chainlesschain 0.162.82 → 0.162.84

package/src/lib/image-input.js

@@ -44,6 +44,96 @@ export function resolveImages(paths, deps = {}) {
   });
 }
 
+/**
+ * Claude-Code-style auto-detection: find local image-file paths mentioned in a
+ * prompt so the turn can attach them as vision input. A token is treated as an
+ * image only when (a) its extension is one we support and (b) the file actually
+ * exists on disk — so prose like "I edited a.png" attaches only when a.png is
+ * really there, and a typo'd/remote path stays as plain text. Matched path
+ * tokens are stripped from the returned text ("describe ./a.png" → "describe" +
+ * an attachment); an all-path message leaves empty text, which the caller turns
+ * into an image-only turn. URLs / data: URIs are never auto-attached (local
+ * files only). Pure except for the existence check (inject `deps.existsSync`).
+ *
+ * @param {string} text
+ * @param {object} deps  { existsSync }
+ * @returns {{ images: string[], text: string }}
+ */
+export function detectImagePaths(text, deps = {}) {
+  const existsSync = deps.existsSync || fs.existsSync;
+  if (typeof text !== "string" || !text.trim()) {
+    return { images: [], text: typeof text === "string" ? text : "" };
+  }
+  const exts = Object.keys(EXT_MEDIA)
+    .map((e) => e.slice(1))
+    .join("|"); // png|jpg|jpeg|gif|webp
+  // Quoted ("…"/'…') paths first (may contain spaces), then bare
+  // whitespace-delimited tokens ending in a supported image extension.
+  const re = new RegExp(
+    `"([^"]+?\\.(?:${exts}))"|'([^']+?\\.(?:${exts}))'|(\\S+?\\.(?:${exts}))(?=$|[\\s)\\]'",])`,
+    "gi",
+  );
+  const images = [];
+  const seen = new Set();
+  const ranges = [];
+  let m;
+  while ((m = re.exec(text)) !== null) {
+    const raw = m[1] || m[2] || m[3];
+    if (!raw) continue;
+    // Local files only — never auto-attach URLs / data: URIs.
+    if (/^[a-z][a-z0-9+.-]*:\/\//i.test(raw) || /^data:/i.test(raw)) continue;
+    let exists = false;
+    try {
+      exists = existsSync(raw);
+    } catch {
+      exists = false;
+    }
+    if (!exists) continue;
+    if (!seen.has(raw)) {
+      seen.add(raw);
+      images.push(raw);
+    }
+    ranges.push([m.index, m.index + m[0].length]);
+  }
+  if (!images.length) return { images: [], text };
+  let out = text;
+  for (let i = ranges.length - 1; i >= 0; i--) {
+    out = out.slice(0, ranges[i][0]) + out.slice(ranges[i][1]);
+  }
+  out = out.replace(/ {2,}/g, " ").trim();
+  return { images, text: out };
+}
+
+/**
+ * Compose a REPL/interactive turn from raw typed text: auto-detect local image
+ * paths, build the user-message content, and (on an image turn) the per-turn
+ * vision-LLM override. Pure orchestration over detectImagePaths/resolveImages/
+ * buildUserContent/resolveVisionLlm so the interactive wiring is unit-testable
+ * (the REPL itself forces a terminal and can't be driven headlessly).
+ *
+ * @param {string} text     the user's typed message
+ * @param {object} llm      { provider, baseUrl, apiKey, visionModel } current LLM
+ * @param {object} deps     { existsSync, fs } injected for tests
+ * @returns {{ content, images: string[], visionLlm: object|null, text: string }}
+ *   content is a plain string when no image is detected, else a multimodal
+ *   array; visionLlm is the per-turn provider/model/baseUrl/apiKey override
+ *   (null when there are no images); text is the path-stripped prompt.
+ */
+export function prepareVisionTurn(text, llm = {}, deps = {}) {
+  const detected = detectImagePaths(text, deps);
+  if (!detected.images.length) {
+    return { content: text, images: [], visionLlm: null, text };
+  }
+  const finalText = detected.text || "Please look at the attached image(s).";
+  const resolved = resolveImages(detected.images, deps); // throws on bad ext
+  return {
+    content: buildUserContent(finalText, resolved),
+    images: detected.images,
+    visionLlm: resolveVisionLlm({ hasImage: true, flags: {}, llm }),
+    text: finalText,
+  };
+}
+
 /**
  * Build a user-message `content`: the plain string when there are no images,
  * else an OpenAI-style multimodal array (the internal representation).

