@xiaoqiong0v0/opencode-file-tool 1.1.1 → 1.2.0

package/dist/file-tool.js

@@ -1,9 +1,6 @@
-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";
+import { fileToolCli } from "./tools/cli.js";
 try {
     loadCfg();
 }
@@ -13,11 +10,6 @@ catch (e) {
 export const fileToolPlugin = async () => {
     log.loaded();
     return {
-        config: async (config) => {
-            const commands = config.command ?? {};
-            commands["file-tool"] = { template: T("cmd_template"), description: T("cmd_desc") };
-            config.command = commands;
-        },
         event: async ({ event }) => {
             const props = event.properties;
             const sid = props?.sessionID;
@@ -71,11 +63,7 @@ export const fileToolPlugin = async () => {
             }
         },
         tool: {
-            analyze_image: analyzeImageTool,
-            text_to_image: textToImageTool,
-            text_to_video: textToVideoTool,
-            text_to_speech: textToSpeechTool,
-            file_tool: fileTool,
+            file_tool: fileToolCli,
         },
     };
 };

package/dist/i18n.js

@@ -40,32 +40,10 @@ export const TX = {
     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。`list-cache input` 看用户图片，`list-cache image|video|tts` 看生成产物，`list-provider` 列出模型，`set-provider [类型:]模型名` 切换，`enable/disable` 临时开关，`enable-save/disable-save` 持久化，`status` 查看状态。", en: "Call file_tool. `list-cache input` user images, `list-cache image|video|tts` generated, `list-provider` list models, `set-provider [type:]model` switch, `enable/disable` temp, `enable-save/disable-save` persist, `status` show state." },
 };
 export const DESC = {
-    analyze_image: {
-        zh: "用多模态模型分析图片。file_id:类型:id，类型: input=用户图片/image=生成图。先用 list-cache 拿 ID 再分析。",
-        en: "Analyze image with multimodal model. file_id:type:id; types: input=user image/image=generated. Get ID via list-cache first.",
-    },
-    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。类型: input=用户图片, image/video/tts=生成产物。查看用户图片用 `list-cache input`；生成产物用 `list-cache image|video|tts`。file_id:类型:id 供 analyze_image 分析。",
-        en: "File cache. Types: input=user images, image/video/tts=generated. View user images via `list-cache input`; generated via `list-cache image|video|tts`. file_id:type:id for analyze_image.",
-    },
-    file_tool_args: {
-        zh: "list-cache [类型] [数量]: input=用户图片(最常用)/image/video/tts=生成产物; list-provider; set-provider [类型:]模型; enable/disable; enable-save/disable-save; status; main 前缀查主会话",
-        en: "list-cache [type] [count]: input=user images(common)/image/video/tts=generated; list-provider; set-provider [type:]model; enable/disable; enable-save/disable-save; status; main prefix for root session",
-    },
-    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)" },
+    file_tool: { zh: "统一命令行工具。子命令: analyze 分析图 / imagine 文生图 / video 文生视频 / tts 文生语音 / list 缓存列表 / providers 模型列表 / set 切换模型 / status / enable-disable。空参数或 help 看用法。", en: "Single CLI tool. Subcommands: analyze image / imagine t2i / video t2v / tts / list cache / providers models / set model / status / enable-disable. Empty or help for usage." },
+    file_tool_args: { zh: "完整命令行字符串，如 'list input'、'imagine 一只橘猫'、'analyze file_id:input:1'；空时默认 help。类型: input=用户图片, image/video/tts=生成产物。", en: "Full command string, e.g. 'list input', 'imagine a cat', 'analyze file_id:input:1'; empty defaults to help. Types: input=user image, image/video/tts=generated." },
 };
 export function T(key, params) {
     const entry = TX[key] || { zh: key, en: key };

