@livx.cc/agentx 0.99.84 → 0.99.85

package/README.md

@@ -64,7 +64,7 @@ await fs.createDir('/src');
 await fs.writeFile('/src/x.ts', 'export const add = (a,b) => a - b;\n');
 
 const ai = new AIClient({ apiKeys: { anthropic: process.env.ANTHROPIC_API_KEY } });
-const res = await new Agent({ ai, fs, model: 'anthropic/claude-sonnet-4-6' })
+const res = await new Agent({ ai, fs, model: 'anthropic/sonnet' })
   .run('Fix the add bug in /src/x.ts');
 
 console.log(res.finishReason, await fs.readFile('/src/x.ts'));
@@ -77,7 +77,7 @@ console.log(res.finishReason, await fs.readFile('/src/x.ts'));
 - **`Edit`** — exact unique-substring replace, with a read-before-edit staleness guard.
 - **`Grep`/`Glob`/`Write`/`MultiEdit`** — structured, typed results straight from the VFS (no `bash` parsing). The selectable tool set the self-evolution loop mutates over.
 - **`TodoWrite`** — a planning scratchpad; **`Task`** — spawn a depth-limited child agent over the VFS (`subagents: true`); **`SlashCommand`** — reusable prompt templates from `<dir>/*.md` (`commandsDir`); plus a real **MCP client** (`src/mcp.client.ts`, node-only — stdio/HTTP JSON-RPC handshake + discovery) that feeds the edge-safe **MCP adapter** (`mcpToolsToAgentTools`), so any MCP server's tools become agent tools.
-- **`WebFetch`/`WebSearch`** — fetch a URL as readable text, or search the web. **Keyless by default** (WebSearch uses DuckDuckGo; `provider: 'auto'` upgrades by key presence — Firecrawl `FIRECRAWL_API_KEY` > Tavily `TAVILY_API_KEY`) and **auto-enabled in the CLI**. Factory-built with an injectable `fetch`, so they stay edge-portable and testable. (In the library they're opt-in by name: `tools: [...,'WebSearch']`.) **`WebSearchAnthropic`** is a separate provider-pinned tool — Anthropic's native search relayed through a cheap model (`claude-haiku-4-5`); CLI-default-on when an Anthropic key resolves, API-billed per call, opt out via `anthropicWebSearch: false`.
+- **`WebFetch`/`WebSearch`** — fetch a URL as readable text, or search the web. **Keyless by default** (WebSearch uses DuckDuckGo; `provider: 'auto'` upgrades by key presence — Firecrawl `FIRECRAWL_API_KEY` > Tavily `TAVILY_API_KEY`) and **auto-enabled in the CLI**. Factory-built with an injectable `fetch`, so they stay edge-portable and testable. (In the library they're opt-in by name: `tools: [...,'WebSearch']`.) **`WebSearchAnthropic`** is a separate provider-pinned tool — Anthropic's native search relayed through a cheap model (the newest Haiku); CLI-default-on when an Anthropic key resolves, API-billed per call, opt out via `anthropicWebSearch: false`.
 - **Oversized-output pagination** — any tool result over a byte ceiling (`maxToolResultBytes`, default 60k) is cropped to page 1 with a marker (refine the query / read further), so one big `Grep`/`Read`/MCP/web result can't blow the context window. In the CLI (**on by default**; `--no-scratch` to disable) the full output instead spills **losslessly** to a **scratch** file and the model recovers specifics via `Grep`/`Read` or **`Ask`** — a cheap, context-isolated peek that returns just the answer (the raw blob never re-enters context).
 
