Now I'll compile a comprehensive analysis based on my thorough exploration of the opencode project.

# OpenCode Project Architecture Analysis

## 1. Main Architecture and Entry Points

### Core Entry Point
- **File**: `/home/jm/data/code/opencode/packages/opencode/src/index.ts`
- **Type**: CLI application using `yargs` for command parsing
- **Key Features**: 
  - Command-based architecture with 20+ commands (run, generate, debug, auth, etc.)
  - Global error handling for unhandled rejections and exceptions
  - Logging middleware initialization
  - Plugin-based command registration system

### Core Architectural Layers
1. **CLI Layer** (`/cli/cmd/`): Command definitions and user interface
2. **Session Layer** (`/session/`): Conversation management and message processing
3. **Agent Layer** (`/agent/`): Agent orchestration and configuration
4. **Tool Layer** (`/tool/`): Tool definitions and execution
5. **Provider Layer** (`/provider/`): LLM provider integrations
6. **Project Layer** (`/project/`): Project context and instance management

### Key Architectural Patterns
- **Instance State Pattern**: Uses `Instance.state()` for singleton-like state management across components
- **Lazy Loading**: Extensive use of lazy initialization for performance (e.g., ripgrep download, tree-sitter parser)
- **Plugin Architecture**: MCP (Model Context Protocol) servers and custom tools can be dynamically loaded
- **Permission System**: Granular permission controls per agent (edit, bash, skill, webfetch, etc.)

## 2. Current Models/Tools Integration

### Provider Model Integration
**File**: `/home/jm/data/code/opencode/packages/opencode/src/provider/provider.ts`
- **Supported Providers**: 20+ bundled providers (OpenAI, Anthropic, Google, Azure, etc.)
- **SDK**: Built on AI SDK with provider-specific loaders and transformers
- **Authentication**: Supports API keys, OAuth, and custom auth flows
- **Features**: Cost tracking, model selection, rate limiting, response streaming
- **Custom ...

{
  "message": "Not Found",
  "documentation_url": "https://docs.github.com/rest/repos/repos#get-a-repository",
  "status": "404"
}

404: Not Found

/home/jm/data/code/opencode/packages/opencode/src/tool/write.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/websearch.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/webfetch.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/tool.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/todo.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/task.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/skill.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/registry.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/read.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/patch.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/multiedit.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/lsp.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/ls.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/invalid.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/grep.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/glob.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/edit.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/codesearch.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/batch.ts
/home/jm/data/code/opencode/packages/opencode/src/tool/bash.ts

code-yeongyu/oh-my-opencode	The Best Agent Harness. Meet Sisyphus: The Batteries-Included Agent that codes like you.	public	2026-01-18T03:02:55Z
alvinunreal/oh-my-opencode-slim	Slimmed and cleaned oh-my-opencode, consumes much less tokens; Help us grow, consider giving a ⭐;	public	2026-01-18T03:00:45Z
Yeachan-Heo/oh-my-claude-sisyphus	Sisyphus from OmO (Oh My Opencode), ported to the Claude Code SDK. Written with Claude Code — ironically. Anthropic, what are you gonna do next?	public	2026-01-18T02:57:23Z
safzanpirani/opencode-configs	My OpenCode and Oh-My-OpenCode configuration files with API proxy setup documentation	public	2026-01-17T22:09:46Z
icysaintdx/OpenCode-Config-Manager	OCCM是一个Opencode&Oh my opencode配置管理器	public	2026-01-18T00:13:07Z

/home/jm/data/code/opencode/.git/refs/remotes/upstream/ripgrep-tests
/home/jm/data/code/opencode/.git/logs/refs/remotes/upstream/ripgrep-tests
/home/jm/data/code/opencode/.git/logs/refs/heads/astro-hardening
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-CbQHKStN.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-OEl6i-aI.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-dark-high-contrast-E3gJ1_iC.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-light-high-contrast-BfjtVDDH.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/plastic-3e1v2bzS.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-CbQHKStN.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-OEl6i-aI.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-light-high-contrast-BfjtVDDH.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/plastic-3e1v2bzS.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-dark-high-contrast-E3gJ1_iC.js.map
/home/jm/data/code/opencode/packages/opencode/src/cli/cmd/debug/ripgrep.ts
/home/jm/data/code/opencode/packages/opencode/src/cli/cmd/tui/ui/toast.tsx
/home/jm/data/code/opencode/packages/opencode/src/file/ripgrep.ts
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/tool/file-search.ts
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/tool/web-search-preview.ts
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/tool/web-search.ts
/home/jm/data/code/opencode/packages/opencode/src/session/prompt/beast.txt

<file>
00001| import z from "zod"
00002| import { Tool } from "./tool"
00003| import { Ripgrep } from "../file/ripgrep"
00004| 
00005| import DESCRIPTION from "./grep.txt"
00006| import { Instance } from "../project/instance"
00007| 
00008| const MAX_LINE_LENGTH = 2000
00009| 
00010| export const GrepTool = Tool.define("grep", {
00011|   description: DESCRIPTION,
00012|   parameters: z.object({
00013|     pattern: z.string().describe("The regex pattern to search for in file contents"),
00014|     path: z.string().optional().describe("The directory to search in. Defaults to the current working directory."),
00015|     include: z.string().optional().describe('File pattern to include in the search (e.g. "*.js", "*.{ts,tsx}")'),
00016|   }),
00017|   async execute(params) {
00018|     if (!params.pattern) {
00019|       throw new Error("pattern is required")
00020|     }
00021| 
00022|     const searchPath = params.path || Instance.directory
00023| 
00024|     const rgPath = await Ripgrep.filepath()
00025|     const args = ["-nH", "--field-match-separator=|", "--regexp", params.pattern]
00026|     if (params.include) {
00027|       args.push("--glob", params.include)
00028|     }
00029|     args.push(searchPath)
00030| 
00031|     const proc = Bun.spawn([rgPath, ...args], {
00032|       stdout: "pipe",
00033|       stderr: "pipe",
00034|     })
00035| 
00036|     const output = await new Response(proc.stdout).text()
00037|     const errorOutput = await new Response(proc.stderr).text()
00038|     const exitCode = await proc.exited
00039| 
00040|     if (exitCode === 1) {
00041|       return {
00042|         title: params.pattern,
00043|         metadata: { matches: 0, truncated: false },
00044|         output: "No files found",
00045|       }
00046|     }
00047| 
00048|     if (exitCode !== 0) {
00049|       throw new Error(`ripgrep failed: ${errorOutput}`)
00050|     }
00051| 
00052|     // Handle both Unix (\n) and Windows (\r\n) line endings
00053|     const lines = o...

<file>
00001| import z from "zod"
00002| import { Tool } from "./tool"
00003| import DESCRIPTION from "./codesearch.txt"
00004| import { Config } from "../config/config"
00005| import { Permission } from "../permission"
00006| 
00007| const API_CONFIG = {
00008|   BASE_URL: "https://mcp.exa.ai",
00009|   ENDPOINTS: {
00010|     CONTEXT: "/mcp",
00011|   },
00012| } as const
00013| 
00014| interface McpCodeRequest {
00015|   jsonrpc: string
00016|   id: number
00017|   method: string
00018|   params: {
00019|     name: string
00020|     arguments: {
00021|       query: string
00022|       tokensNum: number
00023|     }
00024|   }
00025| }
00026| 
00027| interface McpCodeResponse {
00028|   jsonrpc: string
00029|   result: {
00030|     content: Array<{
00031|       type: string
00032|       text: string
00033|     }>
00034|   }
00035| }
00036| 
00037| export const CodeSearchTool = Tool.define("codesearch", {
00038|   description: DESCRIPTION,
00039|   parameters: z.object({
00040|     query: z
00041|       .string()
00042|       .describe(
00043|         "Search query to find relevant context for APIs, Libraries, and SDKs. For example, 'React useState hook examples', 'Python pandas dataframe filtering', 'Express.js middleware', 'Next js partial prerendering configuration'",
00044|       ),
00045|     tokensNum: z
00046|       .number()
00047|       .min(1000)
00048|       .max(50000)
00049|       .default(5000)
00050|       .describe(
00051|         "Number of tokens to return (1000-50000). Default is 5000 tokens. Adjust this value based on how much context you need - use lower values for focused queries and higher values for comprehensive documentation.",
00052|       ),
00053|   }),
00054|   async execute(params, ctx) {
00055|     const cfg = await Config.get()
00056|     if (cfg.permission?.webfetch === "ask")
00057|       await Permission.ask({
00058|         type: "codesearch",
00059|         sessionID: ctx.sessionID,
00060|         messageID: ctx.messageID,
0006...

{"description":"The Best Agent Harness. Meet Sisyphus: The Batteries-Included Agent that codes like you.","homepageUrl":"","url":"https://github.com/code-yeongyu/oh-my-opencode"}

[{"name":".github","path":".github","sha":"5c34dabd3eaf5f003a8ba652cb270cd09f87595a","size":0,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/.github?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/tree/dev/.github","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/trees/5c34dabd3eaf5f003a8ba652cb270cd09f87595a","download_url":null,"type":"dir","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/.github?ref=dev","git":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/trees/5c34dabd3eaf5f003a8ba652cb270cd09f87595a","html":"https://github.com/code-yeongyu/oh-my-opencode/tree/dev/.github"}},{"name":".gitignore","path":".gitignore","sha":"e913cc4be8fdb5d0dc7de7c3c247a715292ddc35","size":403,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/.gitignore?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/.gitignore","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/e913cc4be8fdb5d0dc7de7c3c247a715292ddc35","download_url":"https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/dev/.gitignore","type":"file","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/.gitignore?ref=dev","git":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/e913cc4be8fdb5d0dc7de7c3c247a715292ddc35","html":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/.gitignore"}},{"name":".opencode","path":".opencode","sha":"4b8b1e15bcfd483476eec75c0da2916df72e0f49","size":0,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/.opencode?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/tree/dev/.opencode","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/trees/4b8b1e15bcfd483476eec75c0da2916df72e0f49","download_url":null,"type":"dir","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencod...

> [!NOTE]
>
> [![Sisyphus Labs — Sisyphus is the agent that codes like your team.](./.github/assets/sisyphuslabs.png?v=2)](https://sisyphuslabs.ai)
> > **We're building a fully productized version of Sisyphus to define the future of frontier agents. <br />Join the waitlist [here](https://sisyphuslabs.ai).**

> [!TIP]
>
> [![The Orchestrator is now available in beta.](./.github/assets/orchestrator-sisyphus.png?v=3)](https://github.com/code-yeongyu/oh-my-opencode/releases/tag/v3.0.0-beta.7)
> > **The Orchestrator is now available in beta. Use `oh-my-opencode@3.0.0-beta.7` to install it.**
>
> Be with us!
>
> | [<img alt="Discord link" src="https://img.shields.io/discord/1452487457085063218?color=5865F2&label=discord&labelColor=black&logo=discord&logoColor=white&style=flat-square" width="156px" />](https://discord.gg/PUwSMR9XNk) | Join our [Discord community](https://discord.gg/PUwSMR9XNk) to connect with contributors and fellow `oh-my-opencode` users. |
> | :-----| :----- |
> | [<img alt="X link" src="https://img.shields.io/badge/Follow-%40justsisyphus-00CED1?style=flat-square&logo=x&labelColor=black" width="156px" />](https://x.com/justsisyphus) | News and updates for `oh-my-opencode` used to be posted on my X account. <br /> Since it was suspended mistakenly, [@justsisyphus](https://x.com/justsisyphus) now posts updates on my behalf. |
> | [<img alt="GitHub Follow" src="https://img.shields.io/github/followers/code-yeongyu?style=flat-square&logo=github&labelColor=black&color=24292f" width="156px" />](https://github.com/code-yeongyu) | Follow [@code-yeongyu](https://github.com/code-yeongyu) on GitHub for more projects. |

<!-- <CENTERED SECTION FOR GITHUB DISPLAY> -->

<div align="center">

[![Oh My OpenCode](./.github/assets/hero.jpg)](https://github.com/code-yeongyu/oh-my-opencode#oh-my-opencode)

[![Preview](./.github/assets/omo.png)](https://github.com/code-yeongyu/oh-my-opencode#oh-my-opencode)


</div>

> This is coding on steroids—`oh-my-opencode` in action. Ru...

404: Not Found

# PROJECT KNOWLEDGE BASE

**Generated:** 2026-01-17T21:55:00+09:00
**Commit:** 255f535a
**Branch:** dev

## OVERVIEW

OpenCode plugin implementing multi-model agent orchestration (Claude Opus 4.5, GPT-5.2, Gemini 3, Grok, GLM-4.7). 31 lifecycle hooks, 20+ tools (LSP, AST-Grep, delegation), 10 specialized agents, Claude Code compatibility layer. "oh-my-zsh" for OpenCode.

## STRUCTURE

```
oh-my-opencode/
├── src/
│   ├── agents/        # 10 AI agents (Sisyphus, oracle, librarian, explore, frontend, etc.) - see src/agents/AGENTS.md
│   ├── hooks/         # 31 lifecycle hooks (PreToolUse, PostToolUse, Stop, etc.) - see src/hooks/AGENTS.md
│   ├── tools/         # 20+ tools (LSP, AST-Grep, delegation, session) - see src/tools/AGENTS.md
│   ├── features/      # Background agents, Claude Code compat layer - see src/features/AGENTS.md
│   ├── shared/        # 43 cross-cutting utilities - see src/shared/AGENTS.md
│   ├── cli/           # CLI installer, doctor, run - see src/cli/AGENTS.md
│   ├── mcp/           # Built-in MCPs: websearch, context7, grep_app
│   ├── config/        # Zod schema, TypeScript types
│   └── index.ts       # Main plugin entry (568 lines)
├── script/            # build-schema.ts, publish.ts, build-binaries.ts
├── packages/          # 7 platform-specific binaries
└── dist/              # Build output (ESM + .d.ts)
```

## WHERE TO LOOK

| Task | Location | Notes |
|------|----------|-------|
| Add agent | `src/agents/` | Create .ts with factory, add to `builtinAgents` in index.ts |
| Add hook | `src/hooks/` | Create dir with `createXXXHook()`, register in index.ts |
| Add tool | `src/tools/` | Dir with index/types/constants/tools.ts, add to `builtinTools` |
| Add MCP | `src/mcp/` | Create config, add to index.ts |
| Add skill | `src/features/builtin-skills/` | Create dir with SKILL.md |
| LSP behavior | `src/tools/lsp/` | client.ts (connection), tools.ts (handlers) |
| AST-Grep | `src/tools/ast-grep/` | napi.ts for @ast-grep/napi binding |
| Config ...

[{"name":"AGENTS.md","path":"src/tools/AGENTS.md","sha":"03c602e469272c626d5e76eff34fc9b79f4b5365","size":3011,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/AGENTS.md?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/src/tools/AGENTS.md","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/03c602e469272c626d5e76eff34fc9b79f4b5365","download_url":"https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/dev/src/tools/AGENTS.md","type":"file","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/AGENTS.md?ref=dev","git":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/03c602e469272c626d5e76eff34fc9b79f4b5365","html":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/src/tools/AGENTS.md"}},{"name":"ast-grep","path":"src/tools/ast-grep","sha":"e794a5b569d3adec6c3018b756124b83294dab46","size":0,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/ast-grep?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/tree/dev/src/tools/ast-grep","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/trees/e794a5b569d3adec6c3018b756124b83294dab46","download_url":null,"type":"dir","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/ast-grep?ref=dev","git":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/trees/e794a5b569d3adec6c3018b756124b83294dab46","html":"https://github.com/code-yeongyu/oh-my-opencode/tree/dev/src/tools/ast-grep"}},{"name":"background-task","path":"src/tools/background-task","sha":"c89fad5956f8d1656a5511a9581b07b47c71e450","size":0,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/background-task?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/tree/dev/src/tools/background-task","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-ope...