package/dist/tools/cli.js

@@ -0,0 +1,342 @@
+import { tool } from "@opencode-ai/plugin";
+import { existsSync, readFileSync } from "node:fs";
+import { join } from "node:path";
+import stringArgv from "string-argv";
+import { parseArgs } from "node:util";
+import { T, LANG, DESC } from "../i18n.js";
+import { CONFIG_PATH, MODEL_TYPES, CACHE_TYPES, DEFAULT_CFG, getCfg, getProviderNames, readJsonc, ENABLED, setEnabled, saveCfg, requireModelCfg } from "../config.js";
+import { readTypeStore, getRootSession, findFileInChain, readFileData, registerGeneratedFile, fetchToBuffer, buildGenFilename, extForMime } from "../cache.js";
+import { generateVideo } from "../providers.js";
+import { genError } from "../utils.js";
+import { localModelList, listModelsApi, currentModelSummary } from "./file_tool_helpers.js";
+const GEN_TIMEOUT = 300000;
+function helpText() {
+    if (LANG === "zh") {
+        return `用法: file_tool <子命令> [参数]
+子命令:
+  analyze <file_id:类型:id|路径|base64> [提示语]  分析图片
+  imagine <提示词> [--size 1024x1024]            文生图
+  video <提示词> [--duration 5]                  文生视频(异步)
+  tts <文本> [--voice alloy]                     文生语音
+  list [类型] [数量]                              缓存列表 (input=用户图/image/video/tts=生成)
+  providers                                       列出可用模型
+  set [类型:]模型名                                切换模型
+  status                                          查看状态
+  enable|disable|enable-save|disable-save         缓存开关
+  help                                            此帮助
+示例:
+  file_tool analyze file_id:input:1
+  file_tool imagine "一只橘猫" --size 512x512
+  file_tool list input
+  file_tool set image:agnes/agnes-image-2.1-flash`;
+    }
+    return `Usage: file_tool <subcommand> [args]
+Subcommands:
+  analyze <file_id:type:id|path|base64> [prompt]  analyze image
+  imagine <prompt> [--size 1024x1024]             text-to-image
+  video <prompt> [--duration 5]                   text-to-video (async)
+  tts <text> [--voice alloy]                      text-to-speech
+  list [type] [count]                             list cache (input=user image/image/video/tts=generated)
+  providers                                       list available models
+  set [type:]model                                switch model
+  status                                          show state
+  enable|disable|enable-save|disable-save         cache toggle
+  help                                            this help
+Examples:
+  file_tool analyze file_id:input:1
+  file_tool imagine "an orange cat" --size 512x512
+  file_tool list input
+  file_tool set image:agnes/agnes-image-2.1-flash`;
+}
+async function analyzeCmd(args, context) {
+    const { positionals } = parseArgs({ args, allowPositionals: true });
+    let data = positionals[0] || "";
+    let prompt = positionals.slice(1).join(" ") || "";
+    if (!data)
+        return `${T("specify_model")}\n\n${helpText()}`;
+    let imageUrl, fileName = "";
+    if (data.startsWith("file_id:")) {
+        const segs = data.slice(8).split(":");
+        let type = "input";
+        let fid;
+        if (segs.length === 2 && CACHE_TYPES.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 (existsSync(data) || existsSync(join(context.directory, data))) {
+        if (!existsSync(data))
+            data = join(context.directory, data);
+        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 {
+        imageUrl = `data:image/png;base64,${data.replace(/^data:image\/\w+;base64,/, "")}`;
+    }
+    try {
+        const cfg = requireModelCfg("vision");
+        const resp = await fetch(`${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 d = await resp.json();
+        const msg = d.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] ${msg}`;
+    }
+}
+async function imagineCmd(args, context) {
+    try {
+        const { values, positionals } = parseArgs({ args, allowPositionals: true, options: { size: { type: "string" } } });
+        const prompt = positionals.join(" ");
+        const size = values.size;
+        const cfg = requireModelCfg("image");
+        const body = { model: cfg.modelId, prompt };
+        if (size)
+            body.size = size;
+        const resp = await fetch(`${cfg.baseURL}/images/generations`, {
+            method: "POST",
+            headers: { Authorization: `Bearer ${cfg.apiKey}`, "Content-Type": "application/json" },