package/src/repl/agent-repl.js

@@ -81,6 +81,7 @@ import {
 } from "../runtime/agent-core.js";
 import { formatBackgroundTasks } from "./tasks-status.js";
 import { expandFileRefsAsync } from "../runtime/file-ref-expander.js";
+import { prepareVisionTurn } from "../lib/image-input.js";
 import { composeSystemPrompt } from "../runtime/system-prompt.js";
 import {
   makeFallbackChatFn,
@@ -279,6 +280,16 @@ export async function startAgentRepl(options = {}) {
   let _sessionTier = "strict";
   const baseUrl = options.baseUrl || "http://localhost:11434";
   const apiKey = options.apiKey || null;
+  // Configured vision model (config.llm.visionModel) — used when a turn carries
+  // an auto-detected image path so the REPL switches to a vision-capable model
+  // for that turn only (resolveVisionLlm falls back to the default when unset).
+  let _visionModel;
+  try {
+    const { loadConfig } = await import("../lib/config-manager.js");
+    _visionModel = loadConfig()?.llm?.visionModel || undefined;
+  } catch {
+    /* optional — resolveVisionLlm falls back to DEFAULT_VISION_MODEL */
+  }
   // Extra workspace roots (--add-dir): advertised in the system prompt and
   // spanned by search_files.
   const additionalDirectories = Array.isArray(options.additionalDirectories)
@@ -3075,8 +3086,40 @@ export async function startAgentRepl(options = {}) {
       // optional polish — never fail the turn over it
     }
 
+    // Claude-Code-style: auto-attach local image paths typed in the message so
+    // "describe ./shot.png" reads the image via the vision model (same as the
+    // chat panels). CC_AUTO_IMAGE=0 opts out. `_visionLlm` (truthy on an image
+    // turn) overrides this turn's provider/model/baseUrl/apiKey below. The
+    // composition is the unit-tested `prepareVisionTurn` helper.
+    let _visionLlm = null;
+    let _userMessageContent = userContent;
+    if (process.env.CC_AUTO_IMAGE !== "0") {
+      try {
+        const turn = prepareVisionTurn(userContent, {
+          provider,
+          baseUrl,
+          apiKey,
+          visionModel: _visionModel,
+        });
+        if (turn.visionLlm) {
+          _userMessageContent = turn.content;
+          _visionLlm = turn.visionLlm;
+          logger.info(
+            chalk.gray(
+              `[image] ${turn.images.length} attached → vision model ${turn.visionLlm.model}`,
+            ),
+          );
+        }
+      } catch (e) {
+        // Bad attachment (e.g. unreadable file) → send as plain text.
+        _visionLlm = null;
+        _userMessageContent = userContent;
+        logger.info(chalk.yellow(`[image] ${e.message} — sending as text`));
+      }
+    }
+
     // Add user message
-    messages.push({ role: "user", content: userContent });
+    messages.push({ role: "user", content: _userMessageContent });
 
     // Slot-filling: detect intent and fill missing parameters interactively
     try {
@@ -3202,12 +3245,14 @@ export async function startAgentRepl(options = {}) {
             ),
           ),
         signal: _turnAbort.signal,
-        provider,
-        model: activeModel,
+        // On an auto-detected image turn, switch to the vision LLM for this
+        // turn only (provider/baseUrl/apiKey unchanged, model → vision model).
+        provider: _visionLlm ? _visionLlm.provider : provider,
+        model: _visionLlm ? _visionLlm.model : activeModel,
         thinking,
         thinkingBudget,
-        baseUrl,
-        apiKey,
+        baseUrl: _visionLlm ? _visionLlm.baseUrl : baseUrl,
+        apiKey: _visionLlm ? _visionLlm.apiKey : apiKey,
         contextEngine,
         iterationBudget,
         sessionId,