<!DOCTYPE html><html lang="en"><head><meta charSet="utf-8"/><meta name="viewport" content="width=device-width"/><meta name="next-head-count" content="2"/><link rel="preload" href="/_next/static/css/149dca090180f275.css" as="style"/><link rel="stylesheet" href="/_next/static/css/149dca090180f275.css" data-n-g=""/><noscript data-n-css=""></noscript><script defer="" nomodule="" src="/_next/static/chunks/polyfills-42372ed130431b0a.js"></script><script src="/necaptcha.js" defer="" data-nscript="beforeInteractive"></script><script src="/_next/static/chunks/webpack-3c480d76c8a290e6.js" defer=""></script><script src="/_next/static/chunks/framework-cf8bb177a3874f04.js" defer=""></script><script src="/_next/static/chunks/main-d07b685a1deccd51.js" defer=""></script><script src="/_next/static/chunks/pages/_app-96be9648e5ff68ec.js" defer=""></script><script src="/_next/static/chunks/920-ff658830212a258b.js" defer=""></script><script src="/_next/static/chunks/pages/docs/api/chat.en-US-f167de929c445293.js" defer=""></script><script src="/_next/static/2wpwYvykUwWY2N5s9m-ph/_buildManifest.js" defer=""></script><script src="/_next/static/2wpwYvykUwWY2N5s9m-ph/_ssgManifest.js" defer=""></script><style>.data-ant-cssinjs-cache-path{content:"";}</style></head><body dir="ltr"><div id="__next"><script>!function(){try{var d=document.documentElement,c=d.classList;c.remove('light','dark');var e=localStorage.getItem('moonshot-theme');if('system'===e||(!e&&false)){var t='(prefers-color-scheme: dark)',m=window.matchMedia(t);if(m.media!==t||m.matches){d.style.colorScheme = 'dark';c.add('dark')}else{d.style.colorScheme = 'light';c.add('light')}}else if(e){c.add(e|| '')}else{c.add('light')}if(e==='light'||e==='dark'||!e)d.style.colorScheme=e||'light'}catch(e){}}()</script></div><script id="__NEXT_DATA__" type="application/json">{"props":{"pageProps":{}},"page":"/docs/api/chat.en-US","query":{},"buildId":"2wpwYvykUwWY2N5s9m-ph","nextExport":true,"autoExport":true,"isFallback":false,"locale":"en-US",...

<file>
00001| // Ripgrep utility functions
00002| import path from "path"
00003| import { Global } from "../global"
00004| import fs from "fs/promises"
00005| import z from "zod"
00006| import { NamedError } from "@opencode-ai/util/error"
00007| import { lazy } from "../util/lazy"
00008| import { $ } from "bun"
00009| 
00010| import { ZipReader, BlobReader, BlobWriter } from "@zip.js/zip.js"
00011| import { Log } from "@/util/log"
00012| 
00013| export namespace Ripgrep {
00014|   const log = Log.create({ service: "ripgrep" })
00015|   const Stats = z.object({
00016|     elapsed: z.object({
00017|       secs: z.number(),
00018|       nanos: z.number(),
00019|       human: z.string(),
00020|     }),
00021|     searches: z.number(),
00022|     searches_with_match: z.number(),
00023|     bytes_searched: z.number(),
00024|     bytes_printed: z.number(),
00025|     matched_lines: z.number(),
00026|     matches: z.number(),
00027|   })
00028| 
00029|   const Begin = z.object({
00030|     type: z.literal("begin"),
00031|     data: z.object({
00032|       path: z.object({
00033|         text: z.string(),
00034|       }),
00035|     }),
00036|   })
00037| 
00038|   export const Match = z.object({
00039|     type: z.literal("match"),
00040|     data: z.object({
00041|       path: z.object({
00042|         text: z.string(),
00043|       }),
00044|       lines: z.object({
00045|         text: z.string(),
00046|       }),
00047|       line_number: z.number(),
00048|       absolute_offset: z.number(),
00049|       submatches: z.array(
00050|         z.object({
00051|           match: z.object({
00052|             text: z.string(),
00053|           }),
00054|           start: z.number(),
00055|           end: z.number(),
00056|         }),
00057|       ),
00058|     }),
00059|   })
00060| 
00061|   const End = z.object({
00062|     type: z.literal("end"),
00063|     data: z.object({
00064|       path: z.object({
00065|         text: z.string(),
00066|       }),
00067|       binary...

[{"name":"cli.ts","path":"src/tools/ast-grep/cli.ts","sha":"b93c317f5202bc225b46616c68f9035374ce9cce","size":6152,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/ast-grep/cli.ts?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/src/tools/ast-grep/cli.ts","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/b93c317f5202bc225b46616c68f9035374ce9cce","download_url":"https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/dev/src/tools/ast-grep/cli.ts","type":"file","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/ast-grep/cli.ts?ref=dev","git":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/b93c317f5202bc225b46616c68f9035374ce9cce","html":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/src/tools/ast-grep/cli.ts"}},{"name":"constants.ts","path":"src/tools/ast-grep/constants.ts","sha":"baa43539b4c92c2e7b02c930926dc6eb09a9ff3d","size":6559,"url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/ast-grep/constants.ts?ref=dev","html_url":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/src/tools/ast-grep/constants.ts","git_url":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/baa43539b4c92c2e7b02c930926dc6eb09a9ff3d","download_url":"https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/dev/src/tools/ast-grep/constants.ts","type":"file","_links":{"self":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/contents/src/tools/ast-grep/constants.ts?ref=dev","git":"https://api.github.com/repos/code-yeongyu/oh-my-opencode/git/blobs/baa43539b4c92c2e7b02c930926dc6eb09a9ff3d","html":"https://github.com/code-yeongyu/oh-my-opencode/blob/dev/src/tools/ast-grep/constants.ts"}},{"name":"downloader.ts","path":"src/tools/ast-grep/downloader.ts","sha":"6ed228847ae3a0f6f89fca0fdb6d2e217f446524","size":3743,"url":"https://api.github.com/repos/code-yeongyu/oh-my-o...

import { tool, type ToolDefinition } from "@opencode-ai/plugin/tool"
import { CLI_LANGUAGES } from "./constants"
import { runSg } from "./cli"
import { formatSearchResult, formatReplaceResult } from "./utils"
import type { CliLanguage } from "./types"

function showOutputToUser(context: unknown, output: string): void {
  const ctx = context as { metadata?: (input: { metadata: { output: string } }) => void }
  ctx.metadata?.({ metadata: { output } })
}

function getEmptyResultHint(pattern: string, lang: CliLanguage): string | null {
  const src = pattern.trim()

  if (lang === "python") {
    if (src.startsWith("class ") && src.endsWith(":")) {
      const withoutColon = src.slice(0, -1)
      return `💡 Hint: Remove trailing colon. Try: "${withoutColon}"`
    }
    if ((src.startsWith("def ") || src.startsWith("async def ")) && src.endsWith(":")) {
      const withoutColon = src.slice(0, -1)
      return `💡 Hint: Remove trailing colon. Try: "${withoutColon}"`
    }
  }

  if (["javascript", "typescript", "tsx"].includes(lang)) {
    if (/^(export\s+)?(async\s+)?function\s+\$[A-Z_]+\s*$/i.test(src)) {
      return `💡 Hint: Function patterns need params and body. Try "function $NAME($$$) { $$$ }"`
    }
  }

  return null
}

export const ast_grep_search: ToolDefinition = tool({
  description:
    "Search code patterns across filesystem using AST-aware matching. Supports 25 languages. " +
    "Use meta-variables: $VAR (single node), $$$ (multiple nodes). " +
    "IMPORTANT: Patterns must be complete AST nodes (valid code). " +
    "For functions, include params and body: 'export async function $NAME($$$) { $$$ }' not 'export async function $NAME'. " +
    "Examples: 'console.log($MSG)', 'def $FUNC($$$):', 'async function $NAME($$$)'",
  args: {
    pattern: tool.schema.string().describe("AST pattern with meta-variables ($VAR, $$$). Must be complete AST node."),
    lang: tool.schema.enum(CLI_LANGUAGES).describe("Target language"),
    paths: tool.schema.array(tool.sc...

<file>
00001| import { Log } from "../util/log"
00002| import path from "path"
00003| import { pathToFileURL } from "url"
00004| import os from "os"
00005| import z from "zod"
00006| import { Filesystem } from "../util/filesystem"
00007| import { ModelsDev } from "../provider/models"
00008| import { mergeDeep, pipe, unique } from "remeda"
00009| import { Global } from "../global"
00010| import fs from "fs/promises"
00011| import { lazy } from "../util/lazy"
00012| import { NamedError } from "@opencode-ai/util/error"
00013| import { Flag } from "../flag/flag"
00014| import { Auth } from "../auth"
00015| import { type ParseError as JsoncParseError, parse as parseJsonc, printParseErrorCode } from "jsonc-parser"
00016| import { Instance } from "../project/instance"
00017| import { LSPServer } from "../lsp/server"
00018| import { BunProc } from "@/bun"
00019| import { Installation } from "@/installation"
00020| import { ConfigMarkdown } from "./markdown"
00021| 
00022| export namespace Config {
00023|   const log = Log.create({ service: "config" })
00024| 
00025|   // Custom merge function that concatenates plugin arrays instead of replacing them
00026|   function mergeConfigWithPlugins(target: Info, source: Info): Info {
00027|     const merged = mergeDeep(target, source)
00028|     // If both configs have plugin arrays, concatenate them instead of replacing
00029|     if (target.plugin && source.plugin) {
00030|       const pluginSet = new Set([...target.plugin, ...source.plugin])
00031|       merged.plugin = Array.from(pluginSet)
00032|     }
00033|     return merged
00034|   }
00035| 
00036|   export const state = Instance.state(async () => {
00037|     const auth = await Auth.all()
00038|     let result = await global()
00039| 
00040|     // Override with custom config if provided
00041|     if (Flag.OPENCODE_CONFIG) {
00042|       result = mergeConfigWithPlugins(result, await loadFile(Flag.OPENCODE_CONFIG))
00043|       log.debug("loaded custom config", { path: Fla...

Doriandarko/kimi-writer	AI writing agent powered by kimi-k2-thinking - autonomously creates novels and stories with deep reasoning	public	2026-01-17T22:42:27Z
HarleyCoops/KimiK2Manim	Create epic math and physics explainer animations with Kimi K2 Thinking.	public	2026-01-15T20:52:18Z
prnake/kimi-deepresearch	Kimi K2 Thinking Agentic Search Unofficial Implementation	public	2025-12-16T09:16:17Z
chongdashu/cc-kimi-k2-thinking-prompts	Prompts for Claude Code to use with Moonshot's Kimi K2 Thinking Model	public	2025-12-27T17:10:57Z
jeinlee1991/chinese-llm-benchmark	ReLE评测：中文AI大模型能力评测（持续更新）：目前已囊括335个大模型，覆盖chatgpt、gpt-5.2、o4-mini、谷歌gemini-3-pro、Claude-4.5、文心ERNIE-X1.1、ERNIE-5.0-Thinking、qwen3-max、百川、讯飞星火、商汤senseChat等商用模型， 以及kimi-k2、ernie4.5、minimax-M2、deepseek-v3.2、qwen3-2507、llama4、智谱GLM-4.6、gemma3、mistral等开源大模型。不仅提供排行榜，也提供规模超200万的大模型缺陷库！方便广大社区研究分析、改进大模型。	public	2026-01-18T02:49:16Z
syaikhipin/KIMI-AI-Humanizer	Humanize Text Using Kimi K2 Thinking	public	2025-12-01T09:42:24Z
h4sh5/WebWord	WebWord app generated by Kimi K2 Thinking in its launch demo https://moonshotai.github.io/Kimi-K2/thinking.html	public	2025-11-08T06:52:25Z
ayanreza/kimi-k2-thinking-runpod		public	2026-01-08T00:37:32Z
tinfoilsh/confidential-kimi-k2-thinking		public	2026-01-16T06:29:26Z
lassestilvang/todo-kimi-k2-thinking		public	2025-12-26T10:56:22Z

kyegomez/MuonClip	This repository is an open source implementation of the MuonClip strategy from the KIMI K2 Model from Moonshot AI	public	2026-01-04T05:19:59Z

kvcache-ai/ktransformers:doc/en/Kimi-K2-Thinking.md: Please Note This is Quantization Deployment. For Native Kimi K2 Thinking deployment please refer to [here](./Kimi-K2-Thinking-Native.md).
kvcache-ai/ktransformers:doc/en/Kimi-K2-Thinking.md: * huggingface: https://huggingface.co/moonshotai/Kimi-K2-Thinking
kvcache-ai/ktransformers:doc/en/Kimi-K2-Thinking.md: * modelscope: https://modelscope.cn/models/moonshotai/Kimi-K2-Thinking
skypilot-org/skypilot:docs/source/examples/models/kimi-k2-thinking.md: ../../generated-examples/kimi-k2-thinking.md
NVIDIA/TensorRT-LLM:docs/source/deployment-guide/deployment-guide-for-kimi-k2-thinking-on-trtllm.md: * NVFP4 model: [Kimi-K2-Thinking-NVFP4](https://huggingface.co/nvidia/Kimi-K2-Thinking-NVFP4)
radixark/miles:scripts/run-kimi-k2-Thinking.sh: source "${SCRIPT_DIR}/models/kimi-k2-thinking.sh"
radixark/miles:scripts/run-kimi-k2-Thinking.sh: # --hf-checkpoint $BASE_DIR/Kimi-K2-Thinking-bf16/
radixark/miles:scripts/run-kimi-k2-Thinking.sh: --hf-checkpoint $BASE_DIR/Kimi-K2-Thinking-fp8/
browser-use/browser-use:examples/models/moonshot.py: model='kimi-k2-thinking',
exo-explore/exo:tmp/run_llm.sh: \"model\": \"mlx-community/Kimi-K2-Thinking\",
ggml-org/llama.cpp:tools/server/README.md: ├─ Kimi-K2-Thinking-UD-IQ1_S
ggml-org/llama.cpp:tools/server/README.md: │    ├─ Kimi-K2-Thinking-UD-IQ1_S-00001-of-00006.gguf
ggml-org/llama.cpp:tools/server/README.md: │    ├─ Kimi-K2-Thinking-UD-IQ1_S-00002-of-00006.gguf
cline/cline:docs/provider-config/baseten.mdx: **IMPORTANT: For Kimi K2 Thinking:** To use the `moonshotai/Kimi-K2-Thinking` model, you must enable **Native Tool Call (Experimental)** in Cline settings. This setting allows Cline to call tools through their native tool processor and is required for this reasoning model to function properly.
cline/cline:docs/provider-config/baseten.mdx: -   `moonshotai/Kimi-K2-Thinking` (Moonshot AI) - Enhanced reasoning capabilities with step-by-step thought processes (262K context) - \$0.60/\$2.50 pe...

# Kimi Writing Agent

An autonomous agent powered by the **kimi-k2-thinking** model for creating novels, books, and short story collections.

## Features

- 🤖 **Autonomous Writing**: The agent plans and executes creative writing tasks independently
- 📚 **Multiple Formats**: Create novels, books, or short story collections
- ⚡ **Real-Time Streaming**: See the agent's reasoning and writing appear as it's generated
- 💾 **Smart Context Management**: Automatically compresses context when approaching token limits
- 🔄 **Recovery Mode**: Resume interrupted work from saved context summaries
- 📊 **Token Monitoring**: Real-time tracking of token usage with automatic optimization
- 🛠️ **Tool Use**: Agent can create projects, write files, and manage its workspace

## Installation

### Prerequisites

We recommend using [uv](https://github.com/astral-sh/uv) for fast Python package management:

```bash
# Install uv (if you don't have it)
curl -LsSf https://astral.sh/uv/install.sh | sh
```

### Setup

1. Install dependencies:

**Using uv (recommended):**
```bash
uv pip install -r requirements.txt
```

**Or using pip:**
```bash
pip install -r requirements.txt
```

2. Configure your API key:

Create a `.env` file with your API key:
```bash
# Copy the example file
cp env.example .env

# Edit .env and add your API key
# The file should contain:
MOONSHOT_API_KEY=your-api-key-here
```

**Optional:** Set custom base URL (defaults to https://api.moonshot.ai/v1):
```bash
# Add to your .env file:
MOONSHOT_BASE_URL=https://api.moonshot.ai/v1
```

## Usage

### Fresh Start

Run with an inline prompt:
```bash
# Using uv (recommended)
uv run kimi-writer.py "Create a collection of 5 sci-fi short stories about AI"

# Or using python directly
python kimi-writer.py "Create a collection of 5 sci-fi short stories about AI"
```

Or run interactively:
```bash
uv run kimi-writer.py
# or: python kimi-writer.py
```
Then enter your prompt when asked.