 ## Agentic subsystems

package/dist/cli.js

@@ -1000,6 +1000,34 @@ var init_todo = __esm({
   }
 });
 
+// src/models.ts
+import { claudeModelAliases } from "ai.libx.js";
+function latestClaude(family, provider = "anthropic") {
+  const id = MODEL_ALIASES[family];
+  if (!id) throw new Error(`ai.libx.js catalog has no Claude ${family} model`);
+  return `${provider}/${id}`;
+}
+function resolveModelAlias(input) {
+  const slash = input.indexOf("/");
+  const prefix = slash === -1 ? "" : input.slice(0, slash + 1);
+  const rest = slash === -1 ? input : input.slice(slash + 1);
+  if (prefix === "claude-code/") return input;
+  return prefix + (MODEL_ALIASES[rest.toLowerCase()] ?? rest);
+}
+function modelProvider(model) {
+  const i = model.indexOf("/");
+  if (i !== -1) return model.slice(0, i);
+  if (/^claude-/i.test(model)) return "anthropic";
+  return `?${model}`;
+}
+var MODEL_ALIASES;
+var init_models = __esm({
+  "src/models.ts"() {
+    "use strict";
+    MODEL_ALIASES = { ...claudeModelAliases };
+  }
+});
+
 // src/tools.web.ts
 function htmlToText(html) {
   let s = html.replace(/<script[\s\S]*?<\/script>/gi, " ").replace(/<style[\s\S]*?<\/style>/gi, " ").replace(/<title[\s\S]*?<\/title>/gi, " ").replace(/<noscript[\s\S]*?<\/noscript>/gi, " ").replace(/<textarea[\s\S]*?<\/textarea>/gi, " ").replace(/<!--[\s\S]*?-->/g, " ").replace(/<\/(p|div|li|h[1-6]|tr|section|article|header|footer|nav)>/gi, "\n").replace(/<br\s*\/?>/gi, "\n").replace(/<[^>]+>/g, " ");
@@ -1221,7 +1249,7 @@ function makeWebSearchTool(options = {}) {
         if (provider === "anthropic") {
           const akey = options.anthropicApiKey ?? process.env.ANTHROPIC_API_KEY;
           if (!akey) return "Error: WebSearchAnthropic requires ANTHROPIC_API_KEY (set in env)";
-          return await anthropicSearch(q2, { key: akey, model: options.model ?? "claude-haiku-4-5", fetch: doFetch, signal: ctl.signal });
+          return await anthropicSearch(q2, { key: akey, model: options.model ?? latestClaude("haiku").slice("anthropic/".length), fetch: doFetch, signal: ctl.signal });
         }
         if (useFirecrawl) {
           if (!fcKey) return "Error: Firecrawl provider selected but FIRECRAWL_API_KEY is not set";
@@ -1263,6 +1291,7 @@ var init_tools_web = __esm({
   "src/tools.web.ts"() {
     "use strict";
     init_logging();
+    init_models();
     log4 = forComponent("web");
     webFetchTool = makeWebFetchTool();
     webSearchTool = makeWebSearchTool();
@@ -3983,36 +4012,10 @@ ${sections.join("\n\n---\n\n")}`;
 
 // src/subagent.ts
 init_tools();
+init_models();
 import { mkdirSync as mkdirSync2, writeFileSync as writeFileSync2 } from "fs";
 import { join as join4 } from "path";
 
-// src/models.ts
-var MODEL_ALIASES = {
-  fable: "claude-fable-5",
-  "fable-5": "claude-fable-5",
-  opus: "claude-opus-5-5",
-  "opus-5-5": "claude-opus-5-5",
-  "opus-5": "claude-opus-5",
-  "opus-4-8": "claude-opus-4-8",
-  "opus-4-7": "claude-opus-4-7",
-  "opus-4-6": "claude-opus-4-6",
-  sonnet: "claude-sonnet-4-6",
-  haiku: "claude-haiku-4-5-20251001"
-};
-function resolveModelAlias(input) {
-  const slash = input.indexOf("/");
-  const prefix = slash === -1 ? "" : input.slice(0, slash + 1);
-  const rest = slash === -1 ? input : input.slice(slash + 1);
-  if (prefix === "claude-code/") return input;
-  return prefix + (MODEL_ALIASES[rest.toLowerCase()] ?? rest);
-}
-function modelProvider(model) {
-  const i = model.indexOf("/");
-  if (i !== -1) return model.slice(0, i);
-  if (/^claude-/i.test(model)) return "anthropic";
-  return `?${model}`;
-}
-
 // src/permissions.ts
 init_tools_structured();
 function commandMatches(glob, cmd) {
@@ -4500,6 +4503,9 @@ ${r.error ? `ERROR: ${r.error}` : r.text || "(no summary)"}${extra}`);
   };
 }
 