package/src/runtime/headless-stream.js

@@ -20,6 +20,7 @@ import { bootstrap } from "./bootstrap.js";
 import { buildSystemPrompt, agentLoop as coreAgentLoop } from "./agent-core.js";
 import { composeSystemPrompt } from "./system-prompt.js";
 import { expandFileRefsAsync } from "./file-ref-expander.js";
+import { detectImagePaths } from "../lib/image-input.js";
 import {
   resolveAgentMcp,
   resolvePermissionPromptTool,
@@ -113,9 +114,24 @@ export function parseInputEvent(line) {
   // the same image-input pipeline as `cc agent --image`.
   const rawImages =
     obj && typeof obj === "object" ? obj.images || msg.images : null;
-  const images = Array.isArray(rawImages)
-    ? rawImages.filter((p) => typeof p === "string" && p.trim()).slice(0, 8)
+  let images = Array.isArray(rawImages)
+    ? rawImages.filter((p) => typeof p === "string" && p.trim())
     : [];
+  // Claude-Code-style: auto-attach local image-file paths the user typed into
+  // the message (so "describe ./shot.png" reads the image, like Claude Code).
+  // Opt out with CC_AUTO_IMAGE=0. Explicit `images` (paste) still win.
+  if (
+    typeof content === "string" &&
+    content.trim() &&
+    process.env.CC_AUTO_IMAGE !== "0"
+  ) {
+    const detected = detectImagePaths(content);
+    if (detected.images.length) {
+      images = [...images, ...detected.images];
+      content = detected.text;
+    }
+  }
+  images = [...new Set(images)].slice(0, 8);
   if (typeof content !== "string" || !content.trim()) {
     // An image-only turn is valid — give the model something to act on.
     if (images.length)