### Recovery Mode

If the agent is interrupted or y...

# Kimi K2 Thinking Demos

> **As Seen On the YouTube Video:** [Goodbye Claude Code MAX... Kimi K2 THINKING Is All You Need](https://www.youtube.com/watch?v=kpDZ3fcywwk)

This repository contains prompts for building two complete browser-based applications using Kimi K2 Thinking and Claude Code. 

**[🚀 Want the FULL source code + more? Get it at BuilderPack.ai](https://rebrand.ly/6294rql)**

## 🎮 Demo Previews

<img src="kimikraft-preview.gif" alt="KimiKraft Demo" width="480" />
<img src="kimios-preview.gif" alt="KimiOS Demo" width="480" />

---

## 🎒 Want The Complete Source Code?

This repository **only** contains the prompts and documentation. 

If you want access to the complete working source code and additional resources:

**[🚀 Get the BuilderPack at builderpack.ai](https://rebrand.ly/6294rql)**

The BuilderPack includes:
- ✅ Complete working source code for KimiKraft and KimiOS (fully commented)
- ✅ Enhanced iterations (v2 versions with improvements and bug fixes)
- ✅ Bonus experimental applications (AI Debate Arena, Compare Anything, SimWorld, Mystery Mansion)
- ✅ Implementation notes and learnings from the development process
- ✅ Claude Code Skills for Kimi K2 Thinking / Moonshot API (Experimental)

**One-time purchase • Lifetime access • All updates included**

Join the community of BuilderPack Backpackers getting exclusive access to all source code, iterations, and bonus content.

---

---

## 📝 The Prompts

All prompts are located in the `prompts/` directory. These are the actual prompts used to build the applications shown in the video.

### KimiKraft Prompts:

**`minecraft_clone_prompt.txt`** (1KB) - Basic version
- Core requirements and features
- Quick start for building a simple voxel game
- Block types and basic mechanics

**`minecraft_clone_prompt_enhanced.txt`** (6.4KB) - Production version
- Complete technical specifications
- Quaternion camera implementation details
- Advanced raycasting with multiple intersectable objects
- Chunk loading a...

# Kimi K2 Thinking Agentic Search and Browsing

这是 [K2 Thinking 发布博客](https://moonshotai.github.io/Kimi-K2/thinking.html) 中 Agentic Search and Browsing 一节的非官方复现。

在博客最后提到

> To ensure a fast, lightweight experience, we selectively employ a subset of tools and reduce the number of tool call turns under the chat mode on kimi.com. As a result, chatting on kimi.com may not reproduce our benchmark scores. Our agentic mode will be updated soon to reflect the full capabilities of K2 Thinking.

因此 Kimi Web 版本在开启长推理 + 联网查询后并不会连续搜索，而[官方样例](https://platform.moonshot.cn/docs/guide/use-kimi-k2-thinking-model)非常不完整。

不过在博客中，相关轨迹的原始数据被完整的放在了前端中（参考 `samples/official_trace.json` 文件），这让我们可以尝试在官方 Agent 开源前复现一下相关工作。

主要复现的功能有：
- 富有 Kimi 特色的 system prompt
- tool function 相关定义，和只保留最近 3 轮搜索结果的特性
- 对其 tool function 的返回 prompt
- Kimi 给的样例不是用的 kimi 自家的搜索工具，大概用的 Bing 国内版 + 页面全文获取，而我们使用 Jina AI的搜索接口
   * Kimi 自建搜索：在官方 API 和网页版中使用
   * 官方样例搜索：推测使用 Bing 国内版，并额外实现全文获取补充搜索内容
   * Jina AI 搜索：实际调用 Google，提供页面全文

此外，还实现了前端方便查看轨迹。

## 安装依赖

```bash
pip install -r requirements.txt
# 填写你自己的密钥
mv env.example .env
```

## 运行 prompt

简单 prompt，来自官方文档：

```bash
python3 cli.py -q "请帮我生成一份今日新闻报告"
```

复杂 prompt，来自官方博客：

```bash
python3 cli.py -q "The information below is about an individual who - is an alumnus of a university founded after 1860 but before 1890 - was a university athlete and later played for a professional American football team briefly - starred in a science fiction film about an alien invasion that was released after 2010 and before 2020 - played a Corrections Officer in a prison drama that premiered between 2010 and 2020 (in one episode, their character signs out and releases the wrong inmate) - stated in an interview that the character they loved playing the most was the one in a medical drama that premiered after 2001 but before 2010  Name the character they played in the science fiction movie."
```

## 前端查看

```bash
python3 frontend.py
```

应用将在 `http://localhost:5005` 启动。

## 免责声明

1. 本...

404: Not Found

# MuonClip

This is a simple implementation of MuonClip in pure pytorch from the Kimi K2 Paper. [Read it here](https://arxiv.org/pdf/2507.20534)

