@xiaoqiong0v0/opencode-file-tool 1.0.5 → 1.1.0

package/dist/file-tool.js

@@ -1,222 +1,16 @@
-import { tool } from "@opencode-ai/plugin";
-import { existsSync, readFileSync, writeFileSync, mkdirSync, rmSync } from "node:fs";
-import createLogger from "@xiaoqiong0v0/opencode-plugin-logger";
-import { rm } from "node:fs/promises";
-import { join } from "node:path";
-const CONFIG_DIR = process.env.HOME || process.env.USERPROFILE || "";
-const CONFIG_PATH = join(CONFIG_DIR, ".config/opencode/file-tool.jsonc");
-const OPENCODE_CONFIG = join(CONFIG_DIR, ".config/opencode/opencode.json");
-const CACHE_DIR = join(CONFIG_DIR, ".opencode/plugins-cache", "file-tool");
-const log = createLogger("file-tool");
-let _cfg = null;
-const FILE_TOOL_CFG_SAMPLE = `{
-  // 视觉分析模型（provider/modelId），file_tool set-provider 切换
-  "model": "",
-  "maxTokens": 4096,
-  "timeout": 60000,
-  "maxFileSizeMB": 20,
-  // 缓存消息数量上限，超过则删除最早的
-  "maxCacheMessages": 3,
-  // 工具提示语言：zh=中文, en=English
-  "lang": "en"
-}
-`;
-let MAX_CACHE_MSGS = 3;
-let LANG = "en";
-const TX = {
-    file_not_found: { zh: "文件不存在: {path}", en: "File not found: {path}" },
-    file_id_not_found: { zh: "文件ID不存在: {id}", en: "File ID not found: {id}" },
-    file_data_not_found: { zh: "文件数据不存在: {id}", en: "File data not found: {id}" },
-    not_an_image: { zh: "不是图片文件: {name} ({mime})", en: "Not an image: {name} ({mime})" },
-    unsupported_source: { zh: "不支持的图片来源: {source}", en: "Unsupported source: {source}" },
-    describe_image: { zh: "请详细描述这张图片（{name}）的内容", en: "Describe this image ({name})" },
-    current_model: { zh: "当前模型: {model}\n可用模型:\n{list}", en: "Current model: {model}\nAvailable models:\n{list}" },
-    model_not_set: { zh: "未设置", en: "not set" },
-    model_switched: { zh: "视觉模型已切换为: {model}", en: "Vision model set to: {model}" },
-    specify_model: { zh: "请指定模型名", en: "Specify a model name" },
-    unknown_cmd: { zh: "未知命令: {cmd}\n可用: list-provider, set-provider <model>, list-cache [all|N|main|main N]", en: "Unknown command: {cmd}\nAvailable: list-provider, set-provider <model>, list-cache [all|N|main|main N]" },
-    config_error: { zh: "请在 file-tool.jsonc 中配置 model (provider/modelId) 或 apiKey+apiBaseUrl+model", en: "Set model (provider/modelId) or apiKey+apiBaseUrl+model in file-tool.jsonc" },
-    meta_failed: { zh: "分析失败", en: "Failed" },
-    meta_skip: { zh: "跳过", en: "Skip" },
-    meta_not_found: { zh: "文件不存在", en: "Not found" },
-    meta_image: { zh: "图片", en: "Image" },
-    meta_error: { zh: "分析出错", en: "Error" },
-    no_cache: { zh: "[] (无缓存)", en: "[] (no cache)" },
-    vision_prompt_default: { zh: "请详细描述这张图片的内容，返回格式: [文件名] 描述", en: "Describe this image in detail, format: [filename] description" },
-    err_resolve_config: { zh: "无法解析模型配置: {model}。请在 file-tool.jsonc 中配置 model (provider/modelId) 或 apiKey+apiBaseUrl+model", en: "Cannot resolve model config: {model}. Set model (provider/modelId) or apiKey+apiBaseUrl+model in file-tool.jsonc" },
-    err_api: { zh: "API {status}: {msg}", en: "API {status}: {msg}" },
-    empty_response: { zh: "(空)", en: "(empty)" },
-    cmd_desc: { zh: "切换视觉分析模型", en: "Switch vision analysis model" },
-    cmd_template: { zh: "直接调用 file_tool 工具。默认 `list-provider`，`set-provider <模型名>` 切换模型，`list-cache` 查看缓存。", en: "Call file_tool tool directly. Default: `list-provider`. Use `set-provider <model>` to switch. Use `list-cache` to view cached files." },
-};
-const T = (key, params) => {
-    const entry = TX[key] || { zh: key, en: key };
-    const t = LANG === "zh" ? entry.zh : entry.en;
-    if (!params)
-        return t;
-    return Object.entries(params).reduce((s, [k, v]) => s.replace(`{${k}}`, v), t);
-};
-function loadCfg() {
-    if (!existsSync(CONFIG_PATH)) {
-        try {
-            writeFileSync(CONFIG_PATH, FILE_TOOL_CFG_SAMPLE, "utf-8");
-        }
-        catch { }
-    }
-    const raw = existsSync(CONFIG_PATH) ? readJsonc(CONFIG_PATH) : {};
-    _cfg = resolveConfig(raw);
-    MAX_CACHE_MSGS = (raw.maxCacheMessages > 0) ? raw.maxCacheMessages : 3;
-    LANG = (raw.lang === "zh" ? "zh" : "en");
-    return _cfg;
-}
-function reloadCfg() { loadCfg(); }
+import { T } from "./i18n.js";
+import { log, ENABLED, loadCfg } from "./config.js";
+import { registerInputFile, migrateLegacyCache, knownSessions, sessionParents, deleteSession, removeMsgCache, extForMime } from "./cache.js";
+import { analyzeImageTool } from "./tools/analyze.js";
+import { textToImageTool, textToVideoTool, textToSpeechTool } from "./tools/generate.js";