+            body: JSON.stringify(body),
+            signal: AbortSignal.timeout(GEN_TIMEOUT),
+        });
+        if (!resp.ok)
+            throw new Error(T("err_api", { status: String(resp.status), msg: (await resp.text().catch(() => "unknown")).slice(0, 200) }));
+        const data = await resp.json();
+        const item = data?.data?.[0];
+        if (!item)
+            throw new Error(T("gen_no_result"));
+        let buffer, mime = "image/png";
+        if (item.b64_json)
+            buffer = Buffer.from(item.b64_json, "base64");
+        else if (item.url) {
+            const d = await fetchToBuffer(item.url);
+            buffer = d.buffer;
+            mime = d.mime;
+        }
+        else
+            throw new Error(T("gen_no_result"));
+        const store = readTypeStore(context.sessionID, "image");
+        const r = registerGeneratedFile(context.sessionID, "image", buildGenFilename("image", mime, store.nextId), mime, context.messageID, buffer);
+        context.metadata?.({ title: "[ImageGen]", metadata: { sessionID: context.sessionID, messageID: context.messageID } });
+        return T("image_generated", { name: `image_${r.id}.${extForMime(mime)}`, id: `image:${r.id}`, path: r.path });
+    }
+    catch (e) {
+        context.metadata?.({ title: T("meta_error") });
+        return genError(e);
+    }
+}
+async function videoCmd(args, context) {
+    try {
+        const { values, positionals } = parseArgs({ args, allowPositionals: true, options: { duration: { type: "string" } } });
+        const prompt = positionals.join(" ");
+        const duration = values.duration ? parseInt(values.duration, 10) : undefined;
+        const cfg = requireModelCfg("video");
+        const url = await generateVideo(cfg, prompt, duration);
+        const d = await fetchToBuffer(url);
+        const store = readTypeStore(context.sessionID, "video");
+        const r = registerGeneratedFile(context.sessionID, "video", buildGenFilename("video", d.mime, store.nextId), d.mime, context.messageID, d.buffer);
+        context.metadata?.({ title: "[VideoGen]", metadata: { sessionID: context.sessionID, messageID: context.messageID } });
+        return T("video_generated", { name: `video_${r.id}.${extForMime(d.mime)}`, id: `video:${r.id}`, path: r.path });
+    }
+    catch (e) {
+        context.metadata?.({ title: T("meta_error") });
+        return genError(e);
+    }
+}
+async function ttsCmd(args, context) {
+    try {
+        const { values, positionals } = parseArgs({ args, allowPositionals: true, options: { voice: { type: "string" } } });
+        const text = positionals.join(" ");
+        const voice = values.voice;
+        const cfg = requireModelCfg("tts");
+        const body = { model: cfg.modelId, input: text };
+        if (voice)
+            body.voice = voice;
+        const resp = await fetch(`${cfg.baseURL}/audio/speech`, {
+            method: "POST",
+            headers: { Authorization: `Bearer ${cfg.apiKey}`, "Content-Type": "application/json" },
+            body: JSON.stringify(body),
+            signal: AbortSignal.timeout(GEN_TIMEOUT),
+        });
+        if (!resp.ok)
+            throw new Error(T("err_api", { status: String(resp.status), msg: (await resp.text().catch(() => "unknown")).slice(0, 200) }));
+        const buf = Buffer.from(await resp.arrayBuffer());
+        const mime = resp.headers.get("content-type")?.split(";")[0] || "audio/mpeg";
+        const store = readTypeStore(context.sessionID, "tts");
+        const r = registerGeneratedFile(context.sessionID, "tts", buildGenFilename("tts", mime, store.nextId), mime, context.messageID, buf);
+        context.metadata?.({ title: "[TTS]", metadata: { sessionID: context.sessionID, messageID: context.messageID } });
+        return T("tts_generated", { name: `tts_${r.id}.${extForMime(mime)}`, id: `tts:${r.id}`, path: r.path });
+    }
+    catch (e) {
+        context.metadata?.({ title: T("meta_error") });
+        return genError(e);
+    }
+}
+async function providersCmd() {
+    const providerNames = getProviderNames();
+    const models = await listModelsApi(providerNames);
+    const fallbackModels = localModelList();
+    for (const m of fallbackModels)
+        if (!models.includes(m))
+            models.push(m);
