from `codex-bundled-test-helpers.ts`.
+ */
+export function invalidateCodexBundledCache() {
+    catalogCache = null;
+}
+/**
+ * Get supported reasoning effort options for a model.
+ * @param codexHome 可选 codex 配置目录覆盖（与 loadCodexConfig 的 codexHome 对齐，
+ *   与 loadCodexConfig 的 codexHome 对齐）；省略时走 $CODEX_HOME → ~/.codex。
+ */
+export function getReasoningEffortOptions(model, binary = 'codex', codexHome) {
+    const models = getCodexCatalogModels(binary, codexHome);
+    const found = models.find((m) => m.slug === model);
+    // 未命中（含未知模型，codex fallback 元数据 supported 为空）→ []，不虚构档位（P1）
+    return found?.supportedReasoningLevels ?? [];
+}
+/**
+ * Get default reasoning effort for a model.
+ * 未声明/未知模型 → undefined（codex 语义：ModelPreset.default_reasoning_effort
+ * unwrap_or(ReasoningEffort::None)，即不传 effort；P3-12）。
+ */
+export function getDefaultReasoningEffort(model, binary = 'codex', codexHome) {
+    const models = getCodexCatalogModels(binary, codexHome);
+    return models.find((m) => m.slug === model)?.defaultReasoningLevel;
+}
+/**
+ * Extract custom provider name from model_providers config.
+ * Skips built-in openai/anthropic, returns first custom provider.
+ */
+function extractCustomProvider(providers) {
+    const keys = Object.keys(providers);
+    return keys.find((k) => k !== 'openai' && k !== 'anthropic');
+}
+/**
+ * Read codex config.toml, extract provider and model info
+ *
+ * Model filtering strategy:
+ * - Custom provider: model from config.toml (single model)
+ * - openai: all bundled models
+ * - anthropic: bundled models with 'claude-*' prefix
+ */
+export function loadCodexConfig(opts = {}) {
+    const binary = opts.binary ?? 'codex';
+    const codexHome = resolveCodexHome(opts.codexHome);
+    const configFile = path.join(codexHome, 'config.toml');
+    const catalogMode = isCodexCatalogMode(codexHome);
+    // 统一目录来源：catalog 模式 = 活动目录（model_catalog_json 整体替换内置目录）；
+    // 否则 = bundled（与原行为一致）
+    const catalogSlugs = getCodexCatalogModels(binary, codexHome).map((m) => m.slug);
+    const allModels = catalogSlugs.length > 0 ? catalogSlugs : FALLBACK_MODELS;
+    const anthropicModels = allModels.filter((m) => m.startsWith('claude-'));
+    if (!fs.existsSync(configFile)) {
+        // 无 config.toml：默认模型取目录首个可用模型（codex default_model_from_available），
+        // 目录不可用时回退 'o3'（FALLBACK_MODELS 首项，非 catalog 默认）
+        const currentModel = allModels[0] ?? 'o3';
+        // No config file: only openai has bundled models
+        const providerModels = {
+            openai: allModels,
+        };
+        return {
+            currentModel,
+            currentProvider: FALLBACK_PROVIDERS[0],
+            providerNames: FALLBACK_PROVIDERS,
+            providerEnvKeys: { openai: 'OPENAI_API_KEY' },
+            providerModels,
+            modelOptions: (provider) => buildModelOptions(currentModel, FALLBACK_PROVIDERS[0], provider, providerModels),
+        };
+    }
+    try {
+        const raw = fs.readFileSync(configFile, 'utf-8');
+        const parsed = TOML.parse(raw);
+        const modelKeyRaw = typeof parsed.model === 'string' && parsed.model.length > 0 ? parsed.model : undefined;
+        const currentProvider = String(parsed.model_provider ?? FALLBACK_PROVIDERS[0]);
+        // Extract model_providers section
+        const modelProviders = (parsed.model_providers ?? {});
+        const rawProviderNames = Object.keys(modelProviders);
+        let providerNames;
+        let providerEnvKeys;
+        let providerModels;
+        if (catalogMode) {
+            // catalog 模式：provider 列表 = 内置 openai + [model_providers.*]（对齐 codex
+            // merge_configured_model_providers(built_in_model_providers, cfg.model_providers)，
+            // core/src/config/mod.rs:3655；model_provider 未声明时默认 "openai"，mod.rs:3659）。
+            // anthropic 不是 codex 内置 provider，未显式配置时不得出现。模型列表仍只来自
+            // 活动目录——openai+内置 gpt-5.x 的失效路径不因 provider 合并复活。
+            providerNames = rawProviderNames.includes('openai')
+                ? rawProviderNames
+                : ['openai', ...rawProviderNames];
+            providerEnvKeys = { openai: 'OPENAI_API_KEY' };
+            for (const [name, config] of Object.entries(modelProviders)) {
+                if (typeof config.env_key === 'string') {