+import { fileTool } from "./tools/file_tool.js";
 try {
     loadCfg();
 }
 catch (e) {
     log.error("初始化失败", e instanceof Error ? e : Error(String(e)));
 }
-const DESC = {
-    analyze_image: {
-        zh: "用多模态模型分析图片。先调 file_tool list-cache 拿到文件ID，再用 file_id:N 分析。",
-        en: "Analyze images with multimodal model. Call file_tool list-cache first to get file IDs, then use file_id:N.",
-    },
-    file_tool: {
-        zh: "文件缓存管理。当你在上下文中看到 [Image N] 或收到 Cannot read 图片错误时，立即调 list-cache 获取文件ID，再用 analyze_image file_id:N 分析。",
-        en: "File cache manager. When you see [Image N] or a Cannot read image error, call list-cache to get file IDs, then use analyze_image file_id:N.",
-    },
-    file_tool_args: {
-        zh: "list-cache, list-cache main, list-provider, set-provider <model>",
-        en: "list-cache, list-cache main, list-provider, set-provider <model>",
-    },
-    analyze_args_source: { zh: "file_path=file_id:N", en: "file_path=file_id:N" },
-    analyze_args_data: { zh: "file_id:N 或 base64", en: "file_id:N or base64" },
-    analyze_args_prompt: { zh: "分析提示", en: "prompt" },
-};
-function getCfg() {
-    if (_cfg)
-        return _cfg;
-    const raw = existsSync(CONFIG_PATH) ? readJsonc(CONFIG_PATH) : {};
-    _cfg = resolveConfig(raw);
-    return _cfg;
-}
-function resolveConfig(fileConfig) {
-    const model = fileConfig.model;
-    if (!model)
-        throw new Error(T("config_error"));
-    if (fileConfig.apiKey && fileConfig.apiBaseUrl) {
-        const mId = model.includes("/") ? model.split("/").pop() : model;
-        return { model, apiKey: fileConfig.apiKey, baseURL: fileConfig.apiBaseUrl, modelId: mId, maxTokens: fileConfig.maxTokens || 4096, timeout: fileConfig.timeout || 60000 };
-    }
-    if (model.includes("/")) {
-        const [provider, modelId] = model.split("/");
-        try {
-            const raw = readFileSync(OPENCODE_CONFIG, "utf-8");
-            const oc = JSON.parse(raw);
-            const prov = oc.provider?.[provider];
-            if (prov?.options?.apiKey && prov?.options?.baseURL)
-                return { model, apiKey: prov.options.apiKey, baseURL: prov.options.baseURL, modelId, maxTokens: fileConfig.maxTokens || 4096, timeout: fileConfig.timeout || 60000 };
-        }
-        catch { }
-    }
-    throw new Error(T("err_resolve_config", { model }));
-}
-function readJsonc(path) {
-    const raw = readFileSync(path, "utf-8").replace(/\/\/.*$/gm, "").replace(/\/\*[\s\S]*?\*\//g, "");
-    return JSON.parse(raw);
-}
-async function callVisionApi(imageUrl, prompt) {
-    const cfg = getCfg();
-    const resp = await fetch(`${cfg.baseURL}/chat/completions`, {
-        method: "POST",
-        headers: { Authorization: `Bearer ${cfg.apiKey}`, "Content-Type": "application/json" },
-        body: JSON.stringify({ model: cfg.modelId, messages: [{ role: "user", content: [{ type: "text", text: prompt || T("vision_prompt_default") }, { type: "image_url", image_url: { url: imageUrl } }] }], max_tokens: cfg.maxTokens }),
-        signal: AbortSignal.timeout(cfg.timeout || 60000),
-    });
-    if (!resp.ok)
-        throw new Error(T("err_api", { status: String(resp.status), msg: (await resp.text().catch(() => "unknown")).slice(0, 200) }));
-    const data = await resp.json();
-    const msg = data.choices?.[0]?.message;
-    return msg?.content || msg?.reasoning_content || T("empty_response");
-}
-const sessionParents = new Map();
-const knownSessions = new Set();
-function getRootSession(sid) {
-    let current = sid;
-    while (sessionParents.has(current))
-        current = sessionParents.get(current);
-    return current;
-}
-function findFileInChain(sid, fid) {
-    const store = readSession(sid);
-    const file = store.files[fid];
-    if (file)
-        return { store, file };
-    const parentSid = sessionParents.get(sid);
-    if (parentSid)
-        return findFileInChain(parentSid, fid);
-    return null;
-}
-function sessionDir(sid) { return join(CACHE_DIR, sid); }
-function filesDir(sid) {
-    const d = join(sessionDir(sid), "files");
-    if (!existsSync(d))
-        mkdirSync(d, { recursive: true });
-    return d;
-}
-function readSession(sid) {
-    try {
-        return JSON.parse(readFileSync(join(sessionDir(sid), "files.json"), "utf-8"));
-    }
-    catch {
-        return { nextId: 1, files: {}, messages: [] };
-    }
-}
-function writeSession(sid, data) {
-    const dir = sessionDir(sid);
-    if (!existsSync(dir))
-        mkdirSync(dir, { recursive: true });