+    const modelLines = models.map(m => "  " + m).join("\n");
+    return T("current_model", { model: currentModelSummary(), list: modelLines });
+}
+function setCmd(args) {
+    const { positionals } = parseArgs({ args, allowPositionals: true });
+    const arg = positionals.join(":") || positionals[0] || "";
+    if (!arg)
+        return T("specify_model");
+    let type = "vision";
+    let model = arg;
+    const colon = arg.indexOf(":");
+    if (colon > 0 && MODEL_TYPES.includes(arg.slice(0, colon))) {
+        type = arg.slice(0, colon);
+        model = arg.slice(colon + 1).trim();
+    }
+    if (!model)
+        return T("specify_model");
+    const cfg = existsSync(CONFIG_PATH) ? readJsonc(CONFIG_PATH) : {};
+    const models = { ...DEFAULT_CFG.models, ...((cfg.models && typeof cfg.models === "object") ? cfg.models : {}) };
+    models[type] = model;
+    saveCfg({ models });
+    return T("model_switched", { type, model });
+}
+function statusCmd() {
+    const c = getCfg();
+    const fmt = (t) => c.models[t]?.model || T("model_not_set");
+    return T("status", { s: ENABLED ? T("enabled") : T("disabled"), m: fmt("vision"), i: fmt("image"), v: fmt("video"), t: fmt("tts") });
+}
+function listCmd(args, sessionID) {
+    const arg = args.join(" ");
+    let targetSid = sessionID;
+    let rest = arg;
+    if (arg === "main" || arg.startsWith("main ")) {
+        targetSid = getRootSession(sessionID);
+        rest = arg === "main" ? "" : arg.slice(5).trim();
+    }
+    const tokens = rest.split(/\s+/).filter(Boolean);
+    let filter = null;
+    let countStr = "1";
+    if (tokens[0] && CACHE_TYPES.includes(tokens[0])) {
+        filter = tokens[0];
+        countStr = tokens[1] || "all";
+    }
+    else if (tokens[0]) {
+        countStr = tokens[0];
+    }
+    const types = filter ? [filter] : CACHE_TYPES;
+    const lines = [];
+    for (const type of types) {
+        const store = readTypeStore(targetSid, type);
+        const entries = Object.values(store.files);
+        if (entries.length === 0)
+            continue;
+        let show = entries;
+        if (countStr !== "all") {
+            const n = parseInt(countStr, 10);
+            if (!isNaN(n) && n > 0)
+                show = entries.slice(-n);
+        }
+        lines.push(`  ${type}:`);
+        for (const f of show) {
+            const src = f.msgId ? `msg_${f.msgId.slice(-8)}` : type === "input" ? "input" : "generated";
+            lines.push(`    ${f.filename} (${type}:${f.id}) [${src}]`);
+        }
+    }
+    if (lines.length === 0)
+        return `${targetSid}: ${T("no_cache")}`;
+    return `${targetSid}:\n${lines.join("\n")}`;
+}
+async function handleCommand(raw, context) {
+    const tokens = stringArgv(raw);
+    const [cmd, ...rest] = tokens;
+    if (!cmd)
+        return helpText();
+    switch (cmd) {
+        case "help":
+        case "h":
+        case "-h":
+        case "--help": return helpText();
+        case "analyze":
+        case "a": return analyzeCmd(rest, context);
+        case "imagine":
+        case "i": return imagineCmd(rest, context);
+        case "video":
+        case "v": return videoCmd(rest, context);
+        case "tts":
+        case "t": return ttsCmd(rest, context);
+        case "list":
+        case "l": return listCmd(rest, context.sessionID);
+        case "providers":
+        case "list-provider":
+        case "p": return providersCmd();
+        case "set":
+        case "set-provider":
+        case "s": return setCmd(rest);
+        case "status": return statusCmd();
+        case "enable":
+            setEnabled(true);
+            return T("enabled");
+        case "disable":
+            setEnabled(false);
+            return T("disabled");
+        case "enable-save":
+            saveCfg({ enabled: true });
+            return T("enabled");
+        case "disable-save":
+            saveCfg({ enabled: false });
+            return T("disabled");
+        default:
+            return T("unknown_cmd", { cmd: cmd.slice(0, 50) }) + "\n\n" + helpText();
+    }
+}
+export const fileToolCli = tool({
+    description: DESC.file_tool[LANG],
+    args: { command: tool.schema.string().optional().describe(DESC.file_tool_args[LANG]) },
+    execute: async ({ command }, context) => {
+        return handleCommand(command ?? "help", context);
+    },
+});

