ucode-agent 1.58.0 → 1.59.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 om dixit
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -1,8 +1,8 @@
 # ucode
 
 A coding agent that lives in your terminal. It reads your code, edits it, runs
-your commands, and keeps every conversation on disk. It runs on NVIDIA,
-Cohere and Nex AGI models, all of them free.
+your commands, and keeps every conversation on disk. It runs on models served
+by NVIDIA — DeepSeek, Kimi, GLM and Nemotron — with a free key.
 
 It opens on a quiet screen — the name, the place to type, and the version in the
 corner:
@@ -66,14 +66,13 @@ space straight back when it finishes.
 npm i -g ucode-agent
 ```
 
-Then put a key where it will survive upgrades:
+Then save your NVIDIA key (free at [build.nvidia.com](https://build.nvidia.com)):
 
 ```bash
-mkdir -p ~/.ucode
-echo "UCODE_API_KEY=sk-or-..." > ~/.ucode/.env
+ucode login nvapi-...
 ```
 
-Keys are free at [openrouter.ai/keys](https://openrouter.ai/keys). A `.env` in
+It is written to `~/.ucode/.env` as `NVIDIA_API_KEY`. A `.env` in
 the project you are working on wins over that one, and a real environment
 variable wins over both.
 
@@ -87,35 +86,16 @@ Needs Node 22 or newer.
 
 ## The models
 
-Six, and no picker full of names nobody recognises. NVIDIA, Cohere and Nex AGI
-all serve capable models free, and all handle tool calling properly, which is
-the thing an agent actually depends on.
-
-| Model | Context | For |
-| --- | --- | --- |
-| Nemotron 3 Ultra | 1M | deepest reasoning, slowest to first token |
-| Nemotron 3.5 Lightning | 1M | the same enormous window, answers much sooner |
-| Nemotron 3 Super | 262k | strong all-rounder, quick to start |
-| Nemotron 3 Nano Omni | 256k | small, fast, reasoning tuned |
-| **North Mini Code** ★ | 256k | the default — built for code and interface work, quick to answer |
-| Nex N2.5 Pro | 262k | new agentic coder, on trial — can stall on big builds |
-
-`/model` shows them and switches. `ucode -m cohere/north-mini-code:free`
-starts on one.
-
-North Mini Code is the default: it is built for code and interfaces, which is most
-of what ucode is asked to do, and it answers far sooner than the big reasoning
-models. Switch to Ultra when a problem needs the million-token window more than
-the speed.
-
-**The model you chose is the model you keep.** Free endpoints are shared and
-"too many requests" is routine, so ucode waits it out with growing pauses and
-comes back to the same model. It does not quietly hand your build to a
-different one: a build that starts on one model and finishes on another
-finishes to a different standard, and the swap lands exactly when you are least
-placed to work out why the output changed. Set `UCODE_FALLBACK=1` if you would
-rather it moved down the list — North Mini Code, Nemotron 3.5 Lightning, Super,
-Ultra — when a model stays busy.
+Served by NVIDIA (build.nvidia.com), with a free key.
+
+| Model | For |
+| --- | --- |
+| **DeepSeek V4.1 Flash** ★ | the default — fast, and reliable with tools |
+| Kimi K3 | strong agentic coder — can be slow when NVIDIA is busy |
+| GLM 5.3 | strong coder — can be slow when NVIDIA is busy |
+| Nemotron 3 Super | NVIDIA all-rounder |
+
+`/model` shows them and switches. `ucode -m moonshotai/kimi-k3` starts on one.
 
 ## What it does
 
@@ -474,7 +454,7 @@ The tests need no network and no framework — `node test/run.js` runs them all.
 
 ## Licence
 
-ISC. Made with ❤️ by om dixit.
+MIT — see [LICENSE](LICENSE). Made with ❤️ by om dixit.
