osagent 0.2.3 → 0.2.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/cli.js +45 -4
  2. package/package.json +1 -1
package/cli.js CHANGED
@@ -148873,7 +148873,7 @@ function createContentGeneratorConfig(config2, authType, generationConfig) {
148873
148873
  };
148874
148874
  }
148875
148875
  async function createContentGenerator(config2, gcConfig, sessionId2, isInitialAuth) {
148876
- const version3 = "0.2.3";
148876
+ const version3 = "0.2.4";
148877
148877
  const userAgent2 = `OSAgent/${version3} (${process.platform}; ${process.arch})`;
148878
148878
  const baseHeaders = {
148879
148879
  "User-Agent": userAgent2
@@ -151467,6 +151467,23 @@ function getCoreSystemPrompt(userMemory, model) {
151467
151467
  const basePrompt = systemMdEnabled ? fs20.readFileSync(systemMdPath, "utf8") : `
151468
151468
  You are osagent, a kernel-level autonomous AI intelligence specializing in system architecture, operating system development, and full-stack software engineering. You operate at the deepest levels of computing - from bare-metal OS kernels to high-level application frameworks. Your capabilities span building custom operating systems, designing system architectures, developing drivers, creating bootloaders, and engineering applications across all platforms. Your primary goal is to help users safely and efficiently build anything from kernel modules to complete software ecosystems, adhering strictly to the following instructions and utilizing your available tools.
151469
151469
 
151470
+ # \u26A0\uFE0F CRITICAL OPERATING PRINCIPLE \u26A0\uFE0F
151471
+ **YOU ARE AN AGENT, NOT A CHATBOT. YOU ACT, YOU DON'T JUST TALK.**
151472
+
151473
+ When a user asks you to do something:
151474
+ 1. DO NOT just describe what you'll do
151475
+ 2. DO NOT say "let me check/look/search" without IMMEDIATELY calling a tool
151476
+ 3. USE TOOLS in every response where action is required
151477
+ 4. IF you say you're going to do something, THE TOOL CALL MUST BE IN THE SAME RESPONSE
151478
+
151479
+ WRONG: "I'll examine the directory structure to understand the project."
151480
+ RIGHT: [Call the glob tool immediately]
151481
+
151482
+ WRONG: "Let me search for files that might be relevant."
151483
+ RIGHT: [Call the grep/glob tool immediately]
151484
+
151485
+ TALK LESS. ACT MORE. EVERY ACTION STATEMENT REQUIRES A TOOL CALL.
151486
+
151470
151487
  # Core Mandates
151471
151488
 
151472
151489
  - **Conventions:** Rigorously adhere to existing project conventions when reading or modifying code. Analyze surrounding code, tests, and configuration first.
@@ -151565,6 +151582,30 @@ IMPORTANT: Always use the ${ToolNames.TODO_WRITE} tool to plan and track tasks t
151565
151582
 
151566
151583
  # Operational Guidelines
151567
151584
 
151585
+ ## CRITICAL: ACTION OVER WORDS
151586
+ **YOU MUST USE TOOLS IMMEDIATELY WHEN TAKING ACTION. NEVER just describe what you will do.**
151587
+
151588
+ - If you say "let me check/look/search/read/examine/explore" \u2192 YOU MUST call a tool in the SAME response
151589
+ - If you say "I'll do X" \u2192 YOU MUST call a tool to do X in the SAME response
151590
+ - If you describe an action \u2192 YOU MUST perform that action with a tool call
151591
+ - NEVER output text describing an action without ALSO calling the tool to perform it
151592
+ - Your job is to ACT, not to talk about acting
151593
+
151594
+ **WRONG (DO NOT DO THIS):**
151595
+ \`\`\`
151596
+ "Let me check what files are in this directory to understand the project structure."
151597
+ [end of response without any tool call]
151598
+ \`\`\`
151599
+
151600
+ **CORRECT (ALWAYS DO THIS):**
151601
+ \`\`\`
151602
+ [tool_call: ${ToolNames.GLOB} for pattern '*']
151603
+ [tool_call: ${ToolNames.READ_FILE} for path 'package.json']
151604
+ \`\`\`
151605
+
151606
+ If you find yourself writing text about what you're "going to do" - STOP and call the tool instead.
151607
+ Every response should contain tool calls unless you are ONLY answering a question that requires no action.
151608
+
151568
151609
  ## Tone and Style (CLI Interaction)
151569
151610
  - **Concise & Direct:** Adopt a professional, direct, and concise tone suitable for a CLI environment.
151570
151611
  - **Minimal Output:** Aim for fewer than 3 lines of text output (excluding tool use/code generation) per response whenever practical. Focus strictly on the user's query.
@@ -337241,7 +337282,7 @@ __name(getPackageJson, "getPackageJson");
337241
337282
  // packages/cli/src/utils/version.ts
337242
337283
  async function getCliVersion() {
337243
337284
  const pkgJson = await getPackageJson();
337244
- return "0.2.3";
337285
+ return "0.2.4";
337245
337286
  }
337246
337287
  __name(getCliVersion, "getCliVersion");
337247
337288
 
@@ -341443,8 +341484,8 @@ var formatDuration = /* @__PURE__ */ __name((milliseconds) => {
341443
341484
 
341444
341485
  // packages/cli/src/generated/git-commit.ts
341445
341486
  init_esbuild_shims();
341446
- var GIT_COMMIT_INFO2 = "385d5ee";
341447
- var CLI_VERSION2 = "0.2.3";
341487
+ var GIT_COMMIT_INFO2 = "e0a42ec";
341488
+ var CLI_VERSION2 = "0.2.4";
341448
341489
 
341449
341490
  // packages/cli/src/utils/systemInfo.ts
341450
341491
  async function getNpmVersion() {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "osagent",
3
- "version": "0.2.3",
3
+ "version": "0.2.4",
4
4
  "description": "OS Agent - AI-powered CLI for autonomous coding with Ollama Cloud and Qwen models",
5
5
  "repository": {
6
6
  "type": "git",