package/src/assets/web-panel/assets/Terminal-B6R_Mv2d.js

@@ -1,3 +0,0 @@
-const __vite__mapDeps=(i,m=__vite__mapDeps,d=(m.f||(m.f=["./xterm-BZcWGsqw.js","./markdown-CsiA8-E5.js","./markdown-Dfs9RUU9.css","./addon-fit-CK6X9sAG.js","./xterm-DFuMZ0ql.css"])))=>i.map(i=>d[i]);
-import{u as fe,_ as me,d as U,e as V}from"./index-B8GMkcIp.js";import{I as z,J as x,U as _,Q as q,S as K,K as T,V as R,c as A,F as Q,Z as Y,R as E,o as Z,f as G,w as $,n as D,b as ee,r as S,P as y,_ as ve,a2 as we,a3 as he,a4 as _e}from"./vendor-BvqAck49.js";import{R as ye,b as pe,as as ge}from"./icons-DP3uiYxy.js";const k=new Map,C=new Map,I=new Set,N=new Set;let te=!1;function xe(c){te||(te=!0,c.onMessage(o=>{if(!(!o||typeof o.type!="string")){if(o.type==="terminal.stdout"){const{sessionId:s,data:e,seq:i}=o.payload||{};if(!s)return;let w;try{const d=atob(e||""),f=new Uint8Array(d.length);for(let n=0;n<d.length;n++)f[n]=d.charCodeAt(n);w=new TextDecoder("utf-8").decode(f)}catch{w=""}const u={sessionId:s,data:w,seq:i};k.get(s)?.forEach(d=>d(u)),I.forEach(d=>d(u))}else if(o.type==="terminal.exit"){const{sessionId:s,exitCode:e,signal:i}=o.payload||{};if(!s)return;const w={sessionId:s,exitCode:e,signal:i};C.get(s)?.forEach(u=>u(w)),N.forEach(u=>u(w))}}}))}function ne(c){const o=new TextEncoder().encode(c);let s="";for(let e=0;e<o.length;e++)s+=String.fromCharCode(o[e]);return btoa(s)}function ae(c){const o=atob(c||""),s=new Uint8Array(o.length);for(let e=0;e<o.length;e++)s[e]=o.charCodeAt(e);return new TextDecoder("utf-8").decode(s)}function oe(){const c=fe();xe(c);async function o(n={}){const a=await c.sendRaw({type:"terminal.create",payload:{shell:n.shell,cwd:n.cwd,env:n.env,cols:n.cols,rows:n.rows}});if(a.ok===!1)throw new Error(a.error||"terminal_create_failed");return a.result??a}async function s(){const n=await c.sendRaw({type:"terminal.list",payload:{}});if(n.ok===!1)throw new Error(n.error||"terminal_list_failed");const a=n.result??n;return Array.isArray(a.sessions)?a.sessions:[]}async function e(n,a){const r=await c.sendRaw({type:"terminal.stdin",payload:{sessionId:n,data:ne(String(a))}});if(r.ok===!1)throw new Error(r.error||"terminal_stdin_failed");return r.result??r}async function i(n,a,r){const h=await c.sendRaw({type:"terminal.resize",payload:{sessionId:n,cols:a,rows:r}});if(h.ok===!1)throw new Error(h.error||"terminal_resize_failed");return h.result??h}async function w(n){const a=await c.sendRaw({type:"terminal.close",payload:{sessionId:n}});if(a.ok===!1)throw new Error(a.error||"terminal_close_failed");return a.result??a}async function u(n,a=0){const r=await c.sendRaw({type:"terminal.history",payload:{sessionId:n,fromSeq:a}});if(r.ok===!1)throw new Error(r.error||"terminal_history_failed");const h=r.result??r;return{truncated:!!h.truncated,chunks:(h.chunks||[]).map(L=>({seq:L.seq,data:ae(L.data)}))}}function d(n,a){return n?(k.has(n)||k.set(n,new Set),k.get(n).add(a),()=>{k.get(n)?.delete(a),k.get(n)?.size===0&&k.delete(n)}):(I.add(a),()=>I.delete(a))}function f(n,a){return n?(C.has(n)||C.set(n,new Set),C.get(n).add(a),()=>{C.get(n)?.delete(a),C.get(n)?.size===0&&C.delete(n)}):(N.add(a),()=>N.delete(a))}return{create:o,list:s,stdin:e,resize:i,close:w,history:u,onStdout:d,onExit:f,_internal:{stdoutSubs:k,exitSubs:C,toBase64Utf8:ne,fromBase64Utf8:ae}}}const Se={__name:"Terminal",setup(c,{expose:o}){o();const s=oe(),e=S([]),i=S(null),w=S("pwsh"),u=S(!1),d=S(!1),f=S(""),n=S(""),a=S([]),r=[{value:"pwsh",label:"PowerShell"},{value:"cmd",label:"CMD"},{value:"bash",label:"Bash"},{value:"wsl",label:"WSL"}],h=ee(()=>e.value.find(t=>t.id===i.value));function