-    const msgs = data.messages || [];
-    if (msgs.length > MAX_CACHE_MSGS) {
-        const expired = msgs.splice(0, msgs.length - MAX_CACHE_MSGS);
-        for (const msg of expired) {
-            for (const fid of (msg.fileIds || [])) {
-                delete data.files[fid];
-                const path = join(dir, "files", fid + ".b64");
-                rm(path, { force: true }).then(() => {
-                    log.info(`${sid}: Deleted file ${path}`);
-                }).catch((err) => {
-                    log.error(`${sid}: Failed to delete file ${path}`, err);
-                });
-            }
-        }
-    }
-    writeFileSync(join(dir, "files.json"), JSON.stringify(data, null, 2));
-}
-function writeFileData(sid, fid, url) {
-    const b64 = url.replace(/^data:\w+\/\w+;base64,/, "");
-    writeFileSync(join(filesDir(sid), fid + ".b64"), b64, "utf-8");
-}
-function readFileData(sid, fid) {
-    try {
-        const b64 = readFileSync(join(filesDir(sid), fid + ".b64"), "utf-8");
-        const meta = readSession(sid).files[fid];
-        return `data:${meta?.mime || "image/png"};base64,${b64}`;
-    }
-    catch {
-        const parentSid = sessionParents.get(sid);
-        if (parentSid)
-            return readFileData(parentSid, fid);
-        return null;
-    }
-}
-function deleteSession(sid) {
-    const dir = sessionDir(sid);
-    if (existsSync(dir))
-        rmSync(dir, { recursive: true, force: true });
-}
-// V1 export：工具 + 事件
-export const FileTool = async () => {
+export const fileToolPlugin = async () => {
     log.loaded();
     return {
         config: async (config) => {
@@ -231,6 +25,7 @@ export const FileTool = async () => {
                 knownSessions.add(sid);
                 if (props?.parentID)
                     sessionParents.set(sid, props.parentID);
+                migrateLegacyCache(sid);
             }
             if (event.type === "session.updated" && sid) {
                 if (!knownSessions.has(sid))
@@ -249,158 +44,39 @@ export const FileTool = async () => {
                 const part = props?.part;
                 if (part?.type === "file" && (part?.mime || "").startsWith("image/")) {
                     const fn = (part.filename || part.name || "");
-                    if (fn) {
-                        const data = readSession(sid || "");
-                        if (!sid)
-                            return;
-                        const fid = data.nextId++;
+                    if (fn && sid) {
                         const msgId = (part.messageID || "");
-                        data.files[fid] = { id: fid, filename: fn, mime: part.mime, msgId };
-                        writeFileData(sid, fid, (part.url || ""));
-                        const msgs = data.messages;
-                        const last = msgs[msgs.length - 1];
-                        if (last && last.msgId === msgId) {
-                            last.fileIds.push(fid);
+                        if (ENABLED) {
+                            const ext = extForMime(part.mime || "image/png");
+                            const r = registerInputFile(sid, `input_${Date.now() % 100000}.${ext}`, part.mime || "image/png", msgId, (part.url || ""));
+                            log.info(`${sid}: cached ${fn} as input:${r.id}`);
                         }
                         else {
-                            msgs.push({ msgId, fileIds: [fid] });
+                            log.info(`${sid}: skip cache ${fn} (disabled)`);
                         }
-                        writeSession(sid, data);
                     }
                 }
             }
+            if (event.type === "message.removed" && sid) {
+                const msgId = props?.messageID || "";
+                log.info(`${sid}: message.removed msgId=${msgId.slice(-8) || "(empty)"}`);
+                if (msgId)
+                    removeMsgCache(sid, msgId);
+            }
+            if (event.type === "message.part.removed" && sid) {
+                const msgId = props?.messageID || props?.info?.id || "";
+                log.info(`${sid}: message.part.removed msgId=${msgId.slice(-8) || "(empty)"}`);
+                if (msgId)
+                    removeMsgCache(sid, msgId);
+            }
         },
         tool: {
-            analyze_image: tool({
-                description: DESC.analyze_image[LANG],
-                args: {
-                    source: tool.schema.enum(["file_path", "base64"]).describe(DESC.analyze_args_source[LANG]),