 
 The edit matchers, the summary template, the context-overflow and transient-error
 patterns and tool-name repair are adapted from

package/package.json

@@ -1,64 +1,64 @@
-{
-  "name": "ucode-agent",
-  "version": "1.58.0",
-  "description": "ucode - a terminal coding agent that reads, edits and runs your code, on NVIDIA, Cohere and Nex AGI models.",
-  "type": "module",
-  "main": "ucode.js",
-  "bin": {
-    "ucode": "ucode.js"
-  },
-  "files": [
-    "ucode.js",
-    "src/",
-    "skills/",
-    "templates/",
-    "THIRD_PARTY_NOTICES.md"
-  ],
-  "scripts": {
-    "start": "node ucode.js",
-    "test": "node test/run.js",
-    "prepublishOnly": "node scripts/no-bundled-key.js && node test/run.js",
-    "hooks": "node scripts/install-hooks.js"
-  },
-  "repository": {
-    "type": "git",
-    "url": "git+https://github.com/sppideey/ucode-agent.git"
-  },
-  "bugs": {
-    "url": "https://github.com/sppideey/ucode-agent/issues"
-  },
-  "homepage": "https://github.com/sppideey/ucode-agent#readme",
-  "engines": {
-    "node": ">=22"
-  },
-  "keywords": [
-    "agent",
-    "cli",
-    "terminal",
-    "coding-agent",
-    "llm",
-    "nvidia",
-    "nemotron",
-    "cohere",
-    "nex-agi"
-  ],
-  "author": "om dixit",
-  "license": "ISC",
-  "dependencies": {
-    "@babel/parser": "^7.29.8",
-    "chalk": "^6.0.0",
-    "dotenv": "^17.4.2",
-    "jsonrepair": "^3.15.0",
-    "marked": "^15.0.12",
-    "marked-terminal": "^7.3.0",
-    "openai": "^7.4.0",
-    "playwright-core": "^1.63.0"
-  },
-  "devDependencies": {
-    "@types/node": "^26.5.1",
-    "typescript": "^5.9.3"
-  },
-  "publishConfig": {
-    "access": "public"
-  }
-}
+{
+  "name": "ucode-agent",
+  "version": "1.59.0",
+  "description": "ucode - a terminal coding agent that reads, edits and runs your code, on models served by NVIDIA: DeepSeek, Kimi, GLM and Nemotron.",
+  "type": "module",
+  "main": "ucode.js",
+  "bin": {
+    "ucode": "ucode.js"
+  },
+  "files": [
+    "ucode.js",
+    "src/",
+    "skills/",
+    "templates/",
+    "THIRD_PARTY_NOTICES.md"
+  ],
+  "scripts": {
+    "start": "node ucode.js",
+    "test": "node test/run.js",
+    "prepublishOnly": "node scripts/no-bundled-key.js && node test/run.js",
+    "hooks": "node scripts/install-hooks.js"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/sppideey/ucode-agent.git"
+  },
+  "bugs": {
+    "url": "https://github.com/sppideey/ucode-agent/issues"
+  },
+  "homepage": "https://github.com/sppideey/ucode-agent#readme",
+  "engines": {
+    "node": ">=22"
+  },
+  "keywords": [
+    "agent",
+    "cli",
+    "terminal",
+    "coding-agent",
+    "llm",
+    "nvidia",
+    "nemotron",
+    "cohere",
+    "nex-agi"
+  ],
+  "author": "om dixit",
+  "license": "MIT",
+  "dependencies": {
+    "@babel/parser": "^7.29.8",
+    "chalk": "^6.0.0",
+    "dotenv": "^17.4.2",
+    "jsonrepair": "^3.15.0",
+    "marked": "^15.0.12",
+    "marked-terminal": "^7.3.0",
+    "openai": "^7.4.0",
+    "playwright-core": "^1.63.0"
+  },
+  "devDependencies": {
+    "@types/node": "^26.5.1",
+    "typescript": "^5.9.3"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}

package/src/core/doctor.js

@@ -12,7 +12,7 @@ import os from 'node:os';
 import path from 'node:path';
 import { theme, dim, blue } from '../ui/theme.js';