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xiaoqiong0v0/opencode-file-tool",
-  "version": "1.1.1",
+  "version": "1.2.0",
   "type": "module",
   "description": "File cache & image analysis plugin for OpenCode. Auto-caches pasted images, analyzes via multimodal model.",
   "files": [
@@ -28,7 +28,8 @@
     "@opencode-ai/plugin": "*"
   },
   "dependencies": {
-    "@xiaoqiong0v0/opencode-plugin-logger": "^1.0.2"
+    "@xiaoqiong0v0/opencode-plugin-logger": "^1.0.4",
+    "string-argv": "^0.3.2"
   },
   "devDependencies": {
     "@types/node": "^26.1.1",

package/dist/tools/analyze.js

@@ -1,94 +0,0 @@
-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}`;
-        }
-    },
-});

package/dist/tools/file_tool.js

@@ -1,105 +0,0 @@
-import { tool } from "@opencode-ai/plugin";
-import { existsSync } from "node:fs";
-import { DESC, T, LANG } from "../i18n.js";
-import { CONFIG_PATH, MODEL_TYPES, CACHE_TYPES, DEFAULT_CFG, getCfg, getProviderNames, readJsonc, ENABLED, setEnabled, saveCfg } from "../config.js";
-import { readTypeStore, getRootSession } from "../cache.js";
-import { localModelList, listModelsApi, currentModelSummary } from "./file_tool_helpers.js";
-export const fileTool = 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 providerNames = getProviderNames();
-            const models = await listModelsApi(providerNames);
-            const fallbackModels = localModelList();
-            for (const m of fallbackModels)
-                if (!models.includes(m))
-                    models.push(m);
-            const modelLines = models.map(m => "  " + m).join("\n");
-            return T("current_model", { model: currentModelSummary(), list: modelLines });
-        }
-        if (cmd.startsWith("set-provider ")) {
-            const arg = cmd.slice(13).trim();
-            if (!arg)
-                return T("specify_model");
-            let type = "vision";
-            let model = arg;
-            const colon = arg.indexOf(":");
-            if (colon > 0 && MODEL_TYPES.includes(arg.slice(0, colon))) {
-                type = arg.slice(0, colon);
-                model = arg.slice(colon + 1).trim();
-            }
-            if (!model)
-                return T("specify_model");
-            const cfg = existsSync(CONFIG_PATH) ? readJsonc(CONFIG_PATH) : {};
-            const models = { ...DEFAULT_CFG.models, ...((cfg.models && typeof cfg.models === "object") ? cfg.models : {}) };
-            models[type] = model;
-            saveCfg({ models });
-            return T("model_switched", { type, model });
-        }
-        if (cmd === "disable") {
-            setEnabled(false);
-            return T("disabled");
-        }
-        if (cmd === "enable") {
-            setEnabled(true);
-            return T("enabled");
-        }
-        if (cmd === "disable-save") {
-            saveCfg({ enabled: false });
-            return T("disabled");
-        }
-        if (cmd === "enable-save") {
-            saveCfg({ enabled: true });
-            return T("enabled");
-        }
-        if (cmd === "status") {
-            const c = getCfg();
-            const fmt = (t) => c.models[t]?.model || T("model_not_set");
-            return T("status", { s: ENABLED ? T("enabled") : T("disabled"), m: fmt("vision"), i: fmt("image"), v: fmt("video"), t: fmt("tts") });
-        }
-        if (cmd === "list-cache" || cmd.startsWith("list-cache ")) {
-            const rest = cmd === "list-cache" ? "" : cmd.slice(11).trim();
-            let targetSid = context.sessionID;
-            let arg = rest;
-            if (arg === "main" || arg.startsWith("main ")) {
-                targetSid = getRootSession(context.sessionID);
-                arg = arg === "main" ? "" : arg.slice(5).trim();
-            }
-            const tokens = arg.split(/\s+/).filter(Boolean);
-            let filter = null;
-            let countStr = "1";
-            if (tokens[0] && CACHE_TYPES.includes(tokens[0])) {
-                filter = tokens[0];
-                countStr = tokens[1] || "all";
-            }
-            else if (tokens[0]) {
-                countStr = tokens[0];
-            }
-            const types = filter ? [filter] : CACHE_TYPES;
-            const lines = [];
-            for (const type of types) {
-                const store = readTypeStore(targetSid, type);
-                const entries = Object.values(store.files);
-                if (entries.length === 0)
-                    continue;
-                let show = entries;
-                if (countStr !== "all") {
-                    const n = parseInt(countStr, 10);
-                    if (!isNaN(n) && n > 0)
-                        show = entries.slice(-n);
-                }
-                lines.push(`  ${type}:`);
-                for (const f of show) {
-                    const src = f.msgId ? `msg_${f.msgId.slice(-8)}` : type === "input" ? "input" : "generated";
-                    lines.push(`    ${f.filename} (${type}:${f.id}) [${src}]`);
-                }
-            }
-            if (lines.length === 0)
-                return `${targetSid}: ${T("no_cache")}`;
-            return `${targetSid}:\n${lines.join("\n")}`;
-        }
-        return T("unknown_cmd", { cmd });
-    },
-});