## Code

```python
import torch
import torch.nn as nn
from typing import Optional, Dict, Any, List
import math


def newton_schulz_iteration(M: torch.Tensor, num_iterations: int = 5) -> torch.Tensor:
    """
    Perform Newton-Schulz iteration to orthogonalize a matrix.
    
    The Newton-Schulz iteration computes an approximate orthogonal matrix
    from M by iteratively refining: X_{k+1} = X_k * (3I - X_k^T X_k) / 2
    
    Args:
        M: Input matrix to orthogonalize
        num_iterations: Number of iterations (default: 5)
    
    Returns:
        Approximately orthogonal matrix
    """
    # Initialize with normalized input
    X = M / (M.norm() + 1e-7)
    
    # Iteratively refine
    for _ in range(num_iterations):
        # X = X * (3I - X^T X) / 2
        X = 1.5 * X - 0.5 * X @ (X.T @ X)
    
    return X


class MuonClip(torch.optim.Optimizer):
    """
    MuonClip optimizer as described in Kimi K2 paper.
    
    Combines the token-efficient Muon optimizer with QK-Clip for stability.
    
    Args:
        params: Iterable of parameters to optimize
        lr: Learning rate (default: 2e-4)
        momentum: Momentum coefficient (default: 0.95)
        weight_decay: Weight decay coefficient (default: 0.1)
        qk_clip_threshold: Maximum attention logit threshold τ (default: 100.0)
        qk_clip_alpha: Balance parameter for Q/K scaling (default: 0.5)
        newton_schulz_iters: Number of Newton-Schulz iterations (default: 5)
        rms_scale_factor: RMS scaling factor (default: 0.2)
    """
    
    def __init__(
        self,
        params,
        lr: float = 2e-4,
        momentum: float = 0.95,
        weight_decay: float = 0.1,
        qk_clip_threshold: float = 100.0,
        qk_clip_alpha: float = 0.5,
        newton_schulz_iters: int = 5,
        rms_scale_factor: float = 0.2,
    ):
       ...

---
title: "Baseten"
description: "Learn how to configure and use Baseten's Model APIs with Cline. Access frontier open-source models with enterprise-grade performance, reliability, and competitive pricing."
---

Baseten provides on-demand frontier model APIs designed for production applications, not just experimentation. Built on the Baseten Inference Stack, these APIs deliver optimized inference for leading open-source models from OpenAI, DeepSeek, Moonshot AI, and Alibaba Cloud.

**Website:** [https://www.baseten.co/products/model-apis/](https://www.baseten.co/products/model-apis/)

### Getting an API Key

1.  **Sign Up/Sign In:** Go to [Baseten](https://www.baseten.co/) and create an account or sign in.
2.  **Navigate to API Keys:** Access your dashboard and go to the API Keys section.
3.  **Create a Key:** Generate a new API key. Give it a descriptive name (e.g., "Cline").
4.  **Copy the Key:** Copy the API key immediately and store it securely.

### Configuration in Cline

1.  **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
2.  **Select Provider:** Choose "Baseten" from the "API Provider" dropdown.
3.  **Enter API Key:** Paste your Baseten API key into the "Baseten API Key" field.
4.  **Select Model:** Choose your desired model from the "Model" dropdown.

**IMPORTANT: For Kimi K2 Thinking:** To use the `moonshotai/Kimi-K2-Thinking` model, you must enable **Native Tool Call (Experimental)** in Cline settings. This setting allows Cline to call tools through their native tool processor and is required for this reasoning model to function properly.

### Supported Models

Cline supports all current models under Baseten Model APIs, including:
For the most updated pricing, please visit: https://www.baseten.co/products/model-apis/

-   `moonshotai/Kimi-K2-Thinking` (Moonshot AI) - Enhanced reasoning capabilities with step-by-step thought processes (262K context) - \$0.60/\$2.50 per 1M tokens
-   `zai-org/GLM-4.6` (Z AI) - Frontier open model ...

<!DOCTYPE html>
<html lang="en">

<head>  <title>[2507.20534] Kimi K2: Open Agentic Intelligence</title>
  <meta name="viewport" content="width=device-width, initial-scale=1">
  <link rel="apple-touch-icon" sizes="180x180" href="/static/browse/0.3.4/images/icons/apple-touch-icon.png">
  <link rel="icon" type="image/png" sizes="32x32" href="/static/browse/0.3.4/images/icons/favicon-32x32.png">
  <link rel="icon" type="image/png" sizes="16x16" href="/static/browse/0.3.4/images/icons/favicon-16x16.png">
  <link rel="manifest" href="/static/browse/0.3.4/images/icons/site.webmanifest">
  <link rel="mask-icon" href="/static/browse/0.3.4/images/icons/safari-pinned-tab.svg" color="#5bbad5">
  <meta name="msapplication-TileColor" content="#da532c">
  <meta name="theme-color" content="#ffffff">
  <link rel="stylesheet" type="text/css" media="screen" href="/static/browse/0.3.4/css/arXiv.css?v=20241206" />
  <link rel="stylesheet" type="text/css" media="print" href="/static/browse/0.3.4/css/arXiv-print.css?v=20200611" />
  <link rel="stylesheet" type="text/css" media="screen" href="/static/browse/0.3.4/css/browse_search.css" />
  <script language="javascript" src="/static/browse/0.3.4/js/accordion.js" /></script>
  
  <link rel="canonical" href="https://arxiv.org/abs/2507.20534"/>
  <meta name="description" content="Abstract page for arXiv paper 2507.20534: Kimi K2: Open Agentic Intelligence"><meta property="og:type" content="website" />
<meta property="og:site_name" content="arXiv.org" />
<meta property="og:title" content="Kimi K2: Open Agentic Intelligence" />
<meta property="og:url" content="https://arxiv.org/abs/2507.20534v1" />
<meta property="og:image" content="/static/browse/0.3.4/images/arxiv-logo-fb.png" />
<meta property="og:image:secure_url" content="/static/browse/0.3.4/images/arxiv-logo-fb.png" />
<meta property="og:image:width" content="1200" />
<meta property="og:image:height" content="700" />
<meta property="og:image:alt" content="arXiv logo"/>
<meta property="o...

gatencia/AgrI:orchestrator.py: class Tool:
gatencia/AgrI:orchestrator.py: """Base class for tools that the agent can use."""
BerriAI/litellm:litellm/llms/bedrock/chat/invoke_transformations/amazon_moonshot_transformation.py: - `tools` (list) tool definitions (supported on kimi-k2-thinking)
BerriAI/litellm:litellm/llms/bedrock/chat/invoke_transformations/amazon_moonshot_transformation.py: - `tool_choice` (str|dict) tool choice specification (supported on kimi-k2-thinking)
kvcache-ai/ktransformers:kt-kernel/python/cli/utils/model_registry.py: "served-model-name": "Kimi-K2-Thinking",
kvcache-ai/ktransformers:kt-kernel/python/cli/utils/model_registry.py: "tool-call-parser": "minimax-m2",
zixianglhhh/MCPAgentBench:funcs/plot_tool_count_tefs.py: "kimi-k2-thinking",
zixianglhhh/MCPAgentBench:funcs/plot_tool_count_tefs.py: tool_counts = [10, 20, 30, 40]
inference-sh/grid:api/openrouter/kimi-k2-thinking/inference.py: DEFAULT_MODEL = "moonshotai/kimi-k2-thinking"
inference-sh/grid:api/openrouter/kimi-k2-thinking/inference.py: class AppOutput(ReasoningMixin, ToolCallsMixin, LLMOutput, BaseAppOutput):
inference-sh/grid:api/openrouter/kimi-k2-thinking/inference.py: """OpenRouter output model with reasoning, tool calls, and usage information."""
xiehust/awesome-skills-claude-agents:backend/demo_boto3_tool_use.py: def run_converse_with_tools(user_message: str, model_id: str = "moonshot.kimi-k2-thinking"):
xiehust/awesome-skills-claude-agents:backend/demo_boto3_tool_use.py: print("\n📍 Example 4: Conversation Without Tools")
xiehust/awesome-skills-claude-agents:backend/demo_boto3_tool_use.py: run_converse_with_tools("What are some popular tourist attractions in Paris?", model_id=model_id)
thinking-machines-lab/tinker-cookbook:tinker_cookbook/renderers/kimi_k2.py: return tool_calls, unparsed_tool_calls
thinking-machines-lab/tinker-cookbook:tinker_cookbook/renderers/kimi_k2.py: Format for moonshotai/Kimi-K2-Thinking:
Doriandarko/kimi-writer:utils.py: Returns the tool definitions in t...

import asyncio
import os

from dotenv import load_dotenv

from browser_use import Agent, ChatOpenAI

load_dotenv()

# Get API key from environment variable
api_key = os.getenv('MOONSHOT_API_KEY')
if api_key is None:
	print('Make sure you have MOONSHOT_API_KEY set in your .env file')
	print('Get your API key from https://platform.moonshot.ai/console/api-keys ')
	exit(1)

# Configure Moonshot AI model
llm = ChatOpenAI(
	model='kimi-k2-thinking',
	base_url='https://api.moonshot.ai/v1',
	api_key=api_key,
	add_schema_to_system_prompt=True,
	remove_min_items_from_schema=True,  # Moonshot doesn't support minItems in JSON schema
	remove_defaults_from_schema=True,  # Moonshot doesn't allow default values with anyOf
)


async def main():
	agent = Agent(
		task='Search for the latest news about AI and summarize the top 3 articles',
		llm=llm,
		flash_mode=True,
	)
	await agent.run()


if __name__ == '__main__':
	asyncio.run(main())

"""
Utility functions for the Kimi Writing Agent.
"""

import json
import httpx
from typing import List, Dict, Any, Callable


def estimate_token_count(base_url: str, api_key: str, model: str, messages: List[Dict]) -> int:
    """
    Estimate the token count for the given messages using the Moonshot API.
    
    Note: Token estimation uses api.moonshot.ai (not .cn)
    
    Args:
        base_url: The base URL for the API (will be converted to .ai for token endpoint)
        api_key: The API key for authentication
        model: The model name
        messages: List of message dictionaries
        
    Returns:
        Total token count
    """
    # Convert messages to serializable format (remove non-serializable objects)
    serializable_messages = []
    for msg in messages:
        if hasattr(msg, 'model_dump'):
            # OpenAI SDK message object
            msg_dict = msg.model_dump()
        elif isinstance(msg, dict):
            msg_dict = msg.copy()
        else:
            msg_dict = {"role": "assistant", "content": str(msg)}
        
        # Clean up the message to only include serializable fields
        clean_msg = {}
        if 'role' in msg_dict:
            clean_msg['role'] = msg_dict['role']
        if 'content' in msg_dict and msg_dict['content']:
            clean_msg['content'] = msg_dict['content']
        if 'name' in msg_dict:
            clean_msg['name'] = msg_dict['name']
        if 'tool_calls' in msg_dict and msg_dict['tool_calls']:
            clean_msg['tool_calls'] = msg_dict['tool_calls']
        if 'tool_call_id' in msg_dict:
            clean_msg['tool_call_id'] = msg_dict['tool_call_id']
            
        serializable_messages.append(clean_msg)
    
    # Both token estimation and chat use api.moonshot.ai
    token_base_url = base_url
    
    # Make the API call
    with httpx.Client(
        base_url=token_base_url,
        headers={"Authorization": f"Bearer {api_key}"},
        timeout=30.0
    ) as client:
       ...

import copy
import os
import uuid
from collections.abc import AsyncIterator, Sequence
from typing import TYPE_CHECKING, Any, Self, TypedDict, Unpack, cast

import httpx
from openai import AsyncOpenAI, AsyncStream, OpenAIError, omit
from openai.types.chat import (
    ChatCompletion,
    ChatCompletionChunk,
    ChatCompletionMessageFunctionToolCall,
    ChatCompletionMessageParam,
    ChatCompletionToolParam,
)
from openai.types.completion_usage import CompletionUsage

from kosong.chat_provider import (
    ChatProvider,
    ChatProviderError,
    StreamedMessagePart,
    ThinkingEffort,
    TokenUsage,
)
from kosong.chat_provider.openai_common import convert_error, tool_to_openai
from kosong.message import ContentPart, Message, TextPart, ThinkPart, ToolCall, ToolCallPart
from kosong.tooling import Tool

if TYPE_CHECKING:

    def type_check(kimi: "Kimi"):
        _: ChatProvider = kimi


class Kimi:
    """
    A chat provider that uses the Kimi API.

    >>> chat_provider = Kimi(model="kimi-k2-turbo-preview", api_key="sk-1234567890")
    >>> chat_provider.name
    'kimi'
    >>> chat_provider.model_name
    'kimi-k2-turbo-preview'
    >>> chat_provider.with_generation_kwargs(temperature=0)._generation_kwargs
    {'temperature': 0}
    >>> chat_provider._generation_kwargs
    {}
    """

    name = "kimi"

    class GenerationKwargs(TypedDict, total=False):
        """
        See https://platform.moonshot.ai/docs/api/chat#request-body.
        """

        max_tokens: int | None
        temperature: float | None
        top_p: float | None
        n: int | None
        presence_penalty: float | None
        frequency_penalty: float | None
        stop: str | list[str] | None
        prompt_cache_key: str | None
        reasoning_effort: str | None

    def __init__(
        self,
        *,
        model: str,
        api_key: str | None = None,
        base_url: str | None = None,
        stream: bool = True,
        **client_kwargs: Any,
    ):
        if api_ke...

[{"name":"__init__.py","path":"packages/kosong/src/kosong/chat_provider/__init__.py","sha":"622564ad88b1fb5d860b80d051664cc83b574a07","size":3695,"url":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/__init__.py?ref=main","html_url":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/__init__.py","git_url":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/622564ad88b1fb5d860b80d051664cc83b574a07","download_url":"https://raw.githubusercontent.com/MoonshotAI/kimi-cli/main/packages/kosong/src/kosong/chat_provider/__init__.py","type":"file","_links":{"self":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/__init__.py?ref=main","git":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/622564ad88b1fb5d860b80d051664cc83b574a07","html":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/__init__.py"}},{"name":"chaos.py","path":"packages/kosong/src/kosong/chat_provider/chaos.py","sha":"0436781a9d01b50dfca7477896d2d72312253cf7","size":9793,"url":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/chaos.py?ref=main","html_url":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/chaos.py","git_url":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/0436781a9d01b50dfca7477896d2d72312253cf7","download_url":"https://raw.githubusercontent.com/MoonshotAI/kimi-cli/main/packages/kosong/src/kosong/chat_provider/chaos.py","type":"file","_links":{"self":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/chaos.py?ref=main","git":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/0436781a9d01b50dfca7477896d2d72312253cf7","html":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/chaos.py"}},{"name":"echo....

<file>
00251|                 const modelRequiresPrefix = ["claude", "nova-lite", "nova-micro", "nova-pro"].some((m) =>
00252|                   modelID.includes(m),
00253|                 )
00254|                 if (modelRequiresPrefix) {
00255|                   regionPrefix = "apac"
00256|                   modelID = `${regionPrefix}.${modelID}`
00257|                 }
00258|               }
00259|               break
00260|             }
00261|           }
00262| 
00263|           return sdk.languageModel(modelID)
00264|         },
00265|       }
00266|     },
00267|     openrouter: async () => {
00268|       return {
00269|         autoload: false,
00270|         options: {
00271|           headers: {
00272|             "HTTP-Referer": "https://opencode.ai/",
00273|             "X-Title": "opencode",
00274|           },
00275|         },
00276|       }
00277|     },
00278|     vercel: async () => {
00279|       return {
00280|         autoload: false,
00281|         options: {
00282|           headers: {
00283|             "http-referer": "https://opencode.ai/",
00284|             "x-title": "opencode",
00285|           },
00286|         },
00287|       }
00288|     },
00289|     "google-vertex": async () => {
00290|       const project = Env.get("GOOGLE_CLOUD_PROJECT") ?? Env.get("GCP_PROJECT") ?? Env.get("GCLOUD_PROJECT")
00291|       const location = Env.get("GOOGLE_CLOUD_LOCATION") ?? Env.get("VERTEX_LOCATION") ?? "us-east5"
00292|       const autoload = Boolean(project)
00293|       if (!autoload) return { autoload: false }
00294|       return {
00295|         autoload: true,
00296|         options: {
00297|           project,
00298|           location,
00299|         },
00300|         async getModel(sdk: any, modelID: string) {
00301|           const id = String(modelID).trim()
00302|           return sdk.languageModel(id)
00303|         },
00304|       }
00305|     },
00306|     "google-vertex-anthropic": async () => {
00307|       const project = Env....

<file>
00001| import type { APICallError, ModelMessage } from "ai"
00002| import { unique } from "remeda"
00003| import type { JSONSchema } from "zod/v4/core"
00004| import type { Provider } from "./provider"
00005| import type { ModelsDev } from "./models"
00006| 
00007| type Modality = NonNullable<ModelsDev.Model["modalities"]>["input"][number]
00008| 
00009| function mimeToModality(mime: string): Modality | undefined {
00010|   if (mime.startsWith("image/")) return "image"
00011|   if (mime.startsWith("audio/")) return "audio"
00012|   if (mime.startsWith("video/")) return "video"
00013|   if (mime === "application/pdf") return "pdf"
00014|   return undefined
00015| }
00016| 
00017| export namespace ProviderTransform {
00018|   function normalizeMessages(msgs: ModelMessage[], model: Provider.Model): ModelMessage[] {
00019|     if (model.api.id.includes("claude")) {
00020|       return msgs.map((msg) => {
00021|         if ((msg.role === "assistant" || msg.role === "tool") && Array.isArray(msg.content)) {
00022|           msg.content = msg.content.map((part) => {
00023|             if ((part.type === "tool-call" || part.type === "tool-result") && "toolCallId" in part) {
00024|               return {
00025|                 ...part,
00026|                 toolCallId: part.toolCallId.replace(/[^a-zA-Z0-9_-]/g, "_"),
00027|               }
00028|             }
00029|             return part
00030|           })
00031|         }
00032|         return msg
00033|       })
00034|     }
00035|     if (model.providerID === "mistral" || model.api.id.toLowerCase().includes("mistral")) {
00036|       const result: ModelMessage[] = []
00037|       for (let i = 0; i < msgs.length; i++) {
00038|         const msg = msgs[i]
00039|         const nextMsg = msgs[i + 1]
00040| 
00041|         if ((msg.role === "assistant" || msg.role === "tool") && Array.isArray(msg.content)) {
00042|           msg.content = msg.content.map((part) => {
00043|             if ((part.type === "tool-call" ...

<file>
00078|       model.capabilities.interleaved &&
00079|       typeof model.capabilities.interleaved === "object" &&
00080|       model.capabilities.interleaved.field === "reasoning_content"
00081|     ) {
00082|       return msgs.map((msg) => {
00083|         if (msg.role === "assistant" && Array.isArray(msg.content)) {
00084|           const reasoningParts = msg.content.filter((part: any) => part.type === "reasoning")
00085|           const reasoningText = reasoningParts.map((part: any) => part.text).join("")
00086| 
00087|           // Filter out reasoning parts from content
00088|           const filteredContent = msg.content.filter((part: any) => part.type !== "reasoning")
00089| 
00090|           // Include reasoning_content directly on the message for all assistant messages
00091|           if (reasoningText) {
00092|             return {
00093|               ...msg,
00094|               content: filteredContent,
00095|               providerOptions: {
00096|                 ...msg.providerOptions,
00097|                 openaiCompatible: {