+// src/Agent.ts
+init_models();
+
 // src/agents.ts
 function parseAgentFrontmatter(md) {
   const m = md.match(/^---\s*\n([\s\S]*?)\n---\s*\n?/);
@@ -4635,7 +4641,7 @@ var AgentOptions = class {
    *  browser build (which always passes its own fs) never pulls in `node:fs`. Pass `fs` explicitly to
    *  use any other backend (and to keep full control of the jail). */
   fs;
-  model = "anthropic/claude-sonnet-4-6";
+  model = latestClaude("sonnet");
   systemPrompt = "You are a coding agent operating on a virtual filesystem. Use the `bash` tool to explore (ls, grep, find, cat) and the `Read`/`Edit` tools to view and modify files. Always Read a file before Editing it \u2014 and before overwriting it with `Write` or a whole-file `ApplyEdits`: Write replaces the ENTIRE file, so blindly writing over an existing one destroys its unseen contents. If a file already exists, Read it first, then merge your changes (or Edit the specific part). When multiple tool calls are independent (e.g. several Reads or Greps with no dependency between them), issue them together in a single turn rather than one at a time. When the task is complete, reply with a short summary and make no further tool calls.\nRecovery: every turn's file changes are checkpointed. If the user asks to undo/revert/recover edits (including a file you overwrote), tell them to run `/rewind` (or `/undo` for the last turn) to restore the working tree \u2014 checkpoints cover files you edited even in .gitignored dirs. You cannot run `/rewind` yourself; surface it as the recovery path.\nWhen working with a knowledge base or document workspace: use `RepoMap` with scope \"docs\" to see all document headings and summaries (orient first); use `MemorySearch` to find memories by content when you don't know the exact slug; use `Recall` with multiple slugs to batch-load related facts in one call. Prefer targeted loads over reading everything \u2014 orient \u2192 search \u2192 load \u2192 act.";
   tools = defaultTools();
   /** claude-code/* only: which agentx tools reach it as the in-process `agentx` MCP server (ccBridge.ts). 'memory' =
@@ -6070,6 +6076,7 @@ function toolKey(name) {
 }
 
 // src/index.ts
+init_models();
 init_NodeDiskFilesystem();
 
 // src/BodDbFilesystem.ts
@@ -6778,6 +6785,7 @@ function digestRun(messages, maxChars) {
 import { MemFilesystem as MemFilesystem2 } from "@livx.cc/wcli/core";
 init_llm();
 init_logging();
+init_models();
 