package/dist/tools/generate.js

@@ -1,107 +0,0 @@
-import { tool } from "@opencode-ai/plugin";
-import { DESC, T, LANG } from "../i18n.js";
-import { requireModelCfg } from "../config.js";
-import { readTypeStore, registerGeneratedFile, fetchToBuffer, buildGenFilename, extForMime } from "../cache.js";
-import { generateVideo } from "../providers.js";
-import { genError } from "../utils.js";
-const GEN_TIMEOUT = 300000;
-export const textToImageTool = tool({
-    description: DESC.text_to_image[LANG],
-    args: {
-        prompt: tool.schema.string().describe(DESC.gen_args_prompt[LANG]),
-        size: tool.schema.string().optional().describe(DESC.gen_args_size[LANG]),
-    },
-    execute: async ({ prompt, size }, context) => {
-        try {
-            const cfg = requireModelCfg("image");
-            const body = { model: cfg.modelId, prompt };
-            if (size)
-                body.size = size;
-            const resp = await fetch(`${cfg.baseURL}/images/generations`, {
-                method: "POST",
-                headers: { Authorization: `Bearer ${cfg.apiKey}`, "Content-Type": "application/json" },
-                body: JSON.stringify(body),
-                signal: AbortSignal.timeout(GEN_TIMEOUT),
-            });
-            if (!resp.ok)
-                throw new Error(T("err_api", { status: String(resp.status), msg: (await resp.text().catch(() => "unknown")).slice(0, 200) }));
-            const data = await resp.json();
-            const item = data?.data?.[0];
-            if (!item)
-                throw new Error(T("gen_no_result"));
-            let buffer, mime = "image/png";
-            if (item.b64_json)
-                buffer = Buffer.from(item.b64_json, "base64");
-            else if (item.url) {
-                const d = await fetchToBuffer(item.url);
-                buffer = d.buffer;
-                mime = d.mime;
-            }
-            else
-                throw new Error(T("gen_no_result"));
-            const store = readTypeStore(context.sessionID, "image");
-            const r = registerGeneratedFile(context.sessionID, "image", buildGenFilename("image", mime, store.nextId), mime, context.messageID, buffer);
-            context.metadata?.({ title: "[ImageGen]", metadata: { sessionID: context.sessionID, messageID: context.messageID } });
-            return T("image_generated", { name: `image_${r.id}.${extForMime(mime)}`, id: `image:${r.id}`, path: r.path });
-        }
-        catch (e) {
-            context.metadata?.({ title: T("meta_error") });
-            return genError(e);
-        }
-    },
-});
-export const textToVideoTool = tool({
-    description: DESC.text_to_video[LANG],
-    args: {
-        prompt: tool.schema.string().describe(DESC.gen_args_prompt[LANG]),