00098|                   ...(msg.providerOptions as any)?.openaiCompatible,
00099|                   reasoning_content: reasoningText,
00100|                 },
00101|               },
00102|             }
00103|           }
00104| 
00105|           return {
00106|             ...msg,
00107|             content: filteredContent,
00108|           }
00109|         }
00110| 
00111|         return msg
00112|       })
00113|     }
00114| 
00115|     return msgs
00116|   }
00117| 

(File has more lines. Use 'offset' parameter to read beyond line 117)
</file>

/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:        const reasoningMessages: Record<string, OpenAIResponsesReasoning> = {}
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:            case "reasoning": {
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:              const reasoningId = providerOptions?.itemId
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:              if (reasoningId != null) {
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:                const reasoningMessage = reasoningMessages[reasoningId]
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:                  if (reasoningMessage === undefined) {
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:                    // use item references to refer to reasoning (single reference)
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:                    input.push({ type: "item_reference", id: reasoningId })
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:                    // store unused reasoning message to mark id as used
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/convert-to-openai-responses-input.ts:                    reasoningMessages[reasoningId] = {
/home/jm/data/code/opencode/packages/opencode/src/prov...

<file>
00501|       input_select_left: z.string().optional().default("shift+left").describe("Select left in input"),
00502|       input_select_right: z.string().optional().default("shift+right").describe("Select right in input"),
00503|       input_select_up: z.string().optional().default("shift+up").describe("Select up in input"),
00504|       input_select_down: z.string().optional().default("shift+down").describe("Select down in input"),
00505|       input_line_home: z.string().optional().default("ctrl+a").describe("Move to start of line in input"),
00506|       input_line_end: z.string().optional().default("ctrl+e").describe("Move to end of line in input"),
00507|       input_select_line_home: z
00508|         .string()
00509|         .optional()
00510|         .default("ctrl+shift+a")
00511|         .describe("Select to start of line in input"),
00512|       input_select_line_end: z.string().optional().default("ctrl+shift+e").describe("Select to end of line in input"),
00513|       input_visual_line_home: z.string().optional().default("alt+a").describe("Move to start of visual line in input"),
00514|       input_visual_line_end: z.string().optional().default("alt+e").describe("Move to end of visual line in input"),
00515|       input_select_visual_line_home: z
00516|         .string()
00517|         .optional()
00518|         .default("alt+shift+a")
00519|         .describe("Select to start of visual line in input"),
00520|       input_select_visual_line_end: z
00521|         .string()
00522|         .optional()
00523|         .default("alt+shift+e")
00524|         .describe("Select to end of visual line in input"),
00525|       input_buffer_home: z.string().optional().default("home").describe("Move to start of buffer in input"),
00526|       input_buffer_end: z.string().optional().default("end").describe("Move to end of buffer in input"),
00527|       input_select_buffer_home: z
00528|         .string()
00529|         .optional()
00530|         .default("shift+...

/home/jm/data/code/opencode/packages/opencode/src/agent/agent.ts:    const agentPermission = mergeAgentPermissions(defaultPermission, cfg.permission ?? {})
/home/jm/data/code/opencode/packages/opencode/src/agent/agent.ts:  export async function defaultAgent(): Promise<string> {
/home/jm/data/code/opencode/packages/opencode/src/agent/agent.ts:    const defaultCandidate = Object.values(agents).find((a) => a.default)
/home/jm/data/code/opencode/packages/opencode/src/cli/cmd/github.ts:          // agent is omitted - server will use default_agent from config or fall back to "build"
/home/jm/data/code/opencode/packages/opencode/src/cli/cmd/run.ts:            `agent "${args.agent}" not found. Falling back to default agent`,
/home/jm/data/code/opencode/packages/opencode/src/cli/cmd/run.ts:            `agent "${args.agent}" is a subagent, not a primary agent. Falling back to default agent`,
/home/jm/data/code/opencode/packages/opencode/src/config/config.ts:      agent_list: z.string().optional().default("<leader>a").describe("List agents"),
/home/jm/data/code/opencode/packages/opencode/src/config/config.ts:      agent_cycle: z.string().optional().default("tab").describe("Next agent"),
/home/jm/data/code/opencode/packages/opencode/src/config/config.ts:      agent_cycle_reverse: z.string().optional().default("shift+tab").describe("Previous agent"),
/home/jm/data/code/opencode/packages/opencode/src/server/server.ts:          let currentAgent = await Agent.defaultAgent()

/home/jm/data/code/opencode/packages/opencode/src/session/compaction.ts:  // goes backwards through parts until there are 40_000 tokens worth of tool

<file>
00002| import os from "os"
00003| import fs from "fs/promises"
00004| import z from "zod"
00005| import { Identifier } from "../id/id"
00006| import { MessageV2 } from "./message-v2"
00007| import { Log } from "../util/log"
00008| import { SessionRevert } from "./revert"
00009| import { Session } from "."
00010| import { Agent } from "../agent/agent"
00011| import { Provider } from "../provider/provider"
00012| import { type Tool as AITool, tool, jsonSchema } from "ai"
00013| import { SessionCompaction } from "./compaction"
00014| import { Instance } from "../project/instance"
00015| import { Bus } from "../bus"
00016| import { ProviderTransform } from "../provider/transform"
00017| import { SystemPrompt } from "./system"
00018| import { Plugin } from "../plugin"
00019| import PROMPT_PLAN from "../session/prompt/plan.txt"
00020| import BUILD_SWITCH from "../session/prompt/build-switch.txt"
00021| import MAX_STEPS from "../session/prompt/max-steps.txt"
00022| import { defer } from "../util/defer"
00023| import { clone, mergeDeep, pipe } from "remeda"
00024| import { ToolRegistry } from "../tool/registry"
00025| import { Wildcard } from "../util/wildcard"
00026| import { MCP } from "../mcp"
00027| import { LSP } from "../lsp"
00028| import { ReadTool } from "../tool/read"
00029| import { ListTool } from "../tool/ls"
00030| import { FileTime } from "../file/time"
00031| import { Flag } from "../flag/flag"
00032| import { ulid } from "ulid"
00033| import { spawn } from "child_process"
00034| import { Command } from "../command"
00035| import { $, fileURLToPath } from "bun"
00036| import { ConfigMarkdown } from "../config/markdown"
00037| import { SessionSummary } from "./summary"
00038| import { NamedError } from "@opencode-ai/util/error"
00039| import { fn } from "@/util/fn"
00040| import { SessionProcessor } from "./processor"
00041| import { TaskTool } from "@/tool/task"
00042| import { SessionStatus } from "./status"
00043| import { LLM } from "./llm"
00044| ...

<file>
00001| import { Config } from "../config/config"
00002| import z from "zod"
00003| import { Provider } from "../provider/provider"
00004| import { generateObject, type ModelMessage } from "ai"
00005| import { SystemPrompt } from "../session/system"
00006| import { Instance } from "../project/instance"
00007| import { mergeDeep } from "remeda"
00008| import { Log } from "../util/log"
00009| 
00010| const log = Log.create({ service: "agent" })
00011| 
00012| import PROMPT_GENERATE from "./generate.txt"
00013| import PROMPT_COMPACTION from "./prompt/compaction.txt"
00014| import PROMPT_EXPLORE from "./prompt/explore.txt"
00015| import PROMPT_SUMMARY from "./prompt/summary.txt"
00016| import PROMPT_TITLE from "./prompt/title.txt"
00017| 
00018| export namespace Agent {
00019|   export const Info = z
00020|     .object({
00021|       name: z.string(),
00022|       description: z.string().optional(),
00023|       mode: z.enum(["subagent", "primary", "all"]),
00024|       native: z.boolean().optional(),
00025|       hidden: z.boolean().optional(),
00026|       default: z.boolean().optional(),
00027|       topP: z.number().optional(),
00028|       temperature: z.number().optional(),
00029|       color: z.string().optional(),
00030|       permission: z.object({
00031|         edit: Config.Permission,
00032|         bash: z.record(z.string(), Config.Permission),
00033|         skill: z.record(z.string(), Config.Permission),
00034|         webfetch: Config.Permission.optional(),
00035|         doom_loop: Config.Permission.optional(),
00036|         external_directory: Config.Permission.optional(),
00037|       }),
00038|       model: z
00039|         .object({
00040|           modelID: z.string(),
00041|           providerID: z.string(),
00042|         })
00043|         .optional(),
00044|       prompt: z.string().optional(),
00045|       tools: z.record(z.string(), z.boolean()),
00046|       options: z.record(z.string(), z.any()),
00047|       maxSteps: z.number().int(...

/home/jm/data/code/opencode/packages/opencode/src/cli/cmd/stats.ts:  describe: "show token usage and cost statistics",
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:        inputTokens: response.usage.input_tokens,
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:        outputTokens: response.usage.output_tokens,
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:        totalTokens: response.usage.input_tokens + response.usage.output_tokens,
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:        reasoningTokens: response.usage.output_tokens_details?.reasoning_tokens ?? undefined,
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:        cachedInputTokens: response.usage.input_tokens_details?.cached_tokens ?? undefined,
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:              usage.inputTokens = value.response.usage.input_tokens
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:              usage.outputTokens = value.response.usage.output_tokens
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:              usage.totalTokens = value.response.usage.input_tokens + value.response.usage.output_tokens
/home/jm/data/code/opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts:              usage.reasoningTokens = value.response.usage.output_tokens_details?.reasoning_tokens ...

<file>
00001| import { Provider } from "@/provider/provider"
00002| import { Log } from "@/util/log"
00003| import { streamText, wrapLanguageModel, type ModelMessage, type StreamTextResult, type Tool, type ToolSet } from "ai"
00004| import { clone, mergeDeep, pipe } from "remeda"
00005| import { ProviderTransform } from "@/provider/transform"
00006| import { Config } from "@/config/config"
00007| import { Instance } from "@/project/instance"
00008| import type { Agent } from "@/agent/agent"
00009| import type { MessageV2 } from "./message-v2"
00010| import { Plugin } from "@/plugin"
00011| import { SystemPrompt } from "./system"
00012| import { ToolRegistry } from "@/tool/registry"
00013| import { Flag } from "@/flag/flag"
00014| 
00015| export namespace LLM {
00016|   const log = Log.create({ service: "llm" })
00017| 
00018|   export const OUTPUT_TOKEN_MAX = Flag.OPENCODE_EXPERIMENTAL_OUTPUT_TOKEN_MAX || 32_000
00019| 
00020|   export type StreamInput = {
00021|     user: MessageV2.User
00022|     sessionID: string
00023|     model: Provider.Model
00024|     agent: Agent.Info
00025|     system: string[]
00026|     abort: AbortSignal
00027|     messages: ModelMessage[]
00028|     small?: boolean
00029|     tools: Record<string, Tool>
00030|     retries?: number
00031|   }
00032| 
00033|   export type StreamOutput = StreamTextResult<ToolSet, unknown>
00034| 
00035|   export async function stream(input: StreamInput) {
00036|     const l = log
00037|       .clone()
00038|       .tag("providerID", input.model.providerID)
00039|       .tag("modelID", input.model.id)
00040|       .tag("sessionID", input.sessionID)
00041|       .tag("small", (input.small ?? false).toString())
00042|       .tag("agent", input.agent.name)
00043|     l.info("stream", {
00044|       modelID: input.model.id,
00045|       providerID: input.model.providerID,
00046|     })
00047|     const [language, cfg] = await Promise.all([Provider.getLanguage(input.model), Config.get()])
00048| 
00049|     con...

  export async function defaultAgent(): Promise<string> {
    const agents = await state()
    const defaultCandidate = Object.values(agents).find((a) => a.default)
    return defaultCandidate?.name ?? "build"
  }

  export async function generate(input: { description: string; model?: { providerID: string; modelID: string } }) {
    const cfg = await Config.get()
    const defaultModel = input.model ?? (await Provider.defaultModel())
    const model = await Provider.getModel(defaultModel.providerID, defaultModel.modelID)
    const language = await Provider.getLanguage(model)
    const system = SystemPrompt.header(defaultModel.providerID)
    system.push(PROMPT_GENERATE)
    const existing = await list()
    const result = await generateObject({
      experimental_telemetry: {
        isEnabled: cfg.experimental?.openTelemetry,
        metadata: {
          userId: cfg.username ?? "unknown",
        },
      },
      temperature: 0.3,
      messages: [
        ...system.map(
          (item): ModelMessage => ({
            role: "system",
            content: item,
          }),
        ),
        {
          role: "user",

    const agents = await state()
    const defaultCandidate = Object.values(agents).find((a) => a.default)

<file>
00181|         options: {},
00182|         native: true,
00183|         hidden: true,
00184|         permission: agentPermission,
00185|         prompt: PROMPT_TITLE,
00186|         tools: {},
00187|       },
00188|       summary: {
00189|         name: "summary",
00190|         mode: "primary",
00191|         options: {},
00192|         native: true,
00193|         hidden: true,
00194|         permission: agentPermission,
00195|         prompt: PROMPT_SUMMARY,
00196|         tools: {},
00197|       },
00198|     }
00199|     for (const [key, value] of Object.entries(cfg.agent ?? {})) {
00200|       if (value.disable) {
00201|         delete result[key]
00202|         continue
00203|       }
00204|       let item = result[key]
00205|       if (!item)
00206|         item = result[key] = {
00207|           name: key,
00208|           mode: "all",
00209|           permission: agentPermission,
00210|           options: {},
00211|           tools: {},
00212|           native: false,
00213|         }
00214|       const {
00215|         name,
00216|         model,
00217|         prompt,
00218|         tools,
00219|         description,
00220|         temperature,
00221|         top_p,
00222|         mode,
00223|         permission,
00224|         color,
00225|         maxSteps,
00226|         ...extra
00227|       } = value
00228|       item.options = {
00229|         ...item.options,
00230|         ...extra,
00231|       }
00232|       if (model) item.model = Provider.parseModel(model)
00233|       if (prompt) item.prompt = prompt
00234|       if (tools)
00235|         item.tools = {
00236|           ...item.tools,
00237|           ...tools,
00238|         }
00239|       item.tools = {
00240|         ...defaultTools,
00241|         ...item.tools,
00242|       }
00243|       if (description) item.description = description
00244|       if (temperature != undefined) item.temperature = temperature
00245|       if (top_p != undefined) item.topP = top_p
00246|       if (...

      model_list: z.string().optional().default("<leader>m").describe("List available models"),
      model_cycle_recent: z.string().optional().default("f2").describe("Next recently used model"),
      model_cycle_recent_reverse: z.string().optional().default("shift+f2").describe("Previous recently used model"),
      model_cycle_favorite: z.string().optional().default("none").describe("Next favorite model"),
      model_cycle_favorite_reverse: z.string().optional().default("none").describe("Previous favorite model"),
      model: z.string().describe("Model to use in the format of provider/model, eg anthropic/claude-2").optional(),
      small_model: z
        .describe("Small model to use for tasks like title generation in the format of provider/model")
      default_agent: z
        const { provider, model, ...rest } = mod.default
        if (provider && model) result.model = `${provider}/${model}`

<file>
00131|             toolName: lower,
00132|           }
00133|         }
00134|         return {
00135|           ...failed.toolCall,
00136|           input: JSON.stringify({
00137|             tool: failed.toolCall.toolName,
00138|             error: failed.error.message,
00139|           }),
00140|           toolName: "invalid",
00141|         }
00142|       },
00143|       temperature: params.temperature,
00144|       topP: params.topP,
00145|       topK: params.topK,
00146|       providerOptions: ProviderTransform.providerOptions(input.model, params.options),
00147|       activeTools: Object.keys(tools).filter((x) => x !== "invalid"),
00148|       tools,
00149|       maxOutputTokens,
00150|       abortSignal: input.abort,
00151|       headers: {
00152|         ...(input.model.providerID.startsWith("opencode")
00153|           ? {
00154|               "x-opencode-project": Instance.project.id,
00155|               "x-opencode-session": input.sessionID,
00156|               "x-opencode-request": input.user.id,
00157|               "x-opencode-client": Flag.OPENCODE_CLIENT,
00158|             }
00159|           : undefined),
00160|         ...input.model.headers,
00161|       },
00162|       maxRetries: input.retries ?? 0,
00163|       messages: [
00164|         ...system.map(
00165|           (x): ModelMessage => ({
00166|             role: "system",
00167|             content: x,
00168|           }),
00169|         ),
00170|         ...input.messages,
00171|       ],
00172|       model: wrapLanguageModel({
00173|         model: language,
00174|         middleware: [
00175|           {
00176|             async transformParams(args) {
00177|               if (args.type === "stream") {
00178|                 // @ts-expect-error
00179|                 args.params.prompt = ProviderTransform.message(args.params.prompt, input.model)
00180|               }

(File has more lines. Use 'offset' parameter to read beyond line 180)
</file>

695:      default_agent: z

<file>
00951| 
00952|       throw new JsonError({
00953|         path: configFilepath,
00954|         message: `\n--- JSONC Input ---\n${text}\n--- Errors ---\n${errorDetails}\n--- End ---`,
00955|       })
00956|     }
00957| 
00958|     const parsed = Info.safeParse(data)
00959|     if (parsed.success) {
00960|       if (!parsed.data.$schema) {

(File has more lines. Use 'offset' parameter to read beyond line 960)
</file>

<file>
00694|         .optional(),
00695|       default_agent: z
00696|         .string()
00697|         .optional()