 // node_modules/@bod.ee/voice/src/core/logging.ts
 function consoleForComponent(name) {
@@ -9650,9 +9658,9 @@ var DuplexAgentOptions = class {
   // mislabels gpt-oss harmony channels under load, leaking raw analysis into the spoken `final` channel
   // (and misfiring Hold). 120b is the same price tier (~$0.15/$0.60) — the quality/cost trade is free.
   reflexModel = "groq/openai/gpt-oss-120b";
-  actModel = "anthropic/claude-sonnet-4-6";
+  actModel = latestClaude("sonnet");
   /** Premium reasoning model. Set to `false` to disable the Think tier entirely. */
-  thinkModel = "anthropic/claude-opus-4-8";
+  thinkModel = latestClaude("opus");
   /** Per-worker providerOptions, derived from the worker's actual model at spawn time (IoC — keeps duplex
    *  provider-agnostic). Workers override the reflex/main model, so provider-specific options (e.g. cursor's
    *  cwd/cursorSession) must be recomputed for the worker's model, never inherited from the main template —
@@ -11230,11 +11238,12 @@ Another agent just implemented the above. Independently check the CURRENT state
 // src/modelRouter.ts
 init_llm();
 init_logging();
+init_models();
 var log22 = forComponent("ModelRouter");
 var ModelRouterOptions = class {
   classifier = jevClassifier();
   /** Capability tiers → model id (any provider). basic/standard run the main session; premium is only ever an offload. */
-  models = { basic: "anthropic/claude-haiku-4-5-20251001", standard: "anthropic/claude-sonnet-4-6", premium: "anthropic/claude-opus-5-5" };
+  models = { basic: latestClaude("haiku"), standard: latestClaude("sonnet"), premium: latestClaude("opus") };
   /** Escalate to a stronger tier when it holds at least this much probability mass (upward bias). */
   threshold = 0.25;
   /** P(same thread) at or above which the previous premium side-conversation is resumed. */
@@ -12533,7 +12542,7 @@ Reference files in them by their mount path (the left side).`;
     ai: o.ai,
     jobs: jobRegistry,
     fs,
-    model: o.model ?? "anthropic/claude-sonnet-4-6",
+    model: o.model ?? latestClaude("sonnet"),
     // PDF reads (disk mode only — VFS paths aren't real files): poppler's pdftotext when installed.
     ...!virtual ? { pdfText: pdfTextViaPoppler } : {},
     // Anchor cursor to the launch dir (its adapter defaults to TMPDIR otherwise) and forward the
@@ -12584,7 +12593,7 @@ Windows paths: file tools accept ${cwd} style paths and the same path written ${
       if (o.anthropicWebSearch !== false && o.anthropicKey && !requested.includes("WebSearchAnthropic")) {
         tail.push(makeWebSearchAnthropicTool({ anthropicApiKey: o.anthropicKey }));
       }
-      if (scratch) tail.push(makeAskTool({ fs, ai: o.ai, model: o.scratchAskModel ?? o.model ?? "anthropic/claude-sonnet-4-6", dir: scratchDir }));
+      if (scratch) tail.push(makeAskTool({ fs, ai: o.ai, model: o.scratchAskModel ?? o.model ?? latestClaude("sonnet"), dir: scratchDir }));
       tail.push(makeNotifyTool());
       if (!realShell.length) return [...base, ...tail];
       const filtered = base.filter((t) => t.name !== "bash");
@@ -16425,7 +16434,7 @@ var HELP = `agentx \u2014 agent runtime CLI
   agentx -c "<task>"              continue the most recent session
 