-                    data: tool.schema.string().describe(DESC.analyze_args_data[LANG]),
-                    prompt: tool.schema.string().optional().describe(DESC.analyze_args_prompt[LANG]),
-                },
-                execute: async ({ source, data, prompt }, context) => {
-                    let imageUrl, fileName = "";
-                    if (source === "file_path" && data.startsWith("file_id:")) {
-                        const fid = parseInt(data.slice(8), 10);
-                        const found = findFileInChain(context.sessionID, fid);
-                        if (!found) {
-                            context.metadata?.({ title: T("meta_failed") });
-                            return T("file_id_not_found", { id: String(fid) });
-                        }
-                        const file = found.file;
-                        if (!file.mime.startsWith("image/")) {
-                            context.metadata?.({ title: T("meta_skip") });
-                            return T("not_an_image", { name: file.filename, mime: file.mime });
-                        }
-                        fileName = file.filename;
-                        imageUrl = readFileData(context.sessionID, fid) || "";
-                        if (!imageUrl) {
-                            context.metadata?.({ title: T("meta_failed") });
-                            return T("file_data_not_found", { id: String(fid) });
-                        }
-                        prompt = prompt || T("describe_image", { name: fileName });
-                    }
-                    else if (source === "file_path") {
-                        if (!existsSync(data)) {
-                            const tryPath = join(context.directory, data);
-                            if (existsSync(tryPath))
-                                data = tryPath;
-                        }
-                        if (!existsSync(data)) {
-                            context.metadata?.({ title: T("meta_not_found") });
-                            return T("file_not_found", { path: data });
-                        }
-                        const ext = data.split(".").pop()?.toLowerCase() || "";
-                        const mimeMap = { png: "image/png", jpg: "image/jpeg", jpeg: "image/jpeg", bmp: "image/bmp", gif: "image/gif", webp: "image/webp" };
-                        const mime = mimeMap[ext] || "image/png";
-                        fileName = data.split(/[/\\]/).pop() || "";
-                        imageUrl = `data:${mime};base64,${readFileSync(data).toString("base64")}`;
-                    }
-                    else if (source === "base64") {
-                        imageUrl = `data:image/png;base64,${data.replace(/^data:image\/\w+;base64,/, "")}`;
-                    }
-                    else {
-                        return T("unsupported_source", { source });
-                    }
-                    try {
-                        const result = await callVisionApi(imageUrl, prompt || "");
-                        context.metadata?.({ title: `[Vision] ${fileName || T("meta_image")}`, metadata: { sessionID: context.sessionID, messageID: context.messageID } });
-                        return "[Vision] " + result;
-                    }
-                    catch (e) {
-                        const msg = e instanceof Error ? e.message : String(e);
-                        context.metadata?.({ title: T("meta_error") });
-                        return `[Vision Error] ${msg}`;
-                    }
-                },
-            }),
-            file_tool: tool({
-                description: DESC.file_tool[LANG],
-                args: { command: tool.schema.string().describe(DESC.file_tool_args[LANG]) },
-                execute: async ({ command }, context) => {
-                    const cmd = command.trim();
-                    if (cmd === "list-provider") {
-                        const cfg = existsSync(CONFIG_PATH) ? readJsonc(CONFIG_PATH) : {};
-                        const models = [];
-                        const oc = JSON.parse(readFileSync(OPENCODE_CONFIG, "utf-8"));
-                        for (const [pName, pVal] of Object.entries(oc.provider || {}))
-                            for (const mId of Object.keys(pVal.models || {}))
-                                models.push(`${pName}/${mId}`);
-                        return T("current_model", {
-                            model: cfg.model || T("model_not_set"),