 import { VERSION } from './version.js';
-import { DEFAULT_MODEL, ENV_FILE } from './provider.js';
+import { DEFAULT_MODEL, ENV_FILE, BASE_URL, nvidiaKey } from './provider.js';
 import { newer } from './updater.js';
 
 const DEADLINE = 6000;
@@ -24,10 +24,10 @@ function version(cmd) {
 }
 
 async function checkKey() {
-  const key = process.env.UCODE_API_KEY || process.env.OPENROUTER_API_KEY;
-  if (!key) return { ok: false, name: 'API key', detail: 'not set', fix: `Add UCODE_API_KEY=... to ${ENV_FILE}` };
+  const key = nvidiaKey();
+  if (!key) return { ok: false, name: 'API key', detail: 'not set', fix: `Add NVIDIA_API_KEY=nvapi-... to ${ENV_FILE} (free at build.nvidia.com)` };
   try {
-    const r = await fetch('https://openrouter.ai/api/v1/chat/completions', {
+    const r = await fetch(`${BASE_URL}/chat/completions`, {
       method: 'POST',
       headers: { Authorization: `Bearer ${key}`, 'Content-Type': 'application/json' },
       body: JSON.stringify({ model: DEFAULT_MODEL, messages: [{ role: 'user', content: 'ok' }], max_tokens: 1 }),

package/src/core/login.js

@@ -14,11 +14,11 @@ import { promises as fs } from 'node:fs';
 import path from 'node:path';
 import { ENV_FILE } from './provider.js';
 
-const NAME = 'OPENROUTER_API_KEY';
+const NAME = 'NVIDIA_API_KEY';
 
-/** A plausible OpenRouter key, so a typo is caught here and not mid-answer. */
+/** A plausible NVIDIA key, so a typo is caught here and not mid-answer. */
 export function looksLikeKey(key) {
-  return typeof key === 'string' && /^sk-[A-Za-z0-9_-]{20,}$/.test(key.trim());
+  return typeof key === 'string' && /^nvapi-[A-Za-z0-9_-]{20,}$/.test(key.trim());
 }
 
 /** Put `key` in the machine-wide env file, keeping whatever else is in it. */
@@ -44,10 +44,10 @@ export async function saveKey(key) {
   const trimmed = String(key ?? '').trim();
 
   if (!trimmed) {
-    return `  Usage: ucode login <key>\n\n  Get one free at https://openrouter.ai/keys\n  It is saved to ${ENV_FILE} and used by every project on this machine.`;
+    return `  Usage: ucode login <key>\n\n  Get one free at https://build.nvidia.com\n  It is saved to ${ENV_FILE} and used by every project on this machine.`;
   }
   if (!looksLikeKey(trimmed)) {
-    return `  That does not look like an OpenRouter key — they start with "sk-".\n  Get one at https://openrouter.ai/keys`;
+    return `  That does not look like an NVIDIA key — they start with "nvapi-".\n  Get one at https://build.nvidia.com`;
   }
 
   const existing = await fs.readFile(ENV_FILE, 'utf8').catch(() => '');

package/src/core/provider.js

@@ -32,8 +32,8 @@ const PACKAGE_ROOT = join(HERE, '..', '..');
 export const UCODE_HOME = join(homedir(), '.ucode');
 export const ENV_FILE = join(UCODE_HOME, '.env');
 
-export const BASE_URL = 'https://openrouter.ai/api/v1';
-export const PROVIDER = 'OpenRouter';
+export const BASE_URL = 'https://integrate.api.nvidia.com/v1';
+export const PROVIDER = 'NVIDIA';
 
 // First definition wins — dotenv never overwrites a variable that already
 // exists — so the order here is the precedence order:
@@ -44,60 +44,36 @@ dotenv.config({ path: ENV_FILE, quiet: true });
 dotenv.config({ path: join(PACKAGE_ROOT, '.env'), quiet: true });
 
 /**
- * The whole model list. Not a starting point — the list.
