# opencode-prompt-tracker **Repository Path**: CloudStrolling/opencode-prompt-tracker ## Basic Information - **Project Name**: opencode-prompt-tracker - **Description**: 一个 OpenCode 插件，可自动将每次对话的提示词、模型、Agent 调用链、耗时和 Token ，费用等使用情况记录到每个对话的 Markdown 文件中。 - **Primary Language**: Unknown - **License**: Apache-2.0 - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2026-05-08 - **Last Updated**: 2026-05-15 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # OpenCode Prompt Tracker Plugin An OpenCode plugin that automatically records each conversation's prompt, model, agent call chain, duration, and token usage, including costs, to Markdown files. ## Features - **Automatic Logging**: Records each conversation to Markdown files after completion - **Agent Chain Tracking**: Records the complete agent call chain (e.g., `oracle → build → explore`) - **Token Statistics**: Records Input/Output tokens with cached/uncached breakdown - **Duration Tracking**: Records conversation processing duration - **Step-by-Step Logging**: Logs each assistant message as it completes - **Daily Archiving**: Generates log files organized by date - **Cost Calculation**: Optional billing based on model pricing (when configured) ## Important Note **This project was entirely generated by OpenCode without strict code review or complete testing. At present, basic functionality has been verified to work.** ## Installation ### Method 1: Install via npm (Recommended) Add the plugin configuration in `opencode.json`: ```json { "$schema": "https://opencode.ai/config.json", "plugin": ["opencode-prompt-tracker"] } ``` Restart OpenCode, and the plugin will be automatically downloaded and installed. > **Note**: The plugin is installed to `~/.cache/opencode/node_modules/` directory. ### Method 2: Local Installation (Development) Suitable for modifying the plugin or contributing to development. #### Prerequisites - OpenCode editor - Node.js 18+ or Bun runtime #### Installation Steps ```bash # Clone the plugin repository git clone https://github.com/CloudStrolling/opencode-prompt-tracker.git or git clone https://gitee.com/CloudStrolling/opencode-prompt-tracker.git cd opencode-prompt-tracker # Install dependencies npm install # or bun install # Build the plugin npm run build # or bun run build # Copy to OpenCode plugins directory mkdir -p ~/.opencode/plugins/opencode-prompt-tracker cp -r dist/release/* ~/.opencode/plugins/ ``` ## Usage The project supports zero-config startup - the plugin activates automatically when OpenCode starts. By default, output files are written to `/.opencode/prompts/opencode-prompt-YYYY-MM-DD_.md` You can also create a config file in the project root: `opencode-prompt-tracker.config.json` ## Configuration Create `opencode-prompt-tracker.config.json` in your project root: ```json { "outputPath": "prompts", "filePrefix": "opencode-prompt-tracker", "saveAllLogs": true, "billing": { "enabled": true, "models": [ { "model": "opencode/sonnet-4", "input": 3.75, "output": 15.0, "cacheRead": 0.3, "cacheWrite": 3.75 } ] } } ``` | Field | Default | Description | |-------|---------|-------------| | `outputPath` | `.opencode/prompts` | Output directory relative to project root | | `filePrefix` | `opencode-prompt-` | File name prefix | | `saveAllLogs` | `false` | Save all AI thinking logs - if true, additional `.md` files with `opencode-prompt-log` prefix are created in the output directory to record detailed AI reasoning | | `billing.enabled` | `false` | Enable cost calculation | | `billing.models` | `[]` | Model pricing config (price per 1M tokens) | When billing is enabled and the current model has pricing configured, the summary includes a cost line: ```markdown - **Cost**: $0.0123 (input: $0.005, output: $0.007, cache: $0.0003) ``` Without a config file, the plugin uses all defaults and works as before. ### Log File Format ```markdown # Prompt-Tracker ## Prompt Update detailed design document, including Chinese and English ### Step 1 — 13:24:15 - **Agent**: Sisyphus - Ultraworker - **Model**: opencode/minimax-m2.5-free - **Duration**: 11.16s - **Total Tokens**: 30168 (input: 30001, output: 167) - **Cached Tokens**: 0 (read: 0, write: 0) - **Uncached Tokens**: 30001 - **Task**: Update detailed design document, including Chinese and English ### Step 2 — 13:24:24 - **Agent**: Sisyphus - Ultraworker - **Model**: opencode/minimax-m2.5-free - **Duration**: 9.82s - **Total Tokens**: 3020 (input: 2723, output: 297) - **Cached Tokens**: 29952 (read: 29952, write: 0) - **Uncached Tokens**: 0 - **Task**: I found the current DESIGN.md and DESIGN_CN.md files. Let me read them to understand the current state... ### Step 3 — 13:24:41 - **Agent**: Sisyphus - Ultraworker - **Model**: opencode/minimax-m2.5-free - **Duration**: 16.39s - **Total Tokens**: 66792 (input: 66605, output: 187) - **Cached Tokens**: 0 (read: 0, write: 0) - **Uncached Tokens**: 66605 - **Task**: 1. New `saveAllLogs` feature - need to record complete conversation logs ### Step 4 — 13:24:48 - **Agent**: Sisyphus - Ultraworker - **Model**: opencode/minimax-m2.5-free - **Duration**: 7.31s - **Total Tokens**: 37404 (input: 37295, output: 109) - **Cached Tokens**: 32640 (read: 32640, write: 0) - **Uncached Tokens**: 4655 - **Task**: Let me also check the config.ts and other utility files to have a complete picture --- ### Summary — From 13:24:03 To 13:26:58 - **Model**: opencode/minimax-m2.5-free - **Agent Chain**: Sisyphus - Ultraworker - **Total Duration**: 174.20s - **Steps**: 4 - **Total Tokens**: 137384 (input: 136624, output: 760) - **Cached Tokens**: 62592 (read: 62592, write: 0) - **Uncached Tokens**: 74032 - **Cost**: $0.3078 (input: $0.2776, output: $0.0114, cache: $0.0188) ``` ### Field Description | Field | Description | |-------|-------------| | Time | Conversation start time (HH:MM:SS) | | Agent Chain | Complete agent call chain, e.g., `main → code-review → test` | | Model | AI model used, e.g., `opencode/hy3-preview-free` | | Prompt | User's original input text | | Duration(s) | Conversation processing duration in seconds | | Input Tokens | Total input token count (includes cached portions) | | Output Tokens | Output token count | | Cached Tokens | Total cache read + write tokens | | Uncached Tokens | Input tokens minus cached tokens | | Cache Read | Tokens read from cache (billed at discount) | | Cache Write | Tokens written to cache (billed at premium) | ## Development ### Project Structure ``` opencode-prompt-tracker/ ├── src/ │ ├── index.ts # Main plugin — hook handlers │ ├── types.ts # TypeScript interfaces │ └── utils/ │ ├── file-writer.ts # Writes .md logs to .opencode/prompts/ │ ├── agent-extractor.ts # Parses agent chain from message parts │ ├── logger.ts # Plugin logging to OpenCode console │ ├── config.ts # Config loader │ └── billing.ts # Cost calculation ├── tests/ # Test files ├── dist/ # Build output ├── package.json ├── tsconfig.json ├── README.md # English README ├── README_CN.md # Chinese README ├── AGENTS.md # Developer instructions └── opencode-prompt-tracker.config.example.json ``` ### Developer Commands ```bash npm run build # Build TypeScript + esbuild → dist/release/opencode-prompt-tracker.js npm run dev # Watch mode: tsc --watch npm test # Run tests with Bun (bun test) ``` ### Local Testing ```bash # 1. Build plugin npm run build # 2. Copy to test project mkdir -p /.opencode/plugins/ cp dist/release/opencode-prompt-tracker.js /.opencode/plugins/ # 3. Add to opencode.json { "plugin": ["opencode-prompt-tracker"] } ``` ## Publishing ```bash # Update version (patch/minor/major) npm version patch # Build npm run build # Publish npm publish # For scoped packages: npm publish --access public ``` ## Technical Details - **Hooks**: `chat.message`, `event.message.part.updated`, `event.message.updated`, `event.session.idle` - **Data Storage**: Markdown format with step entries + summary - **Runtime**: Supports both Bun and Node.js - **Type Safety**: TypeScript with strict mode ## License Licensed under the Apache License 2.0. See [LICENSE](./LICENSE) file for details. Copyright 2026 jenemy8023