 Flags:
-  -m, --model <id>     model (default auto: per-turn haiku/sonnet + opus offload when Jev is configured, else anthropic/claude-sonnet-4-6); bare -m or -m list = list all models
+  -m, --model <id>     model (default auto: per-turn haiku/sonnet + opus offload when Jev is configured, else the newest anthropic Sonnet); bare -m or -m list = list all models
   -C, --cwd <dir>      working dir / project root (default .)
   -c, --continue       resume the most recent session in this dir
   -r, --resume [id]    resume a session \u2014 with an id, or bare for an interactive picker
@@ -16457,7 +16466,7 @@ Flags:
                        with SONIOX_API_KEY + CARTESIA_API_KEY(+VOICE_ID) set: real voice I/O \u2014 mic in,
                        spoken replies out (echo-cancelled; speak over it to interrupt)
   --voice-model <id>   with --duplex: the fast voice model (default groq/openai/gpt-oss-120b)
-  --think-model <id>   with --duplex: the premium deep-reasoning model (default anthropic/claude-opus-4-6)
+  --think-model <id>   with --duplex: the premium deep-reasoning model (default: the newest anthropic Opus)
   --no-think           with --duplex: disable the Think tier (Act handles everything)
   --voice-system-prompt <text>  with --duplex: appended to the voice (reflex) prompt only (--append-system-prompt \u2192 workers)
   --add-dir <path>     mount another directory into the workspace (repeatable; disk mode only)
@@ -16508,8 +16517,8 @@ REPL shortcuts: Shift+Tab cycles permission posture (ask \u2192 accept-edits \u2
 REPL steer/stash: type while a turn is running \u2192 Enter STEERS it (the text reaches the model at its next step, alongside the next tool result). Alt+S (or Ctrl+S) with text stashes it for AFTER the turn instead; on an empty prompt pops the next stashed entry for editing.
 REPL editing (emacs/readline): Ctrl-A/E line start/end \xB7 Ctrl-B/F char \xB7 Alt-B/F or Alt/Ctrl-\u2190/\u2192 word \xB7 Ctrl-W kill word \xB7 Ctrl-U/K kill to start/end \xB7 Ctrl-Y yank \xB7 Ctrl-_ undo \xB7 Alt-D kill word fwd \xB7 Ctrl-L clear screen. Set editorMode:'vim' (or /config) for modal vim editing.
 REPL paste: large/multi-line pastes collapse to a [Pasted #N "\u2026" +N lines] preview (expands on send); a pasted image/file path attaches as [Image]/[File]; Ctrl-V or /paste grabs a clipboard image (macOS).`;
-function newestModel() {
-  return listModels().slice().sort((a, b) => (getModelInfo(b)?.releaseDate ?? "").localeCompare(getModelInfo(a)?.releaseDate ?? ""))[0] ?? "";
+function newestModel(provider) {
+  return listModels(provider).slice().sort((a, b) => (getModelInfo(b)?.releaseDate ?? "").localeCompare(getModelInfo(a)?.releaseDate ?? ""))[0] ?? "";
 }
 function printModels() {
   const byProvider = /* @__PURE__ */ new Map();
@@ -16533,7 +16542,8 @@ function resolveModelOrNewest(model) {
   const want = model ?? new AgentOptions().model;
   const resolved = resolveModel(want);
   if (isModelSupported(resolved)) return resolved;
-  const fallback = newestModel();
+  const provider = getProviderFromModel(want) ?? void 0;
+  const fallback = newestModel(provider && listModels(provider).length ? provider : void 0);
   if (!fallback) return want;
   err(yellow(`  \u26A0 model "${want}" not available \u2014 using ${fallback}
 `));
@@ -16603,7 +16613,7 @@ function duplexModels(o) {
   let reflexModel = asked ? resolveModelOrNewest(asked) : void 0;
   const reflex = reflexModel ?? dflt.reflexModel;
   if (!modelUsable(reflex, o.apiKeys)) {
-    const to = anthropic ? "anthropic/claude-haiku-4-5" : cc ? "claude-code/haiku" : o.actModel.startsWith("cursor/") ? void 0 : o.actModel;
+    const to = anthropic ? latestClaude("haiku") : cc ? "claude-code/haiku" : o.actModel.startsWith("cursor/") ? void 0 : o.actModel;
     if (to) reflexModel = to;
     notes.push(`voice model ${reflex} can't run here (${why(reflex)})${to ? ` \u2014 using ${to}` : ""}`);
   }
@@ -17060,7 +17070,7 @@ async function evaluateGoal(ai, condition, transcript, log35) {
   }).join("\n---\n");
   try {
     const r = await ai.chat({
-      model: "anthropic/claude-haiku-4-5",
+      model: latestClaude("haiku"),
       stream: false,
       messages: [
         { role: "system", content: 'You judge whether a goal condition has been met based on conversation transcript. Respond ONLY with JSON: {"met": boolean, "reason": "one sentence"}. Be strict \u2014 only met:true if there is clear evidence in the transcript.' },
@@ -18000,7 +18010,7 @@ async function repl(args, ai, cfg, cwd) {
         err(green("  \u2713 think tier disabled\n"));
         return;
       }
-      const picked2 = await pickModel(dx.options.thinkModel === false ? "anthropic/claude-opus-4-6" : String(dx.options.thinkModel));
+      const picked2 = await pickModel(dx.options.thinkModel === false ? latestClaude("opus") : String(dx.options.thinkModel));
       if (picked2) applyThinkModel(picked2);
       return;
     }
@@ -18641,7 +18651,7 @@ ${task}`;
           return;
         }
         const current = dx.options.thinkModel === false ? void 0 : dx.options.thinkModel;
-        const picked = await pickModel(current ?? "anthropic/claude-opus-4-6");
+        const picked = await pickModel(current ?? latestClaude("opus"));
         if (picked) applyThinkModel(picked);
         else err(dim(`  think ${dx.options.thinkModel === false ? "off" : dx.options.thinkModel}
 `));
@@ -18997,7 +19007,7 @@ ${task}`;
           await fs2.writeFile(p, [
             "---",
             `description: What "${name}" is for \u2014 the Task tool picks agents by this line`,
-            "# model: anthropic/claude-haiku-4-5    # optional per-agent model override",
+            "# model: anthropic/haiku    # optional per-agent model override",
             "# tools: Read, Grep, Glob              # optional tool allowlist (omit = all tools)",
             "---",
             "",