+                    providerEnvKeys[name] = config.env_key;
+                }
+            }
+            // 活动目录全局：codex 运行时 provider 与模型无绑定（StaticModelsManager），
+            // 每个 provider 下拉都展示目录全部 slug
+            // 活动目录不可用（命令失败/空）时回退 [currentModel]——无目录元数据时档位/能力
+            // 不可信；禁止泄漏 FALLBACK_MODELS/内置目录。model 键缺失时不得虚构 'o3'（P1-2）。
+            const catalogCurrentModel = modelKeyRaw ?? catalogSlugs[0] ?? '';
+            const providerModelList = catalogSlugs.length > 0 ? catalogSlugs : modelKeyRaw ? [modelKeyRaw] : [];
+            providerModels = {};
+            for (const name of providerNames) {
+                providerModels[name] = providerModelList;
+            }
+            const resolvedProvider = typeof parsed.model_provider === 'string' && parsed.model_provider.length > 0
+                ? parsed.model_provider
+                : 'openai';
+            return {
+                currentModel: catalogCurrentModel,
+                currentProvider: resolvedProvider,
+                providerNames,
+                providerEnvKeys,
+                providerModels,
+                modelOptions: (provider) => buildModelOptions(catalogCurrentModel, resolvedProvider, provider, providerModels),
+            };
+        }
+        // fallback 模式：保持原行为（openai/anthropic 内置 + bundled 模型）
+        const currentModel = modelKeyRaw ?? catalogSlugs[0] ?? 'o3';
+        providerNames = rawProviderNames;
+        if (!providerNames.includes('openai')) {
+            providerNames.unshift('openai');
+        }
+        const customProvider = extractCustomProvider(modelProviders);
+        providerEnvKeys = {};
+        providerEnvKeys['openai'] = 'OPENAI_API_KEY';
+        for (const [name, config] of Object.entries(modelProviders)) {
+            if (typeof config.env_key === 'string') {
+                providerEnvKeys[name] = config.env_key;
+            }
+        }
+        // anthropic 显式配置但未写 env_key 时保留默认键（P3-9：恢复旧版无条件映射的
+        // 配置场景行为；未配置时不再虚构）
+        if (rawProviderNames.includes('anthropic') &&
+            !modelProviders['anthropic']?.env_key &&
+            providerEnvKeys['anthropic'] === undefined) {
+            providerEnvKeys['anthropic'] = 'ANTHROPIC_API_KEY';
+        }
+        providerModels = { openai: allModels };
+        // anthropic 仅在用户显式配置时才存在（非 codex 内置 provider）
+        if (rawProviderNames.includes('anthropic')) {
+            providerModels['anthropic'] = anthropicModels;
+        }
+        if (customProvider) {
+            // Custom provider: use the model from config.toml
+            providerModels[customProvider] = [currentModel];
+        }
+        return {
+            currentModel,
+            currentProvider,
+            providerNames,
+            providerEnvKeys,
+            providerModels,
+            modelOptions: (provider) => buildModelOptions(currentModel, currentProvider, provider, providerModels),
+        };
+    }
+    catch {
+        // config.toml 解析失败（readModelCatalogJsonPath 已把此场景标记为 catalog 声明退化 ''）：
+        // catalogMode 为 true 时绝不回退 bundled/FALLBACK（P2-1），返回空 provider/模型列表；
+        // 非 catalog 才走 FALLBACK 兜底。
+        if (catalogMode) {
+            return {
+                currentModel: '',
+                currentProvider: '',
+                providerNames: [],
+                providerEnvKeys: {},
+                providerModels: {},
+                modelOptions: () => [],
+            };
+        }
+        const currentModel = catalogSlugs[0] ?? 'o3';
+        const providerModels = {
+            openai: allModels,
+        };
+        return {
+            currentModel,
+            currentProvider: FALLBACK_PROVIDERS[0],
+            providerNames: FALLBACK_PROVIDERS,
+            providerEnvKeys: { openai: 'OPENAI_API_KEY' },
+            providerModels,
+            modelOptions: (provider) => buildModelOptions(currentModel, FALLBACK_PROVIDERS[0], provider, providerModels),
+        };
+    }
+}
+/**
+ * Build model options for a provider.
+ * - If provider specified: return that provider's models.
+ *   Only prepend currentModel if it belongs to currentProvider (same provider filter).
+ *   Cross-provider currentModel is NOT leaked into another provider's list.