-        duration: tool.schema.number().optional().describe(DESC.gen_args_duration[LANG]),
-    },
-    execute: async ({ prompt, duration }, context) => {
-        try {
-            const cfg = requireModelCfg("video");
-            const url = await generateVideo(cfg, prompt, duration);
-            const d = await fetchToBuffer(url);
-            const store = readTypeStore(context.sessionID, "video");
-            const r = registerGeneratedFile(context.sessionID, "video", buildGenFilename("video", d.mime, store.nextId), d.mime, context.messageID, d.buffer);
-            context.metadata?.({ title: "[VideoGen]", metadata: { sessionID: context.sessionID, messageID: context.messageID } });
-            return T("video_generated", { name: `video_${r.id}.${extForMime(d.mime)}`, id: `video:${r.id}`, path: r.path });
-        }
-        catch (e) {
-            context.metadata?.({ title: T("meta_error") });
-            return genError(e);
-        }
-    },
-});
-export const textToSpeechTool = tool({
-    description: DESC.text_to_speech[LANG],
-    args: {
-        text: tool.schema.string().describe(DESC.gen_args_prompt[LANG]),
-        voice: tool.schema.string().optional().describe(DESC.gen_args_voice[LANG]),
-    },
-    execute: async ({ text, voice }, context) => {
-        try {
-            const cfg = requireModelCfg("tts");
-            const body = { model: cfg.modelId, input: text };
-            if (voice)
-                body.voice = voice;
-            const resp = await fetch(`${cfg.baseURL}/audio/speech`, {
-                method: "POST",
-                headers: { Authorization: `Bearer ${cfg.apiKey}`, "Content-Type": "application/json" },
-                body: JSON.stringify(body),
-                signal: AbortSignal.timeout(GEN_TIMEOUT),
-            });
-            if (!resp.ok)
-                throw new Error(T("err_api", { status: String(resp.status), msg: (await resp.text().catch(() => "unknown")).slice(0, 200) }));
-            const buf = Buffer.from(await resp.arrayBuffer());
-            const mime = resp.headers.get("content-type")?.split(";")[0] || "audio/mpeg";
-            const store = readTypeStore(context.sessionID, "tts");
-            const r = registerGeneratedFile(context.sessionID, "tts", buildGenFilename("tts", mime, store.nextId), mime, context.messageID, buf);
-            context.metadata?.({ title: "[TTS]", metadata: { sessionID: context.sessionID, messageID: context.messageID } });
-            return T("tts_generated", { name: `tts_${r.id}.${extForMime(mime)}`, id: `tts:${r.id}`, path: r.path });
-        }
-        catch (e) {
-            context.metadata?.({ title: T("meta_error") });
-            return genError(e);
-        }
-    },
-});