- *
- * ucode runs on NVIDIA, Cohere and Nex AGI only. All three serve genuinely
- * capable models free through OpenRouter, all handle tool calling properly,
- * and keeping the set to six means every one of them has
- * been used in anger rather than listed on the strength of a benchmark. A
- * picker offering sixty models is a picker nobody reads.
+ * The whole model list, served by NVIDIA (build.nvidia.com).
  *
  * `name` is what the status bar shows. `note` is what the picker shows.
  */
 export const MODELS = {
-  'nvidia/nemotron-3-ultra-550b-a55b:free': {
-    name: 'Nemotron 3 Ultra',
-    context: 1_000_000,
+  'deepseek-ai/deepseek-v4.1-flash': {
+    name: 'DeepSeek V4.1 Flash',
+    context: 128_000,
     star: true,
-    note: 'deepest reasoning, 1M context — slowest to answer',
-  },
-  'nvidia/nemotron-3.5-lightning:free': {
-    name: 'Nemotron 3.5 Lightning',
-    context: 1_000_000,
-    note: 'same huge window, answers much sooner',
+    note: 'the default — fast, and reliable with tools',
   },
-  'nvidia/nemotron-3-super-120b-a12b:free': {
-    name: 'Nemotron 3 Super',
-    context: 262_144,
-    note: 'strong all-rounder, quick to first token',
+  'moonshotai/kimi-k3': {
+    name: 'Kimi K3',
+    context: 128_000,
+    note: 'strong agentic coder — can be slow when NVIDIA is busy',
   },
-  'nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free': {
-    name: 'Nemotron 3 Nano Omni',
-    context: 256_000,
-    note: 'small and fast, reasoning tuned',
-  },
-  'cohere/north-mini-code:free': {
-    name: 'North Mini Code',
-    context: 256_000,
-    star: true,
-    note: 'the default — built for code and interface work, quick to answer',
+  'z-ai/glm-5.3': {
+    name: 'GLM 5.3',
+    context: 128_000,
+    note: 'strong coder — can be slow when NVIDIA is busy',
   },
-  // On trial, so not the default and not in FALLBACKS yet. Nex AGI is its only
-  // upstream, so a stall there has nowhere else to go.
-  'nex-agi/nex-n2.5-pro:free': {
-    name: 'Nex N2.5 Pro',
-    context: 262_144,
-    note: 'new agentic coder, on trial — can stall on big builds',
+  'nvidia/nemotron-3-super-120b-a12b': {
+    name: 'Nemotron 3 Super',
+    context: 128_000,
+    note: 'NVIDIA all-rounder',
   },
 };
 
-/**
- * North Mini Code is the default: it is built for code and interface work,
- * which is what ucode is mostly asked to do, and it answers far sooner than
- * the big reasoning models. /model moves to Ultra when a problem needs the
- * million-token window and the long think more than it needs the speed.
- */
-export const DEFAULT_MODEL = 'cohere/north-mini-code:free';
+/** The model a session starts on. */
+export const DEFAULT_MODEL = 'deepseek-ai/deepseek-v4.1-flash';
 
 /**
  * Where to go when a model is busy, in order of preference. Each is served by
@@ -106,10 +82,10 @@ export const DEFAULT_MODEL = 'cohere/north-mini-code:free';
  * the first "too many requests".