L(t){return t?t.slice(0,8):""}let O=null,M=null;async function F(){if(O)return{xtermMod:O,fitAddonMod:M};try{O=await V(()=>import("./xterm-BZcWGsqw.js").then(t=>t.x),__vite__mapDeps([0,1,2]),import.meta.url),M=await V(()=>import("./addon-fit-CK6X9sAG.js").then(t=>t.a),__vite__mapDeps([3,1,2]),import.meta.url),await V(()=>Promise.resolve({}),__vite__mapDeps([4]),import.meta.url)}catch(t){throw n.value="xterm 资源加载失败："+(t?.message||"未知错误"),t}return{xtermMod:O,fitAddonMod:M}}async function B(t){await D();const{xtermMod:l,fitAddonMod:p}=await F(),b=a.value.find(v=>v?.dataset?.sessionId===t.id);if(!b)return;const m=new l.Terminal({cursorBlink:!0,fontFamily:'Consolas, "Courier New", monospace',fontSize:13,theme:{background:"#1e1e1e",foreground:"#d4d4d4"},convertEol:!1}),P=new p.FitAddon;m.loadAddon(P),m.open(b);try{P.fit()}catch{}t.xterm=m,t.fitAddon=P;const ie=m.onData(v=>{s.stdin(t.id,v).catch(g=>{String(g?.message||"").includes("dangerous_keyword_blocked")?U.warning("该命令被桌面端拦截（高危关键字）"):U.error("stdin 失败: "+(g?.message||g))})}),ce=s.onStdout(t.id,({data:v,seq:g})=>{t.lastSeq=g,m.write(v)}),de=s.onExit(t.id,({exitCode:v,signal:g})=>{t.alive=!1,t.exitCode=v,t.signal=g,m.writeln(`\r
-\x1B[33m[session exited, code=${v}, signal=${g??"-"}]\x1B[0m`)});t.offs=()=>{try{ie.dispose?.()}catch{}ce(),de()};try{const{chunks:v,truncated:g}=await s.history(t.id,0);g&&m.writeln("\x1B[2m[history truncated — earlier output was evicted]\x1B[0m");for(const J of v)m.write(J.data),t.lastSeq=J.seq}catch(v){m.writeln(`\x1B[31m[history fetch failed: ${v?.message||v}]\x1B[0m`)}const H=new ResizeObserver(()=>{try{P.fit(),s.resize(t.id,m.cols,m.rows).catch(()=>{})}catch{}});H.observe(b);const ue=t.offs;t.offs=()=>{try{H.disconnect()}catch{}ue()}}async function re(){u.value=!0,f.value="";try{const t=await s.create({shell:w.value,cols:80,rows:24}),l={id:t.sessionId,shell:t.shell,cwd:"",alive:!0,lastSeq:0,exitCode:null,xterm:null,fitAddon:null,offs:()=>{}};e.value.push(l),i.value=l.id,await B(l)}catch(t){f.value=t?.message||String(t)}finally{u.value=!1}}async function se(t){try{await s.close(t)}catch(l){f.value=l?.message||String(l)}setTimeout(()=>W(t),500)}function W(t){const l=e.value.findIndex(b=>b.id===t);if(l===-1)return;const p=e.value[l];try{p.offs?.()}catch{}try{p.xterm?.dispose?.()}catch{}e.value.splice(l,1),i.value===t&&(i.value=e.value[0]?.id||null)}function le(t){i.value=t,D(()=>{const l=e.value.find(p=>p.id===t);try{l?.fitAddon?.fit()}catch{}})}async function X(){d.value=!0,f.value="";try{const t=await s.list();for(const l of t){const p=e.value.find(m=>m.id===l.id);if(p){p.alive=l.alive,p.lastSeq=l.lastSeq;continue}const b={id:l.id,shell:l.shell,cwd:l.cwd,alive:l.alive,lastSeq:l.lastSeq,exitCode:null,xterm:null,fitAddon:null,offs:()=>{}};e.value.push(b),await B(b)}!i.value&&e.value.length>0&&(i.value=e.value[0].id)}catch(t){f.value=t?.message||String(t)}finally{d.value=!1}}Z(async()=>{await X()}),G(()=>{for(const t of e.value){try{t.offs?.()}catch{}try{t.xterm?.dispose?.()}catch{}}}),$(i,()=>{D(()=>{const t=h.value;try{t?.fitAddon?.fit()}catch{}})});const j={term:s,sessions:e,activeId:i,newShell:w,creating:u,loadingList:d,error:f,warning:n,xtermContainers:a,shellOptions:r,active:h,shortId:L,get xtermMod(){return O},set xtermMod(t){O=t},get fitAddonMod(){return M},set fitAddonMod(t){M=t},loadXterm:F,mountXterm:B,onCreate:re,onClose:se,removeSession:W,activate:le,refreshList:X,ref:S,computed:ee,onMounted:Z,onBeforeUnmount:G,nextTick:D,watch:$,get message(){return U},get PlusOutlined(){return ge},get CloseOutlined(){return pe},get