+ * - If no provider: return all models from all providers (with currentModel prepended if not in any list)
+ */
+function buildModelOptions(currentModel, currentProvider, provider, providerModels) {
+    // 不为空 currentModel 虚构默认值（catalog 退化场景由上层显式传入，P1-2）；
+    // 空值时直接返回原始列表
+    const cm = currentModel;
+    if (provider && providerModels[provider]) {
+        // Return models for the specified provider
+        const models = providerModels[provider];
+        // Only prepend currentModel if it belongs to this provider (currentProvider matches)
+        // Cross-provider currentModel must NOT leak into another provider's list
+        if (cm && !models.includes(cm) && currentProvider === provider) {
+            return [cm, ...models];
+        }
+        return models;
+    }
+    // No provider specified: return all unique models
+    const allModels = Object.values(providerModels).flat();
+    const unique = [...new Set(allModels)];
+    // Ensure currentModel is always first (even if bundled failed)
+    if (cm && !unique.includes(cm)) {
+        return [cm, ...unique];
+    }
+    // If currentModel already in list but not first, move it to front
+    if (cm && unique[0] !== cm) {
+        const filtered = unique.filter((m) => m !== cm);
+        return [cm, ...filtered];
+    }
+    return unique;
+}
+//# sourceMappingURL=codex-config.js.map

package/dist/config/dir.d.ts

@@ -0,0 +1,17 @@
+interface CliArgs {
+    configDir?: string;
+    settings?: string;
+    help?: boolean;
+    version?: boolean;
+}
+export declare function resolveConfigDir(configDirArg: string | undefined): string;
+export declare function getConfigDir(): string;
+export declare function setConfigDir(dir: string): void;
+export declare function parseCliArgs(args?: string[]): CliArgs;
+/** Read the project version from package.json (single source of truth). */
+export declare function getVersion(): string;
+/** Print version to stdout. */
+export declare function printVersion(): void;
+/** Print CLI help to stdout. */
+export declare function printHelp(): void;
+export {};

package/dist/config/dir.js

@@ -0,0 +1,81 @@
+import path from 'node:path';
+import os from 'node:os';
+import { readFileSync } from 'node:fs';
+import { fileURLToPath } from 'node:url';
+const DEFAULT_DIR = '.lark-remote';
+let cachedConfigDir = null;
+export function resolveConfigDir(configDirArg) {
+    if (configDirArg) {
+        let p = configDirArg;
+        if (p === '~')
+            p = os.homedir();
+        else if (p.startsWith('~/'))
+            p = path.join(os.homedir(), p.slice(2));
+        return path.resolve(p);
+    }
+    return path.join(os.homedir(), DEFAULT_DIR);
+}
+export function getConfigDir() {
+    if (!cachedConfigDir) {
+        cachedConfigDir = path.join(os.homedir(), DEFAULT_DIR);
+    }
+    return cachedConfigDir;
+}
+export function setConfigDir(dir) {
+    cachedConfigDir = dir;
+}
+export function parseCliArgs(args = process.argv.slice(2)) {
+    const result = {};
+    for (let i = 0; i < args.length; i++) {
+        const arg = args[i];
+        if (arg === '-h' || arg === '--help') {
+            result.help = true;
+        }
+        else if (arg === '-v' || arg === '--version') {
+            result.version = true;
+        }
+        else if (arg === '--config-dir' && i + 1 < args.length) {
+            // P1-20: peek, don't consume — when the next arg is itself a flag,
+            // roll back so it continues parsing (previously --settings/--help
+            // right after --config-dir were silently swallowed).
+            const nextArg = args[i + 1];
+            if (nextArg && !nextArg.startsWith('--')) {
+                result.configDir = args[++i];
+            }
+        }
+        else if (arg === '--settings' && i + 1 < args.length) {
+            const nextArg = args[i + 1];
+            if (nextArg && !nextArg.startsWith('--')) {
+                result.settings = args[++i];
+            }
+        }
+    }
+    return result;
+}
+/** Read the project version from package.json (single source of truth). */
+export function getVersion() {
+    const pkgPath = path.join(path.dirname(fileURLToPath(import.meta.url)), '..', '..', 'package.json');
+    const pkg = JSON.parse(readFileSync(pkgPath, 'utf8'));
+    return pkg.version ?? '0.0.0';
+}
+/** Print version to stdout. */
+export function printVersion() {
+    process.stdout.write(`lark-remote ${getVersion()}\n`);
+}
+/** Print CLI help to stdout. */
+export function printHelp() {
+    const lines = [
+        'lark-remote — 飞书私聊 ↔ 本地 Coding Agent CLI 桥接',
+        '',
+        'Usage:',
+        '  lark-remote [options]',
+        '',
+        'Options:',
+        '  --config-dir <path>   自定义配置目录（默认 ~/.lark-remote，可用于同机多实例）',
+        '  --settings <path>     指定 Claude 配置文件路径',
+        '  -h, --help            显示本帮助信息',
+        '  -v, --version         显示版本号',
+    ];
+    process.stdout.write(lines.join('\n') + '\n');
+}
+//# sourceMappingURL=dir.js.map