  */
 export const FALLBACKS = [
-  'cohere/north-mini-code:free',
-  'nvidia/nemotron-3.5-lightning:free',
-  'nvidia/nemotron-3-super-120b-a12b:free',
-  'nvidia/nemotron-3-ultra-550b-a55b:free',
+  'deepseek-ai/deepseek-v4.1-flash',
+  'nvidia/nemotron-3-super-120b-a12b',
+  'moonshotai/kimi-k3',
+  'z-ai/glm-5.3',
 ];
 
 /** The next model to try after `id`, skipping any already tried this round. */
@@ -173,7 +149,7 @@ export function setModel(id) {
     throw new Failure({
       kind: 'bad_model',
       attempted: `switching to "${wanted}"`,
-      failed: 'ucode only runs NVIDIA, Cohere and Nex AGI models, and that is not one of them.',
+      failed: 'That model is not in ucode\'s list.',
       fix: `Run /model to choose from: ${Object.keys(MODELS).join(', ')}`,
     });
   }
@@ -238,19 +214,27 @@ export function estimateConversation(messages) {
  * package is readable by anyone who runs `npm pack ucode-agent`, and no amount
  * of first-run convenience is worth handing out a live credential.
  */
+/** The NVIDIA key: NVIDIA_API_KEY, or a UCODE_API_KEY that is one (nvapi-...). */
+export function nvidiaKey() {
+  const direct = (process.env.NVIDIA_API_KEY || '').trim();
+  if (direct) return direct;
+  const shared = (process.env.UCODE_API_KEY || '').trim();
+  return shared.startsWith('nvapi-') ? shared : '';
+}
+
 function apiKey() {
   // UCODE_API_KEY is the documented name. The provider's own variable name is
   // still read, so a key set up for another tool keeps working here.
-  const key = (process.env.UCODE_API_KEY || process.env.OPENROUTER_API_KEY || '').trim();
+  const key = nvidiaKey();
   if (!key) {
     throw new Failure({
       kind: 'no_api_key',
       attempted: 'connecting to the model',
-      failed: 'No API key is set - UCODE_API_KEY is missing from the environment and from every .env file.',
+      failed: 'No NVIDIA API key is set - NVIDIA_API_KEY is missing from the environment and from every .env file.',
       fix:
-        `Put UCODE_API_KEY=your-key in ${ENV_FILE} — that applies to every ` +
+        `Put NVIDIA_API_KEY=nvapi-... in ${ENV_FILE} — that applies to every ` +
         'project on this machine — or in a .env file beside your code. ' +
-        'Free keys: https://openrouter.ai/keys',
+        'Free keys: https://build.nvidia.com',
     });
   }
   return key;
@@ -672,7 +656,6 @@ export async function ask(messages, tools = [], opts = {}) {
     // nothing, and the provider eventually drops the request as idle. Asking
     // for it fixes the blank screen and the dropped request together. Models
     // that do not reason ignore the flag.
-    include_reasoning: true,
   };
 
   const wired = wireTools(tools);

package/src/tools/browser.js

@@ -20,7 +20,7 @@ import { ToolFailure } from '../core/failure.js';
 import { ask } from '../core/provider.js';
 import { getRoot, result } from './shared.js';
 
-const VISION_MODEL = 'nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free';
+const VISION_MODEL = 'nvidia/nemotron-3-nano-omni-30b-a3b-reasoning';
 const WIDTHS = [
   { name: 'phone', width: 375, height: 812 },
   { name: 'desktop', width: 1440, height: 900 },

package/ucode.js

@@ -1,4 +1,4 @@
-#!/usr/bin/env node
+#!/usr/bin/env node
 /**
  * ucode.js — the command.
  *
@@ -64,8 +64,8 @@ async function usage() {
     '    doctor             check that everything ucode needs is working\n' +
     '    login <key>        save your key for every folder on this machine\n\n' +
     `  ${sky('Models')}\n${models}\n\n` +
-    `  Needs UCODE_API_KEY in the environment or in ${ENV_FILE}\n` +
-    '  Free keys: https://openrouter.ai/keys\n\n'
+    `  Needs NVIDIA_API_KEY in the environment or in ${ENV_FILE}\n` +
+    '  Free keys: https://build.nvidia.com\n\n'
   );
 }