-                            list: models.map(m => "  " + m).join("\n"),
-                        });
-                    }
-                    if (cmd.startsWith("set-provider ")) {
-                        const model = cmd.slice(13).trim();
-                        if (!model)
-                            return T("specify_model");
-                        const cfg = existsSync(CONFIG_PATH) ? readJsonc(CONFIG_PATH) : {};
-                        cfg.model = model;
-                        delete cfg.apiKey;
-                        delete cfg.apiBaseUrl;
-                        writeFileSync(CONFIG_PATH, JSON.stringify(cfg, null, 2));
-                        reloadCfg();
-                        return T("model_switched", { model });
-                    }
-                    if (cmd === "list-cache" || cmd.startsWith("list-cache ")) {
-                        const arg = cmd === "list-cache" ? "1" : cmd.slice(11).trim();
-                        let targetSid = context.sessionID;
-                        let limit = arg;
-                        if (arg === "main") {
-                            targetSid = getRootSession(context.sessionID);
-                            limit = "1";
-                        }
-                        if (arg.startsWith("main ")) {
-                            targetSid = getRootSession(context.sessionID);
-                            limit = arg.slice(5).trim();
-                        }
-                        const data = readSession(targetSid);
-                        const msgs = data.messages || [];
-                        if (msgs.length === 0)
-                            return `${targetSid}: ${T("no_cache")}`;
-                        let count = msgs.length;
-                        if (limit !== "all") {
-                            const n = parseInt(limit, 10);
-                            if (!isNaN(n) && n > 0)
-                                count = Math.min(n, count);
-                        }
-                        const show = msgs.slice(-count);
-                        let out = `${targetSid}:\n`;
-                        for (const msg of show) {
-                            out += `  msg_${msg.msgId.slice(-8)}:\n`;
-                            for (const fid of msg.fileIds) {
-                                const f = data.files[fid];
-                                if (f)
-                                    out += `    ${f.filename}: ${f.id}\n`;
-                            }
-                        }
-                        return out.trim();
-                    }
-                    return T("unknown_cmd", { cmd });
-                },
-            }),
+            analyze_image: analyzeImageTool,
+            text_to_image: textToImageTool,
+            text_to_video: textToVideoTool,
+            text_to_speech: textToSpeechTool,
+            file_tool: fileTool,
         },
     };
 };
-export { FileTool as default };
+export default fileToolPlugin;

package/dist/i18n.js

@@ -0,0 +1,67 @@
+export let LANG = "en";
+export function setLang(lang) { LANG = lang === "zh" ? "zh" : "en"; }
+export const TX = {
+    file_not_found: { zh: "文件不存在: {path}", en: "File not found: {path}" },
+    file_id_not_found: { zh: "文件ID不存在: {id}", en: "File ID not found: {id}" },
+    file_data_not_found: { zh: "文件数据不存在: {id}", en: "File data not found: {id}" },
+    not_an_image: { zh: "不是图片文件: {name} ({mime})", en: "Not an image: {name} ({mime})" },
+    unsupported_source: { zh: "不支持的图片来源: {source}", en: "Unsupported source: {source}" },
+    describe_image: { zh: "请详细描述这张图片（{name}）的内容", en: "Describe this image ({name})" },
+    current_model: { zh: "当前模型:\n{model}\n可用模型:\n{list}", en: "Current models:\n{model}\nAvailable models:\n{list}" },
+    model_not_set: { zh: "未设置", en: "not set" },
+    model_switched: { zh: "{type} 模型已切换为: {model}", en: "{type} model set to: {model}" },
+    model_types: { zh: "vision: 图像分析, image: 文生图, video: 文生视频, tts: 文生语音", en: "vision: analyze image, image: text-to-image, video: text-to-video, tts: text-to-speech" },
+    model_usage: { zh: "用法: set-provider <模型名>（vision，默认）或 set-provider <类型>:<模型名>，如 set-provider image:xxx/yyy", en: "Usage: set-provider <model> (vision, default) or set-provider <type>:<model>, e.g. set-provider image:xxx/yyy" },
+    specify_model: { zh: "请指定模型名", en: "Specify a model name" },
+    model_not_configured: { zh: "{type} 模型未配置，请在 file-tool.jsonc 的 models 中设置（可用 set-provider 切换）", en: "{type} model not configured, set it in models of file-tool.jsonc (use set-provider)" },