00698|         .describe(
00699|           "Default agent to use when none is specified. Must be a primary agent. Falls back to 'build' if not set or if the specified agent is invalid.",
00700|         ),
00701|       username: z
00702|         .string()
00703|         .optional()
00704|         .describe("Custom username to display in conversations instead of system username"),
00705|       mode: z
00706|         .object({
00707|           build: Agent.optional(),
00708|           plan: Agent.optional(),
00709|         })
00710|         .catchall(Agent)
00711|         .optional()
00712|         .describe("@deprecated Use `agent` field instead."),
00713|       agent: z

(File has more lines. Use 'offset' parameter to read beyond line 713)
</file>

      disabled_providers: z.array(z.string()).optional().describe("Disable providers that are loaded automatically"),
      enabled_providers: z
        .array(z.string())
        .optional()
        .describe("When set, ONLY these providers will be enabled. All other providers will be ignored"),
      model: z.string().describe("Model to use in the format of provider/model, eg anthropic/claude-2").optional(),
      small_model: z
        .string()
        .describe("Small model to use for tasks like title generation in the format of provider/model")
        .optional(),
      default_agent: z
        .string()
        .optional()
        .describe(
          "Default agent to use when none is specified. Must be a primary agent. Falls back to 'build' if not set or if the specified agent is invalid.",
        ),

import { streamText, wrapLanguageModel, type ModelMessage, type StreamTextResult, type Tool, type ToolSet } from "ai"
import { clone, mergeDeep, pipe } from "remeda"
import { ProviderTransform } from "@/provider/transform"
import { Config } from "@/config/config"
import { Instance } from "@/project/instance"
import type { Agent } from "@/agent/agent"
import type { MessageV2 } from "./message-v2"
import { Plugin } from "@/plugin"
import { SystemPrompt } from "./system"
import { ToolRegistry } from "@/tool/registry"
import { Flag } from "@/flag/flag"
--
      model: wrapLanguageModel({
        model: language,
        middleware: [
          {
            async transformParams(args) {
              if (args.type === "stream") {
                // @ts-expect-error
                args.params.prompt = ProviderTransform.message(args.params.prompt, input.model)
              }
              return args.params
            },
          },
        ],

                  const usage = Session.getUsage({
                    usage: value.usage,
                  input.assistantMessage.cost += usage.cost
                  input.assistantMessage.tokens = usage.tokens
                    tokens: usage.tokens,
                    cost: usage.cost,

<file>
00002| import { Log } from "@/util/log"
00003| import { Identifier } from "@/id/id"
00004| import { Session } from "."
00005| import { Agent } from "@/agent/agent"
00006| import { Permission } from "@/permission"
00007| import { Snapshot } from "@/snapshot"
00008| import { SessionSummary } from "./summary"
00009| import { Bus } from "@/bus"
00010| import { SessionRetry } from "./retry"
00011| import { SessionStatus } from "./status"
00012| import { Plugin } from "@/plugin"
00013| import type { Provider } from "@/provider/provider"
00014| import { LLM } from "./llm"
00015| import { Config } from "@/config/config"
00016| 
00017| export namespace SessionProcessor {
00018|   const DOOM_LOOP_THRESHOLD = 3
00019|   const log = Log.create({ service: "session.processor" })
00020| 
00021|   export type Info = Awaited<ReturnType<typeof create>>
00022|   export type Result = Awaited<ReturnType<Info["process"]>>
00023| 
00024|   export function create(input: {
00025|     assistantMessage: MessageV2.Assistant
00026|     sessionID: string
00027|     model: Provider.Model
00028|     abort: AbortSignal
00029|   }) {
00030|     const toolcalls: Record<string, MessageV2.ToolPart> = {}
00031|     let snapshot: string | undefined
00032|     let blocked = false
00033|     let attempt = 0
00034| 
00035|     const result = {
00036|       get message() {
00037|         return input.assistantMessage
00038|       },
00039|       partFromToolCall(toolCallID: string) {
00040|         return toolcalls[toolCallID]
00041|       },
00042|       async process(streamInput: LLM.StreamInput) {
00043|         log.info("process")
00044|         const shouldBreak = (await Config.get()).experimental?.continue_loop_on_deny !== true
00045|         while (true) {
00046|           try {
00047|             let currentText: MessageV2.TextPart | undefined
00048|             let reasoningMap: Record<string, MessageV2.ReasoningPart> = {}
00049|             const stream = await LLM.stream(streamInput)
00050|...

251:                  input.assistantMessage.cost += usage.cost
252:                  input.assistantMessage.tokens = usage.tokens

                    type: "step-start",
                  })
                  break

                case "finish-step":
                  const usage = Session.getUsage({
                    model: input.model,
                    usage: value.usage,
                    metadata: value.providerMetadata,
                  })
                  input.assistantMessage.finish = value.finishReason
                  input.assistantMessage.cost += usage.cost
                  input.assistantMessage.tokens = usage.tokens
                  await Session.updatePart({
                    id: Identifier.ascending("part"),
                    reason: value.finishReason,

30:    const toolcalls: Record<string, MessageV2.ToolPart> = {}
39:      partFromToolCall(toolCallID: string) {
40:        return toolcalls[toolCallID]
99:                case "tool-input-start":
101:                    id: toolcalls[value.id]?.id ?? Identifier.ascending("part"),
104:                    type: "tool",
105:                    tool: value.toolName,
113:                  toolcalls[value.id] = part as MessageV2.ToolPart
116:                case "tool-input-delta":
119:                case "tool-input-end":
122:                case "tool-call": {
123:                  const match = toolcalls[value.toolCallId]
127:                      tool: value.toolName,
137:                    toolcalls[value.toolCallId] = part as MessageV2.ToolPart
146:                          p.type === "tool" &&
147:                          p.tool === value.toolName &&
156:                          pattern: value.toolName,
159:                          callID: value.toolCallId,
160:                          title: `Possible doom loop: "${value.toolName}" called ${DOOM_LOOP_THRESHOLD} times with identical arguments`,
162:                            tool: value.toolName,

      delta,
    })
    return part
  })

  export const getUsage = fn(
    z.object({
      model: z.custom<Provider.Model>(),
      usage: z.custom<LanguageModelUsage>(),
      metadata: z.custom<ProviderMetadata>().optional(),
    }),
    (input) => {
      const cachedInputTokens = input.usage.cachedInputTokens ?? 0
      const excludesCachedTokens = !!(input.metadata?.["anthropic"] || input.metadata?.["bedrock"])
      const adjustedInputTokens = excludesCachedTokens
        ? (input.usage.inputTokens ?? 0)

  export const getUsage = fn(
    z.object({
      model: z.custom<Provider.Model>(),
      usage: z.custom<LanguageModelUsage>(),
      metadata: z.custom<ProviderMetadata>().optional(),
    }),
    (input) => {
      const cachedInputTokens = input.usage.cachedInputTokens ?? 0
      const excludesCachedTokens = !!(input.metadata?.["anthropic"] || input.metadata?.["bedrock"])
      const adjustedInputTokens = excludesCachedTokens
        ? (input.usage.inputTokens ?? 0)
        : (input.usage.inputTokens ?? 0) - cachedInputTokens
      const safe = (value: number) => {
        if (!Number.isFinite(value)) return 0
        return value
      }

      const tokens = {
        input: safe(adjustedInputTokens),
        output: safe(input.usage.outputTokens ?? 0),
        reasoning: safe(input.usage?.reasoningTokens ?? 0),
        cache: {
          write: safe(
            (input.metadata?.["anthropic"]?.["cacheCreationInputTokens"] ??
              // @ts-expect-error
              input.metadata?.["bedrock"]?.["usage"]?.["cacheWriteInputTokens"] ??
              0) as number,
          ),
          read: safe(cachedInputTokens),
        },
      }

      const costInfo =
        input.model.cost?.experimentalOver200K && tokens.input + tokens.cache.read > 200_000
          ? input.model.cost.experimentalOver200K
          : input.model.cost
      return {
        cost: safe(
          new Decimal(0)
            .add(new Decimal(tokens.input).mul(costInfo?.input ?? 0).div(1_000_000))
            .add(new Decimal(tokens.output).mul(costInfo?.output ?? 0).div(1_000_000))
            .add(new Decimal(tokens.cache.read).mul(costInfo?.cache?.read ?? 0).div(1_000_000))
            .add(new Decimal(tokens.cache.write).mul(costInfo?.cache?.write ?? 0).div(1_000_000))
            // TODO: update models.dev to have better pricing model, for now:
            // charge reasoning tokens at the same rate as output tokens
            .add(new Decimal(tokens.reasoning).mul(costInfo?.o...

/home/jm/data/code/opencode/packages/opencode/src/session/index.ts:import { type LanguageModelUsage, type ProviderMetadata } from "ai"

<file>
00001| import z from "zod"
00002| import type { MessageV2 } from "../session/message-v2"
00003| import type { Agent } from "../agent/agent"
00004| 
00005| export namespace Tool {
00006|   interface Metadata {
00007|     [key: string]: any
00008|   }
00009| 
00010|   export interface InitContext {
00011|     agent?: Agent.Info
00012|   }
00013| 
00014|   export type Context<M extends Metadata = Metadata> = {
00015|     sessionID: string
00016|     messageID: string
00017|     agent: string
00018|     abort: AbortSignal
00019|     callID?: string
00020|     extra?: { [key: string]: any }
00021|     metadata(input: { title?: string; metadata?: M }): void
00022|   }
00023|   export interface Info<Parameters extends z.ZodType = z.ZodType, M extends Metadata = Metadata> {
00024|     id: string
00025|     init: (ctx?: InitContext) => Promise<{
00026|       description: string
00027|       parameters: Parameters
00028|       execute(
00029|         args: z.infer<Parameters>,
00030|         ctx: Context,

(File has more lines. Use 'offset' parameter to read beyond line 30)
</file>

-rw-rw-r-- 1 jm jm 253053 Jan 18 00:05 /home/jm/data/code/opencode/session_thinking_text.png

    // Kimi K2 interleaved thinking support (streamed as <think> tags)
    if (model.providerID === "moonshot" || model.api.id.includes("kimi-k2-thinking")) {
      return msgs.map((msg) => {
        if (msg.role === "assistant" && Array.isArray(msg.content)) {
          const reasoningParts = msg.content.filter((part: any) => part.type === "reasoning")
          const reasoningText = reasoningParts.map((part: any) => part.text).join("")

          if (reasoningText) {
            return {
              ...msg,
              providerOptions: {
                ...msg.providerOptions,
                openaiCompatible: {
                  ...(msg.providerOptions as any)?.openaiCompatible,
                  reasoning_content: reasoningText,
                },
              },
            }
          }
        }
        return msg
      })
    }

    return msgs
  }

  function applyCaching(msgs: ModelMessage[], providerID: string): ModelMessage[] {
    const system = msgs.filter((msg) => msg.role === "system").slice(0, 2)
    const final = msgs.filter((msg) => msg.role !== "system").slice(-2)

    const providerOptions = {
      anthropic: {
        cacheControl: { type: "ephemeral" },
      },
      openrouter: {
        cache_control: { type: "ephemeral" },
      },
      bedrock: {
        cachePoint: { type: "ephemeral" },
      },
      openaiCompatible: {
        cache_control: { type: "ephemeral" },
      },
    }

    for (const msg of unique([...system, ...final])) {
      const shouldUseContentOptions = providerID !== "anthropic" && Array.isArray(msg.content) && msg.content.length > 0

      if (shouldUseContentOptions) {
        const lastContent = msg.content[msg.content.length - 1]

    type: z.literal("reasoning"),
      reasoning: z.number(),
      reasoning: z.number(),
            msg.parts.some((part) => part.type !== "step-start" && part.type !== "reasoning")
          if (part.type === "reasoning") {
              type: "reasoning",

        return msg
      })
    }

    // Kimi K2 interleaved thinking support (streamed as <think> tags)
    if (model.providerID === "moonshot" || model.api.id.includes("kimi-k2-thinking")) {
      return msgs.map((msg) => {
        if (msg.role === "assistant" && Array.isArray(msg.content)) {
          const reasoningParts = msg.content.filter((part: any) => part.type === "reasoning")
          const reasoningText = reasoningParts.map((part: any) => part.text).join("")

          if (reasoningText) {
            return {
              ...msg,
              providerOptions: {
                ...msg.providerOptions,
                openaiCompatible: {
                  ...(msg.providerOptions as any)?.openaiCompatible,
                  reasoning_content: reasoningText,
                },
              },
            }
          }
        }
        return msg
      })
    }

    return msgs
  }

  function applyCaching(msgs: ModelMessage[], providerID: string): ModelMessage[] {
    const system = msgs.filter((msg) => msg.role === "system").slice(0, 2)
    const final = msgs.filter((msg) => msg.role !== "system").slice(-2)

    const providerOptions = {
      anthropic: {
        cacheControl: { type: "ephemeral" },
      },
      openrouter: {
        cache_control: { type: "ephemeral" },
      },
      bedrock: {
        cachePoint: { type: "ephemeral" },
      },
      openaiCompatible: {
        cache_control: { type: "ephemeral" },
      },
    }

    for (const msg of unique([...system, ...final])) {
      const shouldUseContentOptions = providerID !== "anthropic" && Array.isArray(msg.content) && msg.content.length > 0

      if (shouldUseContentOptions) {
        const lastContent = msg.content[msg.content.length - 1]
        if (lastContent && typeof lastContent === "object") {
          lastContent.providerOptions = {
            ...lastContent.providerOptions,
            ...providerOptions,
          }
          continue
        }
      }

      msg.prov...

/home/jm/data/code/opencode/.git/logs/refs/heads/astro-hardening
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-CbQHKStN.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-OEl6i-aI.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-dark-high-contrast-E3gJ1_iC.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-light-high-contrast-BfjtVDDH.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/plastic-3e1v2bzS.js
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-CbQHKStN.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/astro-OEl6i-aI.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/github-light-high-contrast-BfjtVDDH.js.map
/home/jm/data/code/opencode/packages/desktop/dist/assets/plastic-3e1v2bzS.js.map

/home/jm/data/code/opencode/packages/opencode/src/tool/registry.ts:import { AstGrepTool } from "./astgrep"
/home/jm/data/code/opencode/packages/opencode/src/tool/registry.ts:      AstGrepTool,
/home/jm/data/code/opencode/packages/opencode/src/tool/astgrep.ts:export const AstGrepTool = Tool.define("astgrep", {

The arguments provided to the tool are invalid: Model tried to call unavailable tool 'ls -la /home/jm/data/code/opencode/packages/opencode/src/tool/</arg_value>'. Available tools: invalid, astgrep, bash, read, glob, grep, edit, write, task, webfetch, todowrite, todoread, websearch, codesearch, skill, lsp, batch, github-triage, cloudflare-docs_search_cloudflare_documentation, cloudflare-docs_migrate_pages_to_workers_guide, cloudflare-workers-bindings_accounts_list, cloudflare-workers-bindings_kv_namespaces_list, cloudflare-workers-bindings_kv_namespace_create, cloudflare-workers-bindings_kv_namespace_delete, cloudflare-workers-bindings_kv_namespace_get, cloudflare-workers-bindings_kv_namespace_update, cloudflare-workers-bindings_workers_list, cloudflare-workers-bindings_workers_get_worker, cloudflare-workers-bindings_workers_get_worker_code, cloudflare-workers-bindings_r2_buckets_list, cloudflare-workers-bindings_r2_bucket_create, cloudflare-workers-bindings_r2_bucket_get, cloudflare-workers-bindings_r2_bucket_delete, cloudflare-workers-bindings_d1_databases_list, cloudflare-workers-bindings_d1_database_create, cloudflare-workers-bindings_d1_database_delete, cloudflare-workers-bindings_d1_database_get, cloudflare-workers-bindings_d1_database_query, cloudflare-workers-bindings_hyperdrive_configs_list, cloudflare-workers-bindings_hyperdrive_config_delete, cloudflare-workers-bindings_hyperdrive_config_get, cloudflare-workers-bindings_hyperdrive_config_edit, cloudflare-workers-bindings_search_cloudflare_documentation, cloudflare-workers-bindings_migrate_pages_to_workers_guide, web-reader_webReader, zread_search_doc, zread_read_file, zread_get_repo_structure, web-search-prime_webSearchPrime.

/home/jm/data/code/opencode/packages/opencode/src/tool/astgrep.ts

<file>
00001| import z from "zod"
00002| import { Tool } from "./tool"
00003| import { Instance } from "../project/instance"
00004| import path from "path"
00005| import fs from "fs/promises"
00006| 
00007| export const AstGrepTool = Tool.define("astgrep", {
00008|   description: "AST search with meta-variables. K2: Use for structural queries",
00009|   parameters: z.object({
00010|     pattern: z.string().describe("AST pattern with meta-variables ($VAR, $$$). Must be complete AST node."),
00011|     lang: z.enum(["typescript", "javascript"]), 
00012|     path: z.string().optional().describe("Directory to search (defaults to current working directory)")
00013|   }),
00014| 
00015|   async execute(params, context) {
00016|     const astGrep = await import("@ast-grep/napi")
00017| 
00018|     const searchPath = params.path || Instance.directory
00019| 
00020|     // Map language strings to ast-grep language modules
00021|     const langMap = {
00022|       typescript: astGrep.ts,
00023|       javascript: astGrep.js
00024|     }
00025| 
00026|     const lang = langMap[params.lang]
00027|     if (!lang) {
00028|       throw new Error(`Unsupported language: ${params.lang}`)
00029|     }
00030| 
00031|     // Find source files
00032|     const files = await findSourceFiles(searchPath, params.lang)
00033|     const results = []
00034| 
00035|     for (const file of files) {
00036|       try {
00037|         const content = await fs.readFile(file, "utf-8")
00038|         const ast = lang.parse(content)
00039|         const root = ast.root()
00040|         const matches = root.findAll(params.pattern)
00041| 
00042|         for (const match of matches) {
00043|           results.push({
00044|             file,
00045|             line: match.range().start.line + 1,
00046|             text: match.text(),
00047|             metaVars: extractMetaVars(match, params.pattern),
00048|           })
00049|         }
00050|       } catch (error) {
00051|         continue
00052|       }
...