ReloadOutlined(){return ye},get useTerminal(){return oe}};return Object.defineProperty(j,"__isScriptSetup",{enumerable:!1,value:!0}),j}},be={class:"terminal-page"},ke={class:"terminal-header"},Ce={class:"page-sub"},Ae={class:"terminal-body"},Ee={class:"session-tabs"},Oe=["onClick"],Te={class:"session-shell"},Re={class:"session-id"},Me={key:0,class:"session-empty"},ze={class:"xterm-host"},Le=["data-session-id"],Pe={key:0,class:"xterm-placeholder"},qe={key:1,class:"terminal-footer"},De={key:0,class:"footer-exit"};function Be(c,o,s,e,i,w){const u=z("a-tag"),d=z("a-select"),f=z("a-button"),n=z("a-space"),a=z("a-alert");return y(),x("div",be,[_("div",ke,[_("div",null,[o[3]||(o[3]=_("h2",{class:"page-title"},"远程终端",-1)),_("p",Ce,[o[2]||(o[2]=q(" 桌面端托管的 PTY 会话；Android 端可远程操控同一通道 ",-1)),e.warning?(y(),K(u,{key:0,color:"orange",style:{"margin-left":"8px"}},{default:T(()=>[q(E(e.warning),1)]),_:1})):R("v-if",!0)])]),A(n,null,{default:T(()=>[A(d,{value:e.newShell,"onUpdate:value":o[0]||(o[0]=r=>e.newShell=r),style:{width:"130px"},size:"small",options:e.shellOptions},null,8,["value"]),A(f,{type:"primary",size:"small",loading:e.creating,onClick:e.onCreate},{icon:T(()=>[A(e.PlusOutlined)]),default:T(()=>[o[4]||(o[4]=q(" 新会话 ",-1))]),_:1},8,["loading"]),A(f,{size:"small",loading:e.loadingList,onClick:e.refreshList},{icon:T(()=>[A(e.ReloadOutlined)]),default:T(()=>[o[5]||(o[5]=q(" 刷新 ",-1))]),_:1},8,["loading"])]),_:1})]),e.error?(y(),K(a,{key:0,message:e.error,type:"error","show-icon":"",closable:"",style:{"margin-bottom":"12px"},onClose:o[1]||(o[1]=r=>e.error="")},null,8,["message"])):R("v-if",!0),_("div",Ae,[_("div",Ee,[(y(!0),x(Q,null,Y(e.sessions,r=>(y(),x("div",{key:r.id,class:ve(["session-tab",{active:r.id===e.activeId,dead:!r.alive}]),onClick:h=>e.activate(r.id)},[_("span",Te,E(r.shell),1),_("span",Re,E(e.shortId(r.id)),1),A(e.CloseOutlined,{class:"session-close",onClick:we(h=>e.onClose(r.id),["stop"])},null,8,["onClick"])],10,Oe))),128)),e.sessions.length===0?(y(),x("div",Me,' 点击 "新会话" 创建第一个终端 ')):R("v-if",!0)]),_("div",ze,[(y(!0),x(Q,null,Y(e.sessions,r=>he((y(),x("div",{key:r.id,ref_for:!0,ref:"xtermContainers","data-session-id":r.id,class:"xterm-container"},null,8,Le)),[[_e,r.id===e.activeId]])),128)),e.sessions.length===0?(y(),x("div",Pe,[...o[6]||(o[6]=[_("span",null,"无活跃会话",-1)])])):R("v-if",!0)])]),e.active?(y(),x("div",qe,[_("span",null,E(e.active.shell)+" · "+E(e.active.cwd||"(默认 cwd)")+" · seq "+E(e.active.lastSeq),1),e.active.alive?R("v-if",!0):(y(),x("span",De,"已退出 (code="+E(e.active.exitCode??"-")+")",1))])):R("v-if",!0)])}const Ne=me(Se,[["render",Be],["__scopeId","data-v-65366a29"],["__file","/tmp/cc-web-panel-dUwIHq/repo/packages/web-panel/src/views/Terminal.vue"]]);export{Ne as default};

package/src/assets/web-panel/assets/devWarning-cfHzirPw.js

@@ -1 +0,0 @@
-import{O as r}from"./index-B8GMkcIp.js";const o=((n,a,e)=>{r(n,`[ant-design-vue: ${a}] ${e}`)});export{o as d};

package/src/assets/web-panel/assets/index-CWB8kLmE.js

@@ -1 +0,0 @@
-import{A as o}from"./Row-DEHR17nP.js";import{U as t}from"./index-B8GMkcIp.js";import"./vendor-BvqAck49.js";import"./responsiveObserve-BvwPauMH.js";import"./useFlexGapSupport-Cp_Cx5AB.js";import"./styleChecker-EtdQYqac.js";import"./index-DBfaQNOv.js";import"./icons-DP3uiYxy.js";const l=t(o);export{l as default};

package/src/assets/web-panel/assets/index-DtaI9LEB.js

@@ -1 +0,0 @@
-import{C as o}from"./Col-BpLy4N8M.js";import{U as t}from"./index-B8GMkcIp.js";import"./vendor-BvqAck49.js";import"./index-DBfaQNOv.js";import"./icons-DP3uiYxy.js";const s=t(o);export{s as default};