+    image_generated: { zh: "图片已生成并缓存: {name} (file_id:{id})\n路径: {path}\n用 analyze_image file_id:{id} 查看", en: "Image generated & cached: {name} (file_id:{id})\nPath: {path}\nuse analyze_image file_id:{id}" },
+    video_generated: { zh: "视频已生成并缓存: {name} (file_id:{id})\n路径: {path}", en: "Video generated & cached: {name} (file_id:{id})\nPath: {path}" },
+    video_pending: { zh: "视频生成中（任务 {id}），已提交，稍后可查", en: "Video generating (task {id}), submitted, check later" },
+    video_failed: { zh: "视频生成失败: {msg}", en: "Video generation failed: {msg}" },
+    tts_generated: { zh: "语音已生成并缓存: {name} (file_id:{id})\n路径: {path}", en: "Audio generated & cached: {name} (file_id:{id})\nPath: {path}" },
+    gen_failed: { zh: "生成失败: {msg}", en: "Generation failed: {msg}" },
+    gen_no_result: { zh: "接口未返回生成结果", en: "No generation result from API" },
+    gen_download_failed: { zh: "下载生成结果失败: {url}", en: "Failed to download result: {url}" },
+    unknown_cmd: { zh: "未知命令: {cmd}\n可用: list-provider, set-provider [<类型>:]<模型名>, list-cache [类型] [数量], enable, disable, enable-save, disable-save, status", en: "Unknown command: {cmd}\nAvailable: list-provider, set-provider [<type>:]<model>, list-cache [type] [count], enable, disable, enable-save, disable-save, status" },
+    config_error: { zh: "请在 file-tool.jsonc 中配置 models.{type}（provider/modelId）或 apiKey+apiBaseUrl+model", en: "Set models.{type} (provider/modelId) or apiKey+apiBaseUrl+model in file-tool.jsonc" },
+    meta_failed: { zh: "分析失败", en: "Failed" },
+    meta_skip: { zh: "跳过", en: "Skip" },
+    meta_not_found: { zh: "文件不存在", en: "Not found" },
+    meta_image: { zh: "图片", en: "Image" },
+    meta_error: { zh: "分析出错", en: "Error" },
+    no_cache: { zh: "[] (无缓存)", en: "[] (no cache)" },
+    enabled: { zh: "已启用", en: "Enabled" },
+    disabled: { zh: "已禁用", en: "Disabled" },
+    status: { zh: "图片缓存: {s}\n视觉模型: {m}\n文生图: {i}\n文生视频: {v}\n文生语音: {t}", en: "Image cache: {s}\nVision: {m}\nImage gen: {i}\nVideo gen: {v}\nTTS: {t}" },
+    status_cmd: { zh: "查看缓存开关状态", en: "Show cache status" },
+    vision_prompt_default: { zh: "请详细描述这张图片的内容，返回格式: [文件名] 描述", en: "Describe this image in detail, format: [filename] description" },
+    err_resolve_config: { zh: "无法解析模型配置: {model}。请在 file-tool.jsonc 中配置 models.{type} (provider/modelId) 或 apiKey+apiBaseUrl+model", en: "Cannot resolve model config: {model}. Set models.{type} (provider/modelId) or apiKey+apiBaseUrl+model in file-tool.jsonc" },
+    err_api: { zh: "API {status}: {msg}", en: "API {status}: {msg}" },
+    empty_response: { zh: "(空)", en: "(empty)" },
+    uncached: { zh: "未缓存", en: "uncached" },
+    uncached_hint: { zh: "文件未缓存（id={id}），请先启用缓存再操作", en: "File not cached (id={id}), enable cache first" },
+    cmd_desc: { zh: "文件缓存管理 + 多模型配置（视觉/文生图/文生视频/文生语音）", en: "File cache manager + multi-model config (vision/image/video/tts)" },
+    cmd_template: { zh: "直接调用 file_tool 工具。`list-provider` 列出模型（优先 API 查询），`set-provider [类型:]模型名` 切换模型（类型: vision/image/video/tts），`list-cache [类型] [数量]` 查看缓存，`enable/disable` 临时开关，`enable-save/disable-save` 持久化开关，`status` 查看状态。", en: "Call file_tool tool directly. `list-provider` list models (API-first), `set-provider [type:]model` switch (type: vision/image/video/tts), `list-cache [type] [count]` view cache, `enable/disable` temp toggle, `enable-save/disable-save` persist toggle, `status` show state." },
+};
+export const DESC = {
+    analyze_image: { zh: "用多模态模型分析图片。先调 file_tool list-cache 拿到文件ID，再用 file_id:类型:id 分析。", en: "Analyze images with multimodal model. Call file_tool list-cache first to get file IDs, then use file_id:type:id." },
+    text_to_image: { zh: "文生图：根据文本提示生成图片，结果缓存并返回 file_id:类型:id，可用 analyze_image 查看。", en: "Text-to-image: generate an image from a prompt, cached and returned as file_id:type:id." },
+    text_to_video: { zh: "文生视频：根据文本提示生成视频（异步提交+轮询），结果缓存并返回 file_id:类型:id。", en: "Text-to-video: generate a video from a prompt (async submit+poll), cached and returned as file_id:type:id." },
+    text_to_speech: { zh: "文生语音：将文本转为语音（TTS），结果缓存并返回 file_id:类型:id。", en: "Text-to-speech: convert text to audio, cached and returned as file_id:type:id." },
+    file_tool: { zh: "文件缓存管理。当你在上下文中看到 [Image N] 或收到 Cannot read 图片错误时，立即调 list-cache 获取文件ID，再用 analyze_image file_id:类型:id 分析。主模型能直接读取图片时建议用 `disable` 关闭缓存。", en: "File cache manager. When you see [Image N] or a Cannot read image error, call list-cache to get file IDs, then use analyze_image file_id:type:id. If the main model can read images directly, use `disable` to turn off caching." },