[{"name":"__init__.py","path":"packages/kosong/src/kosong/chat_provider/__init__.py","sha":"622564ad88b1fb5d860b80d051664cc83b574a07","size":3695,"url":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/__init__.py?ref=main","html_url":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/__init__.py","git_url":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/622564ad88b1fb5d860b80d051664cc83b574a07","download_url":"https://raw.githubusercontent.com/MoonshotAI/kimi-cli/main/packages/kosong/src/kosong/chat_provider/__init__.py","type":"file","_links":{"self":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/__init__.py?ref=main","git":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/622564ad88b1fb5d860b80d051664cc83b574a07","html":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/__init__.py"}},{"name":"chaos.py","path":"packages/kosong/src/kosong/chat_provider/chaos.py","sha":"0436781a9d01b50dfca7477896d2d72312253cf7","size":9793,"url":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/chaos.py?ref=main","html_url":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/chaos.py","git_url":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/0436781a9d01b50dfca7477896d2d72312253cf7","download_url":"https://raw.githubusercontent.com/MoonshotAI/kimi-cli/main/packages/kosong/src/kosong/chat_provider/chaos.py","type":"file","_links":{"self":"https://api.github.com/repos/MoonshotAI/kimi-cli/contents/packages/kosong/src/kosong/chat_provider/chaos.py?ref=main","git":"https://api.github.com/repos/MoonshotAI/kimi-cli/git/blobs/0436781a9d01b50dfca7477896d2d72312253cf7","html":"https://github.com/MoonshotAI/kimi-cli/blob/main/packages/kosong/src/kosong/chat_provider/chaos.py"}},{"name":"echo....

/home/jm/data/code/opencode node_modules (2367)
├── @actions/artifact@5.0.1
├── @ast-grep/napi@0.40.5
├── @aws-sdk/client-s3@3.933.0
├── @opencode-ai/app@workspace:packages/app
├── @opencode-ai/console-app@workspace:packages/console/app
├── @opencode-ai/console-core@workspace:packages/console/core
├── @opencode-ai/console-function@workspace:packages/console/function
├── @opencode-ai/console-mail@workspace:packages/console/mail
├── @opencode-ai/console-resource@workspace:packages/console/resource
├── @opencode-ai/desktop@workspace:packages/desktop
├── @opencode-ai/enterprise@workspace:packages/enterprise
├── @opencode-ai/plugin@workspace:packages/plugin
├── @opencode-ai/plugin@workspace:packages/plugin
├── @opencode-ai/script@workspace:packages/script
├── @opencode-ai/script@workspace:packages/script
├── @opencode-ai/sdk@workspace:packages/sdk/js
├── @opencode-ai/sdk@workspace:packages/sdk/js
├── @opencode-ai/sessions@workspace:packages/sessions
├── @opencode-ai/slack@workspace:packages/slack

bun test v1.3.6 (d530ed99)

packages/sessions/src/index.test.ts:
 6 | 
 7 | export async function generateUUIDv5(shared_secret: string, name: string) {
 8 | 
 9 |   // 0. Validate shared_secret is a non null UUID string
10 |   if (!isValidUUID(shared_secret)) {
11 |     throw new TypeError("The namespace must be a valid UUID string.")
                   ^
TypeError: The namespace must be a valid UUID string.
      at generateUUIDv5 (/home/jm/data/code/opencode/packages/util/src/secret.ts:11:15)
      at generateUUIDv5 (/home/jm/data/code/opencode/packages/util/src/secret.ts:7:38)
      at <anonymous> (/home/jm/data/code/opencode/packages/sessions/src/index.ts:101:31)
      at <anonymous> (/home/jm/data/code/opencode/packages/sessions/src/index.ts:96:10)
      at dispatch (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/compose.js:22:23)
      at dispatch (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/compose.js:6:32)
      at <anonymous> (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/validator/validator.js:81:18)
      at async dispatch (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/compose.js:22:23)
      at async cors2 (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/middleware/cors/index.js:79:11)
      at async dispatch (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/compose.js:22:23)
      at async <anonymous> (/home/jm/data/code/opencode/node_modules/.bun/hono@4.10.7/node_modules/hono/dist/hono-base.js:201:31)

68 |   SESSIONS_STORE: createMockR2Bucket(),
69 |   SESSIONS_SHARED_SECRET: sharedSecret,
70 |   API_DOMAIN: "test.opencode.ai",
71 | })
72 | 
73 | const parseJson = async <T>(response: Response) => (await response.json()) as T
                                                                        ^
SyntaxError: Failed to parse JSON
      at <anonymou...

/home/jm/data/code/opencode/packages/opencode/package.json:    "@ast-grep/napi": "0.40.5",

<file>
00001| {
00002|   "$schema": "https://json.schemastore.org/package.json",
00003|   "version": "1.0.207",
00004|   "name": "opencode",
00005|   "type": "module",
00006|   "private": true,
00007|   "scripts": {
00008|     "typecheck": "tsgo --noEmit",
00009|     "test": "bun test",
00010|     "build": "bun run script/build.ts",
00011|     "dev": "bun run --conditions=browser ./src/index.ts",
00012|     "random": "echo 'Random script updated at $(date)' && echo 'Change queued successfully' && echo 'Another change made' && echo 'Yet another change' && echo 'One more change' && echo 'Final change' && echo 'Another final change' && echo 'Yet another final change'",
00013|     "clean": "echo 'Cleaning up...' && rm -rf node_modules dist",
00014|     "lint": "echo 'Running lint checks...' && bun test --coverage",
00015|     "format": "echo 'Formatting code...' && bun run --prettier --write src/**/*.ts",
00016|     "docs": "echo 'Generating documentation...' && find src -name '*.ts' -exec echo 'Processing: {}' \\;",
00017|     "deploy": "echo 'Deploying application...' && bun run build && echo 'Deployment completed successfully'"
00018|   },
00019|   "bin": {
00020|     "opencode": "./bin/opencode"
00021|   },
00022|   "exports": {
00023|     "./*": "./src/*.ts"
00024|   },
00025|   "devDependencies": {
00026|     "@babel/core": "7.28.4",
00027|     "@octokit/webhooks-types": "7.6.1",
00028|     "@parcel/watcher-darwin-arm64": "2.5.1",
00029|     "@parcel/watcher-darwin-x64": "2.5.1",
00030|     "@parcel/watcher-linux-arm64-glibc": "2.5.1",
00031|     "@parcel/watcher-linux-arm64-musl": "2.5.1",
00032|     "@parcel/watcher-linux-x64-glibc": "2.5.1",
00033|     "@parcel/watcher-linux-x64-musl": "2.5.1",
00034|     "@parcel/watcher-win32-x64": "2.5.1",
00035|     "@standard-schema/spec": "1.0.0",
00036|     "@tsconfig/bun": "catalog:",
00037|     "@types/babel__core": "7.20.5",
00038|     "@types/bun": "catalog:",
00039|     "@types/turndown": "5.0.5",
00040|     "@types...

/usr/bin/bash: line 1: node: command not found
Node import failed

✓ AstGrepTool imported successfully
ID: {
  id: "astgrep",
  init: [AsyncFunction: init],
}
Parameters: undefined

env: ‘node’: No such file or directory
tsgo not available

Resolving dependencies
Resolved, downloaded and extracted [68]
Saved lockfile
env: ‘node’: No such file or directory
tsx not available

Tool ID: astgrep
Tool description: undefined
Init succeeded
Description: AST search with meta-variables. K2: Use for structural queries
Parameters: _ {
  _zod: {
    traits: Set(5) {
      "ZodObject",
      "$ZodObjectJIT",
      "$ZodObject",
      "$ZodType",
      "ZodType",
    },
    def: {
      type: "object",
      shape: [Getter],
    },
    bag: {},
    version: {
      major: 4,
      minor: 1,
      patch: 8,
    },
    deferred: [
      [Function], [Function]
    ],
    constr: [Function: _],
    propValues: [Getter/Setter],
    parse: [Function],
    run: [Function],
  },
  "~standard": {
    validate: [Function: validate],
    vendor: "zod",
    version: 1,
  },
  def: {
    type: "object",
    shape: [Getter],
  },
  type: "object",
  _def: {
    type: "object",
    shape: [Getter],
  },
  check: [Function],
  clone: [Function],
  brand: [Function],
  register: [Function],
  parse: [Function],
  safeParse: [Function],
  parseAsync: [AsyncFunction],
  safeParseAsync: [AsyncFunction],
  spa: [AsyncFunction],
  encode: [Function],
  decode: [Function],
  encodeAsync: [AsyncFunction],
  decodeAsync: [AsyncFunction],
  safeEncode: [Function],
  safeDecode: [Function],
  safeEncodeAsync: [AsyncFunction],
  safeDecodeAsync: [AsyncFunction],
  refine: [Function],
  superRefine: [Function],
  overwrite: [Function],
  optional: [Function],
  nullable: [Function],
  nullish: [Function],
  nonoptional: [Function],
  array: [Function],
  or: [Function],
  and: [Function],
  transform: [Function],
  default: [Function],
  prefault: [Function],
  catch: [Function],
  pipe: [Function],
  readonly: [Function],
  describe: [Function],
  description: [Getter],
  meta: [Function],
  isOptional: [Function],
  isNullable: [Function],
  shape: [Getter/Setter],
  keyof: [Function],
  catchall: [Function],
  passthrough: [Function],
  loose: [Function],
  strict: [Function],
  strip: [Function],
  extend: [Function],
  safeExtend: [Function],
  merge: [Function],
  pick: [F...

=== AstGrep Tool Details ===
ID: astgrep
Description: undefined
Init successful
Description: AST search with meta-variables. K2: Use for structural queries
Parameters Shape: ✓ Defined

<file>
00001| import { AstGrepTool } from "./astgrep"
00002| import { BashTool } from "./bash"
00003| import { EditTool } from "./edit"
00004| import { GlobTool } from "./glob"
00005| import { GrepTool } from "./grep"
00006| import { ListTool } from "./ls"
00007| import { BatchTool } from "./batch"
00008| import { ReadTool } from "./read"
00009| import { TaskTool } from "./task"
00010| import { TodoWriteTool, TodoReadTool } from "./todo"
00011| import { WebFetchTool } from "./webfetch"
00012| import { WriteTool } from "./write"
00013| import { InvalidTool } from "./invalid"
00014| import { SkillTool } from "./skill"
00015| import type { Agent } from "../agent/agent"
00016| import { Tool } from "./tool"
00017| import { Instance } from "../project/instance"
00018| import { Config } from "../config/config"
00019| import path from "path"
00020| import { type ToolDefinition } from "@opencode-ai/plugin"
00021| import z from "zod"
00022| import { Plugin } from "../plugin"
00023| import { WebSearchTool } from "./websearch"
00024| import { CodeSearchTool } from "./codesearch"
00025| import { Flag } from "@/flag/flag"
00026| import { Log } from "@/util/log"
00027| import { LspTool } from "./lsp"
00028| 
00029| export namespace ToolRegistry {
00030|   const log = Log.create({ service: "tool.registry" })
00031| 
00032|   export const state = Instance.state(async () => {
00033|     const custom = [] as Tool.Info[]
00034|     const glob = new Bun.Glob("tool/*.{js,ts}")
00035| 
00036|     for (const dir of await Config.directories()) {
00037|       for await (const match of glob.scan({
00038|         cwd: dir,
00039|         absolute: true,
00040|         followSymlinks: true,
00041|         dot: true,
00042|       })) {
00043|         const namespace = path.basename(match, path.extname(match))
00044|         const mod = await import(match)
00045|         for (const [id, def] of Object.entries<ToolDefinition>(mod)) {
00046|           custom.push(fromPlugin(id === "default" ? names...

<file>
00111|       ...custom,
00112|     ]
00113|   }
00114| 
00115|   export async function ids() {
00116|     return all().then((x) => x.map((t) => t.id))
00117|   }
00118| 
00119|   export async function tools(providerID: string, agent?: Agent.Info) {
00120|     const tools = await all()
00121|     const result = await Promise.all(
00122|       tools
00123|         .filter((t) => {
00124|           // Enable websearch/codesearch for zen users OR via enable flag
00125|           if (t.id === "codesearch" || t.id === "websearch") {
00126|             return providerID === "opencode" || Flag.OPENCODE_ENABLE_EXA
00127|           }
00128|           return true
00129|         })
00130|         .map(async (t) => {
00131|           using _ = log.time(t.id)
00132|           return {
00133|             id: t.id,
00134|             ...(await t.init({ agent })),
00135|           }
00136|         }),
00137|     )
00138|     return result
00139|   }
00140| 
00141|   export async function enabled(agent: Agent.Info): Promise<Record<string, boolean>> {
00142|     const result: Record<string, boolean> = {}
00143| 
00144|     if (agent.permission.edit === "deny") {
00145|       result["edit"] = false
00146|       result["write"] = false
00147|     }
00148|     if (agent.permission.bash["*"] === "deny" && Object.keys(agent.permission.bash).length === 1) {
00149|       result["bash"] = false
00150|     }
00151|     if (agent.permission.webfetch === "deny") {
00152|       result["webfetch"] = false
00153|       result["codesearch"] = false
00154|       result["websearch"] = false
00155|     }
00156|     // Disable skill tool if all skills are denied
00157|     if (agent.permission.skill["*"] === "deny" && Object.keys(agent.permission.skill).length === 1) {
00158|       result["skill"] = false
00159|     }
00160| 

(File has more lines. Use 'offset' parameter to read beyond line 160)
</file>

        log.info("process")
        const shouldBreak = (await Config.get()).experimental?.continue_loop_on_deny !== true
        while (true) {
          try {
            let currentText: MessageV2.TextPart | undefined
            let reasoningMap: Record<string, MessageV2.ReasoningPart> = {}
            const stream = await LLM.stream(streamInput)

            for await (const value of stream.fullStream) {
              input.abort.throwIfAborted()
              switch (value.type) {
                case "start":
                  SessionStatus.set(input.sessionID, { type: "busy" })
                  break

                case "reasoning-start":
                  if (value.id in reasoningMap) {
                    continue
                  }
                  reasoningMap[value.id] = {
                    id: Identifier.ascending("part"),
                    messageID: input.assistantMessage.id,
                    sessionID: input.assistantMessage.sessionID,
                    type: "reasoning",
                    text: "",
                    time: {
                      start: Date.now(),
                    },
                    metadata: value.providerMetadata,
                  }
                  break

                case "reasoning-delta":
                  if (value.id in reasoningMap) {
                    const part = reasoningMap[value.id]
                    part.text += value.text
                    if (value.providerMetadata) part.metadata = value.providerMetadata
                    if (part.text) await Session.updatePart({ part, delta: value.text })
                  }
                  break

                case "reasoning-end":
                  if (value.id in reasoningMap) {
                    const part = reasoningMap[value.id]
                    part.text = part.text.trimEnd()

                    part.time = {
                      ...part.time,
                      end: Date.now(),
                    }
                    if...

<file>
00049|             const stream = await LLM.stream(streamInput)
00050| 
00051|             for await (const value of stream.fullStream) {
00052|               input.abort.throwIfAborted()
00053|               switch (value.type) {
00054|                 case "start":
00055|                   SessionStatus.set(input.sessionID, { type: "busy" })
00056|                   break
00057| 
00058|                 case "reasoning-start":
00059|                   if (value.id in reasoningMap) {
00060|                     continue
00061|                   }
00062|                   reasoningMap[value.id] = {
00063|                     id: Identifier.ascending("part"),
00064|                     messageID: input.assistantMessage.id,
00065|                     sessionID: input.assistantMessage.sessionID,
00066|                     type: "reasoning",
00067|                     text: "",
00068|                     time: {
00069|                       start: Date.now(),
00070|                     },
00071|                     metadata: value.providerMetadata,
00072|                   }
00073|                   break
00074| 
00075|                 case "reasoning-delta":
00076|                   if (value.id in reasoningMap) {
00077|                     const part = reasoningMap[value.id]
00078|                     part.text += value.text
00079|                     if (value.providerMetadata) part.metadata = value.providerMetadata
00080|                     if (part.text) await Session.updatePart({ part, delta: value.text })
00081|                   }
00082|                   break
00083| 
00084|                 case "reasoning-end":
00085|                   if (value.id in reasoningMap) {
00086|                     const part = reasoningMap[value.id]
00087|                     part.text = part.text.trimEnd()
00088| 
00089|                     part.time = {
00090|                       ...part.time,
00091|                       end: Date.now(),
00092|                     }
000...

/home/jm/data/code/opencode/packages/opencode/src/session/message-v2.ts:              providerMetadata: part.metadata,
/home/jm/data/code/opencode/packages/opencode/src/session/message-v2.ts:              providerMetadata: part.metadata,
/home/jm/data/code/opencode/packages/opencode/src/session/message.ts:      providerMetadata: z.record(z.string(), z.any()).optional(),
/home/jm/data/code/opencode/packages/opencode/src/session/message.ts:      providerMetadata: z.record(z.string(), z.any()).optional(),
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    metadata: value.providerMetadata,
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    if (value.providerMetadata) part.metadata = value.providerMetadata
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    if (value.providerMetadata) part.metadata = value.providerMetadata
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                      metadata: value.providerMetadata,
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    metadata: value.providerMetadata,
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    metadata: value.providerMetadata,
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    if (value.providerMetadata) currentText.metadata = value.providerMetadata
/home/jm/data/code/opencode/packages/opencode/src/session/processor.ts:                    if (value.providerMetadata) currentText.metadata = value.providerMetadata

<file>
00076|                   if (value.id in reasoningMap) {
00077|                     const part = reasoningMap[value.id]
00078|                     part.text += value.text
00079|                     if (value.providerMetadata) part.metadata = value.providerMetadata
00080|                     if (part.text) await Session.updatePart({ part, delta: value.text })
00081|                   }
00082|                   break
00083| 
00084|                 case "reasoning-end":
00085|                   if (value.id in reasoningMap) {
00086|                     const part = reasoningMap[value.id]
00087|                     part.text = part.text.trimEnd()
00088| 
00089|                     part.time = {
00090|                       ...part.time,
00091|                       end: Date.now(),
00092|                     }
00093|                     if (value.providerMetadata) part.metadata = value.providerMetadata
00094|                     await Session.updatePart(part)
00095|                     delete reasoningMap[value.id]
00096|                   }
00097|                   break
00098| 
00099|                 case "tool-input-start":
00100|                   const part = await Session.updatePart({
00101|                     id: toolcalls[value.id]?.id ?? Identifier.ascending("part"),
00102|                     messageID: input.assistantMessage.id,
00103|                     sessionID: input.assistantMessage.sessionID,
00104|                     type: "tool",
00105|                     tool: value.toolName,

(File has more lines. Use 'offset' parameter to read beyond line 105)
</file>

<file>
00002| import { Bus } from "@/bus"
00003| import z from "zod"
00004| import { NamedError } from "@opencode-ai/util/error"
00005| import { Message } from "./message"
00006| import { APICallError, convertToModelMessages, LoadAPIKeyError, type ModelMessage, type UIMessage } from "ai"
00007| import { Identifier } from "../id/id"
00008| import { LSP } from "../lsp"
00009| import { Snapshot } from "@/snapshot"
00010| import { fn } from "@/util/fn"
00011| import { Storage } from "@/storage/storage"
00012| import { ProviderTransform } from "@/provider/transform"
00013| import { STATUS_CODES } from "http"
00014| import { iife } from "@/util/iife"
00015| import { type SystemError } from "bun"
00016| 
00017| export namespace MessageV2 {
00018|   export const OutputLengthError = NamedError.create("MessageOutputLengthError", z.object({}))
00019|   export const AbortedError = NamedError.create("MessageAbortedError", z.object({ message: z.string() }))
00020|   export const AuthError = NamedError.create(
00021|     "ProviderAuthError",
00022|     z.object({
00023|       providerID: z.string(),
00024|       message: z.string(),
00025|     }),
00026|   )
00027|   export const APIError = NamedError.create(
00028|     "APIError",
00029|     z.object({
00030|       message: z.string(),
00031|       statusCode: z.number().optional(),
00032|       isRetryable: z.boolean(),
00033|       responseHeaders: z.record(z.string(), z.string()).optional(),
00034|       responseBody: z.string().optional(),
00035|       metadata: z.record(z.string(), z.string()).optional(),
00036|     }),
00037|   )
00038|   export type APIError = z.infer<typeof APIError.Schema>
00039| 
00040|   const PartBase = z.object({
00041|     id: z.string(),
00042|     sessionID: z.string(),
00043|     messageID: z.string(),
00044|   })
00045| 
00046|   export const SnapshotPart = PartBase.extend({
00047|     type: z.literal("snapshot"),
00048|     snapshot: z.string(),
00049|   }).meta({
00050|     ref: "SnapshotPart"...

80:  export const ReasoningPart = PartBase.extend({
91:  export type ReasoningPart = z.infer<typeof ReasoningPart>

/usr/bin/bash: line 44: node: command not found
Original length: 916
Deduplicated length: 385
Reduction: 531 chars
Result: Let me find the default_agent line around 695: 
Let me find the default_agent line around 695:  
Let me find the default_agent line around 695: 
Let me find the default_agent line around 695:  
Let me find the default_agent line around 695: 
Let me find the default_agent line around 695:  
Let me find the default_agent line around 695: 
Let me find the default_agent line around 695:

84:                case "reasoning-end":

<file>
00083| 
00084|                 case "reasoning-end":
00085|                   if (value.id in reasoningMap) {
00086|                     const part = reasoningMap[value.id]
00087|                     part.text = part.text.trimEnd()
00088| 
00089|                     part.time = {
00090|                       ...part.time,
00091|                       end: Date.now(),
00092|                     }
00093|                     if (value.providerMetadata) part.metadata = value.providerMetadata
00094|                     await Session.updatePart(part)
00095|                     delete reasoningMap[value.id]
00096|                   }
00097|                   break
00098| 
00099|                 case "tool-input-start":
00100|                   const part = await Session.updatePart({
00101|                     id: toolcalls[value.id]?.id ?? Identifier.ascending("part"),
00102|                     messageID: input.assistantMessage.id,

(File has more lines. Use 'offset' parameter to read beyond line 102)
</file>

Original:
Thinking about the problem
Let me analyze this
Let me analyze this
Let me analyze this
Let me analyze this
Moving to next step
Processing data
Processing data
Processing data
Done
Done

Deduplicated:
Thinking about the problem
Let me analyze this
Moving to next step
Processing data
Done

/home/jm/data/code/opencode/packages/opencode/src/config/config.ts:  const INVALID_DIRS = new Bun.Glob(`{${["agents", "commands", "plugins", "tools", "skills"].join(",")}}/`)
/home/jm/data/code/opencode/packages/opencode/src/tool/registry.ts:    // Disable skill tool if all skills are denied

/home/jm/data/code/opencode/packages/opencode/src/tool/skill.ts:        const output = [`## Skill: ${skill.name}`, "", `**Base directory**: ${dir}`, "", parsed.content.trim()].join(

/home/jm/data/code/opencode/packages/opencode/src/skill
/home/jm/data/code/opencode/packages/opencode/test/skill
/home/jm/data/code/opencode/.opencode/skill
/home/jm/data/code/opencode/.opencode/skill/test-skill

drwxrwxr-x 1 jm jm   32 Jan 16 23:23 skill

          custom.push(fromPlugin(id === "default" ? namespace : `${namespace}_${id}`, def))
        custom.push(fromPlugin(id, def))
  function fromPlugin(id: string, def: ToolDefinition): Tool.Info {

<file>
00001| import path from "path"
00002| import z from "zod"
00003| import { Tool } from "./tool"
00004| import { Skill } from "../skill"
00005| import { Agent } from "../agent/agent"
00006| import { Permission } from "../permission"
00007| import { Wildcard } from "../util/wildcard"
00008| import { ConfigMarkdown } from "../config/markdown"
00009| 
00010| const parameters = z.object({
00011|   name: z.string().describe("The skill identifier from available_skills (e.g., 'code-review')"),
00012| })
00013| 
00014| export const SkillTool: Tool.Info<typeof parameters> = {
00015|   id: "skill",
00016|   async init(ctx) {
00017|     const skills = await Skill.all()
00018| 
00019|     // Filter skills by agent permissions if agent provided
00020|     let accessibleSkills = skills
00021|     if (ctx?.agent) {
00022|       const permissions = ctx.agent.permission.skill
00023|       accessibleSkills = skills.filter((skill) => {
00024|         const action = Wildcard.all(skill.name, permissions)
00025|         return action !== "deny"
00026|       })
00027|     }
00028| 
00029|     const description =
00030|       accessibleSkills.length === 0
00031|         ? "Load a skill to get detailed instructions for a specific task. No skills are currently available."
00032|         : [
00033|             "Load a skill to get detailed instructions for a specific task.",
00034|             "Skills provide specialized knowledge and step-by-step guidance.",
00035|             "Use this when a task matches an available skill's description.",
00036|             "<available_skills>",
00037|             ...accessibleSkills.flatMap((skill) => [
00038|               `  <skill>`,
00039|               `    <name>${skill.name}</name>`,
00040|               `    <description>${skill.description}</description>`,
00041|               `  </skill>`,
00042|             ]),
00043|             "</available_skills>",
00044|           ].join(" ")
00045| 
00046|     return {
00047|       description,
0004...

total 8
drwxrwxr-x 1 jm jm   32 Jan 16 23:23 .
drwxrwxr-x 1 jm jm  368 Jan 16 23:23 ..
-rw-rw-r-- 1 jm jm   24 Jan 16 23:23 index.ts
-rw-rw-r-- 1 jm jm 2514 Jan 16 23:23 skill.ts

/home/jm/data/code/opencode/packages/opencode/src/skill/index.ts
/home/jm/data/code/opencode/packages/opencode/src/skill/skill.ts

<file>
00001| import z from "zod"
00002| import { Config } from "../config/config"
00003| import { Instance } from "../project/instance"
00004| import { NamedError } from "@opencode-ai/util/error"
00005| import { ConfigMarkdown } from "../config/markdown"
00006| import { Log } from "../util/log"
00007| 
00008| export namespace Skill {
00009|   const log = Log.create({ service: "skill" })
00010|   export const Info = z.object({
00011|     name: z.string(),
00012|     description: z.string(),
00013|     location: z.string(),
00014|   })
00015|   export type Info = z.infer<typeof Info>
00016| 
00017|   export const InvalidError = NamedError.create(
00018|     "SkillInvalidError",
00019|     z.object({
00020|       path: z.string(),
00021|       message: z.string().optional(),
00022|       issues: z.custom<z.core.$ZodIssue[]>().optional(),
00023|     }),
00024|   )
00025| 
00026|   export const NameMismatchError = NamedError.create(
00027|     "SkillNameMismatchError",
00028|     z.object({
00029|       path: z.string(),
00030|       expected: z.string(),
00031|       actual: z.string(),
00032|     }),
00033|   )
00034| 
00035|   const OPENCODE_SKILL_GLOB = new Bun.Glob("skill/**/SKILL.md")
00036|   const CLAUDE_SKILL_GLOB = new Bun.Glob(".claude/skills/**/SKILL.md")
00037| 
00038|   export const state = Instance.state(async () => {
00039|     const directories = await Config.directories()
00040|     const skills: Record<string, Info> = {}
00041| 
00042|     const addSkill = async (match: string) => {
00043|       const md = await ConfigMarkdown.parse(match)
00044|       if (!md) {
00045|         return
00046|       }
00047| 
00048|       const parsed = Info.pick({ name: true, description: true }).safeParse(md.data)
00049|       if (!parsed.success) return
00050| 
00051|       // Warn on duplicate skill names
00052|       if (skills[parsed.data.name]) {
00053|         log.warn("duplicate skill name", {
00054|           name: parsed.data.name,
00055|           existing: ...

total 0
drwxrwxr-x 1 jm jm 202 Jan 18 03:33 .
drwxrwxr-x 1 jm jm 356 Jan 18 22:28 ..
drwxr-xr-x 1 jm jm  36 Jan 18 03:34 ast-grep
drwxr-xr-x 1 jm jm  16 Jan 17 10:56 debugging-protocol
drwxr-xr-x 1 jm jm  16 Jan 17 10:56 handoff-protocol
drwxr-xr-x 1 jm jm  16 Jan 17 10:56 nucleus-protocol
drwxr-xr-x 1 jm jm  16 Jan 17 10:56 research-protocol
drwxr-xr-x 1 jm jm  16 Jan 17 10:56 review-checklist
drwxrwxr-x 1 jm jm  16 Jan 16 23:23 test-skill

import z from "zod"
import { Config } from "../config/config"
import { Instance } from "../project/instance"
import { NamedError } from "@opencode-ai/util/error"
import { ConfigMarkdown } from "../config/markdown"
import { Log } from "../util/log"

export namespace Skill {
  const log = Log.create({ service: "skill" })
  export const Info = z.object({
    name: z.string(),
    description: z.string(),
    location: z.string(),
  })
  export type Info = z.infer<typeof Info>

  export const InvalidError = NamedError.create(
    "SkillInvalidError",
    z.object({
      path: z.string(),
      message: z.string().optional(),
      issues: z.custom<z.core.$ZodIssue[]>().optional(),
    }),
  )

  export const NameMismatchError = NamedError.create(
    "SkillNameMismatchError",
    z.object({
      path: z.string(),
      expected: z.string(),
      actual: z.string(),
    }),
  )

  const OPENCODE_SKILL_GLOB = new Bun.Glob("skill/**/SKILL.md")
  const CLAUDE_SKILL_GLOB = new Bun.Glob(".claude/skills/**/SKILL.md")

  export const state = Instance.state(async () => {
    const directories = await Config.directories()
    const skills: Record<string, Info> = {}

    const addSkill = async (match: string) => {
      const md = await ConfigMarkdown.parse(match)
      if (!md) {
        return
      }

      const parsed = Info.pick({ name: true, description: true }).safeParse(md.data)
      if (!parsed.success) return

total 4
drwxrwxr-x 1 jm jm   16 Jan 16 23:23 .
drwxrwxr-x 1 jm jm  368 Jan 16 23:23 ..
-rw-rw-r-- 1 jm jm 2929 Jan 16 23:23 index.ts

<file>
00036|     for (const dir of await Config.directories()) {
00037|       for await (const match of glob.scan({
00038|         cwd: dir,
00039|         absolute: true,
00040|         followSymlinks: true,
00041|         dot: true,
00042|       })) {
00043|         const namespace = path.basename(match, path.extname(match))
00044|         const mod = await import(match)
00045|         for (const [id, def] of Object.entries<ToolDefinition>(mod)) {
00046|           custom.push(fromPlugin(id === "default" ? namespace : `${namespace}_${id}`, def))
00047|         }
00048|       }
00049|     }
00050| 
00051|     const plugins = await Plugin.list()
00052|     for (const plugin of plugins) {
00053|       for (const [id, def] of Object.entries(plugin.tool ?? {})) {
00054|         custom.push(fromPlugin(id, def))
00055|       }
00056|     }
00057| 
00058|     return { custom }
00059|   })
00060| 
00061|   function fromPlugin(id: string, def: ToolDefinition): Tool.Info {
00062|     return {
00063|       id,
00064|       init: async () => ({
00065|         parameters: z.object(def.args),
00066|         description: def.description,
00067|         execute: async (args, ctx) => {
00068|           const result = await def.execute(args as any, ctx)
00069|           return {
00070|             title: "",
00071|             output: result,
00072|             metadata: {},
00073|           }
00074|         },
00075|       }),
00076|     }
00077|   }
00078| 
00079|   export async function register(tool: Tool.Info) {
00080|     const { custom } = await state()
00081|     const idx = custom.findIndex((t) => t.id === tool.id)
00082|     if (idx >= 0) {
00083|       custom.splice(idx, 1, tool)
00084|       return
00085|     }
00086|     custom.push(tool)
00087|   }
00088| 
00089|   async function all(): Promise<Tool.Info[]> {
00090|     const custom = await state().then((x) => x.custom)
00091|     const config = await Config.get()
00092| 
00093|     return [
00094|       InvalidTool,
0...

  export const state = Instance.state(async () => {
    const custom = [] as Tool.Info[]
    const glob = new Bun.Glob("tool/*.{js,ts}")

    for (const dir of await Config.directories()) {
      for await (const match of glob.scan({
        cwd: dir,
        absolute: true,
        followSymlinks: true,
        dot: true,
      })) {
        const namespace = path.basename(match, path.extname(match))
        const mod = await import(match)
        for (const [id, def] of Object.entries<ToolDefinition>(mod)) {
          custom.push(fromPlugin(id === "default" ? namespace : `${namespace}_${id}`, def))
        }
      }
    }

    const plugins = await Plugin.list()
    for (const plugin of plugins) {
      for (const [id, def] of Object.entries(plugin.tool ?? {})) {
        custom.push(fromPlugin(id, def))
      }
    }

    return { custom }
  })

  function fromPlugin(id: string, def: ToolDefinition): Tool.Info {
    return {
      id,
      init: async () => ({
        parameters: z.object(def.args),
        description: def.description,
        execute: async (args, ctx) => {
          const result = await def.execute(args as any, ctx)
          return {
            title: "",
            output: result,
            metadata: {},
          }
        },
      }),
    }
  }

  export async function register(tool: Tool.Info) {
    const { custom } = await state()

<file>
00001| import type { Hooks, PluginInput, Plugin as PluginInstance } from "@opencode-ai/plugin"
00002| import { Config } from "../config/config"
00003| import { Bus } from "../bus"
00004| import { Log } from "../util/log"
00005| import { createOpencodeClient } from "@opencode-ai/sdk"
00006| import { Server } from "../server/server"
00007| import { BunProc } from "../bun"
00008| import { Instance } from "../project/instance"
00009| import { Flag } from "../flag/flag"
00010| 
00011| export namespace Plugin {
00012|   const log = Log.create({ service: "plugin" })
00013| 
00014|   const state = Instance.state(async () => {
00015|     const client = createOpencodeClient({
00016|       baseUrl: "http://localhost:4096",
00017|       // @ts-ignore - fetch type incompatibility
00018|       fetch: async (...args) => Server.App().fetch(...args),
00019|     })
00020|     const config = await Config.get()
00021|     const hooks = []
00022|     const input: PluginInput = {
00023|       client,
00024|       project: Instance.project,
00025|       worktree: Instance.worktree,
00026|       directory: Instance.directory,
00027|       $: Bun.$,
00028|     }
00029|     const plugins = [...(config.plugin ?? [])]
00030|     if (!Flag.OPENCODE_DISABLE_DEFAULT_PLUGINS) {
00031|       plugins.push("opencode-copilot-auth@0.0.9")
00032|       plugins.push("opencode-anthropic-auth@0.0.5")
00033|     }
00034|     for (let plugin of plugins) {
00035|       log.info("loading plugin", { path: plugin })
00036|       if (!plugin.startsWith("file://")) {
00037|         const lastAtIndex = plugin.lastIndexOf("@")
00038|         const pkg = lastAtIndex > 0 ? plugin.substring(0, lastAtIndex) : plugin
00039|         const version = lastAtIndex > 0 ? plugin.substring(lastAtIndex + 1) : "latest"
00040|         plugin = await BunProc.install(pkg, version)
00041|       }
00042|       const mod = await import(plugin)
00043|       for (const [_name, fn] of Object.entries<PluginInstance>(mod)) {
00044|   ...

21

0

import { tool, type ToolDefinition } from "@opencode-ai/plugin/tool"
import { z } from "zod"

// Kimi K2 optimized web search with folding
export const websearch_kimi: ToolDefinition = tool({
  description: "Search web with Kimi K2 optimized folding. K2: Plan query sequence, fold after 3 results",
  args: {
    query: z.string().describe("Search query"),
    iteration: z.number().optional().describe("Search iteration number for folding"),
    previousResults: z.number().optional().describe("Number of previous results to retain"),
  },
  execute: async (args, ctx) => {
    // Implementation with automatic folding after 3 results
    return {
      title: "Web Search Results",
      output: results,
      metadata: { totalResults: results.length, folded: previousResultsCount }
    }
  }
})

// Kimi K2 context-aware delegation
export const delegate_kimi: ToolDefinition = tool({
  description: "Delegate to specialist agents with Kimi K2 planning. K2: Use for parallel execution",
  args: {
    agent: z.enum(["oracle", "librarian", "frontend", "backend"]),
    task: z.string().describe("Task description"),
    context: z.string().describe("Context to pass to agent"),
    reasoning: z.string().optional().describe("Kimi's reasoning for this delegation"),
  },
  execute: async (args, ctx) => {
    // Delegate to agent and return result
    return {
      title: `Delegated to ${args.agent}`,
      output: result,
      metadata: { agent: args.agent, task: args.task }
    }
  }
})

export default { websearch_kimi, delegate_kimi }

{
  "builtin_tools": {
    "count": 21,
    "load_time": "0ms (pre-loaded)",
    "type_safety": "Full TypeScript",
    "update_method": "Code changes + release",
    "examples": ["astgrep", "read", "edit", "grep", "bash"]
  },
  "plugin_tools": {
    "load_time": "10-50ms (dynamic import)",
    "type_safety": "Runtime validation",
    "update_method": "File change or plugin update",
    "examples": ["websearch", "codesearch", "screenshot"]
  },
  "kimi_k2_recommendations": {
    "keep_builtin": ["astgrep", "read", "edit", "grep"],
    "extract_to_plugin": ["websearch", "codesearch", "screenshot"],
    "new_as_plugin": ["websearch_kimi", "delegate_kimi", "multihop_search"]
  }
}