+    file_tool_args: { zh: "list-cache [类型] [数量]（类型: input/image/video/tts，如 list-cache image 3），list-provider, set-provider [<类型>:]<模型名>, enable/disable（临时）, enable-save/disable-save（持久化）, status — main 前缀查主会话（list-cache main [类型] [数量]）", en: "list-cache [type] [count] (type: input/image/video/tts, e.g. list-cache image 3), list-provider, set-provider [<type>:]<model>, enable/disable (temp), enable-save/disable-save (persist), status — main prefix for root session (list-cache main [type] [count])" },
+    analyze_args_source: { zh: "file_path=file_id:类型:id（如 file_id:image:2）", en: "file_path=file_id:type:id (e.g. file_id:image:2)" },
+    analyze_args_data: { zh: "file_id:类型:id 或 base64", en: "file_id:type:id or base64" },
+    analyze_args_prompt: { zh: "分析提示", en: "prompt" },
+    gen_args_prompt: { zh: "生成提示词", en: "generation prompt" },
+    gen_args_size: { zh: "图片尺寸（如 1024x1024）", en: "image size (e.g. 1024x1024)" },
+    gen_args_duration: { zh: "视频时长（秒）", en: "video duration (seconds)" },
+    gen_args_voice: { zh: "音色（如 alloy）", en: "voice (e.g. alloy)" },
+};
+export function T(key, params) {
+    const entry = TX[key] || { zh: key, en: key };
+    const t = LANG === "zh" ? entry.zh : entry.en;
+    if (!params)
+        return t;
+    return Object.entries(params).reduce((s, [k, v]) => s.replace(`{${k}}`, v), t);
+}

package/dist/providers.js

@@ -0,0 +1,61 @@
+import { T } from "./i18n.js";
+import { isAgnesProvider, postJson, getJson } from "./utils.js";
+// 视频生成 - agnes 私有接口
+export async function generateVideoAgnes(cfg, prompt, duration) {
+    const dur = duration || 5;
+    const fps = 24;
+    const frames = dur * fps + 1;
+    const data = await postJson(`${cfg.baseURL}/videos`, cfg.apiKey, {
+        model: cfg.modelId, prompt, num_frames: Math.min(frames, 441), frame_rate: fps,
+    });
+    const taskId = data?.id || data?.task_id || data?.video_id;
+    if (!taskId)
+        throw new Error(T("gen_no_result"));
+    const deadline = Date.now() + 600000;
+    while (Date.now() < deadline) {
+        await new Promise(r => setTimeout(r, 10000));
+        const pd = await getJson(`https://apihub.agnes-ai.com/agnesapi?video_id=${taskId}&model_name=${cfg.modelId}`, cfg.apiKey);
+        const st = pd?.status || pd?.state || "";
+        if (st === "completed") {
+            const url = pd?.metadata?.url || pd?.url || "";
+            if (url)
+                return url;
+        }
+        else if (st === "failed" || st === "error") {
+            throw new Error(T("video_failed", { msg: pd?.error || st }));
+        }
+    }
+    throw new Error(T("video_failed", { msg: "timeout" }));
+}
+// 视频生成 - OpenAI 标准接口
+export async function generateVideoOpenAI(cfg, prompt, duration) {
+    const body = { model: cfg.modelId, prompt };
+    if (duration)
+        body.duration = duration;
+    const data = await postJson(`${cfg.baseURL}/videos/generations`, cfg.apiKey, body);
+    const taskId = data?.id || data?.task_id || data?.data?.[0]?.id;
+    if (!taskId)
+        throw new Error(T("gen_no_result"));
+    const deadline = Date.now() + 600000;
+    while (Date.now() < deadline) {
+        await new Promise(r => setTimeout(r, 10000));
+        const pd = await getJson(`${cfg.baseURL}/videos/generations/${taskId}`, cfg.apiKey);
+        const st = pd?.status || pd?.state || "";
+        if (st === "completed" || st === "succeeded") {
+            const out = pd?.output?.[0] || pd?.output;
+            const url = typeof out === "string" ? out : out?.url || "";
+            if (url)
+                return url;
+        }
+        else if (st === "failed" || st === "error") {
+            throw new Error(T("video_failed", { msg: pd?.error || st }));
+        }
+    }
+    throw new Error(T("video_failed", { msg: "timeout" }));
+}
+// 视频生成 - 路由：按域名选择适配器
+export async function generateVideo(cfg, prompt, duration) {
+    if (isAgnesProvider(cfg.baseURL))
+        return generateVideoAgnes(cfg, prompt, duration);
+    return generateVideoOpenAI(cfg, prompt, duration);
+}

package/dist/tools/analyze.js

@@ -0,0 +1,94 @@
+import { tool } from "@opencode-ai/plugin";
+import { existsSync, readFileSync } from "node:fs";
+import { join } from "node:path";
+import { DESC, T, LANG } from "../i18n.js";
+import { requireModelCfg, getCfg } from "../config.js";
+import { findFileInChain, readFileData } from "../cache.js";
+const CACHE_TYPES_LIST = ["input", "image", "video", "tts"];
+export const analyzeImageTool = tool({
+    description: DESC.analyze_image[LANG],
+    args: {
+        source: tool.schema.enum(["file_path", "base64"]).describe(DESC.analyze_args_source[LANG]),
+        data: tool.schema.string().describe(DESC.analyze_args_data[LANG]),
+        prompt: tool.schema.string().optional().describe(DESC.analyze_args_prompt[LANG]),
+    },
+    execute: async ({ source, data, prompt }, context) => {
+        let imageUrl, fileName = "";
+        if (source === "file_path" && data.startsWith("file_id:")) {
+            const segs = data.slice(8).split(":");
+            let type = "input";
+            let fid;
+            if (segs.length === 2 && CACHE_TYPES_LIST.includes(segs[0])) {
+                type = segs[0];
+                fid = parseInt(segs[1], 10);
+            }
+            else {
+                fid = parseInt(segs[0], 10);
+            }
+            const found = findFileInChain(context.sessionID, type, fid);
+            if (!found) {
+                context.metadata?.({ title: T("meta_failed") });
+                return T("file_id_not_found", { id: `${type}:${fid}` });
+            }
+            const file = found.file;
+            if (!file.cached) {
+                context.metadata?.({ title: T("meta_skip") });
+                return T("uncached_hint", { id: `${type}:${fid}` });
+            }
+            if (!file.mime.startsWith("image/")) {
+                context.metadata?.({ title: T("meta_skip") });
+                return T("not_an_image", { name: file.filename, mime: file.mime });
+            }
+            fileName = file.filename;
+            imageUrl = readFileData(context.sessionID, type, fid) || "";
+            if (!imageUrl) {
+                context.metadata?.({ title: T("meta_failed") });
+                return T("file_data_not_found", { id: `${type}:${fid}` });
+            }
+            prompt = prompt || T("describe_image", { name: fileName });
+        }
+        else if (source === "file_path") {
+            if (!existsSync(data)) {
+                const tryPath = join(context.directory, data);
+                if (existsSync(tryPath))
+                    data = tryPath;
+            }
+            if (!existsSync(data)) {
+                context.metadata?.({ title: T("meta_not_found") });
+                return T("file_not_found", { path: data });
+            }
+            const ext = data.split(".").pop()?.toLowerCase() || "";
+            const mimeMap = { png: "image/png", jpg: "image/jpeg", jpeg: "image/jpeg", bmp: "image/bmp", gif: "image/gif", webp: "image/webp" };
+            const mime = mimeMap[ext] || "image/png";
+            fileName = data.split(/[/\\]/).pop() || "";
+            imageUrl = `data:${mime};base64,${readFileSync(data).toString("base64")}`;
+        }
+        else if (source === "base64") {
+            imageUrl = `data:image/png;base64,${data.replace(/^data:image\/\w+;base64,/, "")}`;
+        }
+        else {
+            return T("unsupported_source", { source });
+        }
+        try {
+            const cfg = requireModelCfg("vision");
+            const resp = await fetch(`${cfg.baseURL}/chat/completions`, {
+                method: "POST",
+                headers: { Authorization: `Bearer ${cfg.apiKey}`, "Content-Type": "application/json" },
+                body: JSON.stringify({ model: cfg.modelId, messages: [{ role: "user", content: [{ type: "text", text: prompt || T("vision_prompt_default") }, { type: "image_url", image_url: { url: imageUrl } }] }], max_tokens: getCfg().maxTokens }),
+                signal: AbortSignal.timeout(getCfg().timeout || 60000),
+            });
+            if (!resp.ok)
+                throw new Error(T("err_api", { status: String(resp.status), msg: (await resp.text().catch(() => "unknown")).slice(0, 200) }));
+            const data = await resp.json();
+            const msg = data.choices?.[0]?.message;
+            const result = msg?.content || msg?.reasoning_content || T("empty_response");
+            context.metadata?.({ title: `[Vision] ${fileName || T("meta_image")}`, metadata: { sessionID: context.sessionID, messageID: context.messageID } });
+            return "[Vision] " + result;
+        }
+        catch (e) {
+            const msg = e instanceof Error ? e.message : String(e);
+            context.metadata?.({ title: T("meta_error") });
+            return `[Vision Error] ${msg}`;
+        }
+    },
+});