// Public Markdown
ByteRover
Manages project knowledge using ByteRover context tree. Provides two operations: query (retrieve knowledge) and curate (store knowledge). Invoke when user requests information lookup, pattern discovery, or knowledge persistence. Developed by ByteRover Inc. (https://byterover.dev/)
MarkdownSecurityShared Jan 27, 2026
Prompt content
# ByteRover Context Tree A project-level knowledge repository that persists across sessions. Use it to avoid re-discovering patterns, conventions, and decisions. ## Why Use ByteRover - **Query before working**: Get existing knowledge about patterns, conventions, and past decisions before implementing - **Curate after learning**: Capture insights, decisions, and bug fixes so future sessions start informed ## Quick Reference | Command | When | Example | |---------|------|---------| | `brv query "question"` | Before starting work | `brv query "How is auth implemented?"` | | `brv curate "context" -f file` | After completing work | `brv curate "JWT 24h expiry" -f auth.ts` | | `brv status` | To check prerequisites | `brv status` | ## When to Use **Query** when you need to understand something: - "How does X work in this codebase?" - "What patterns exist for Y?" - "Are there conventions for Z?" **Curate** when you learned or created something valuable: - Implemented a feature using specific patterns - Fixed a bug and found root cause - Made an architecture decision ## Curate Quality Context must be **specific** and **actionable**: ```bash # Good - specific, explains where and why brv curate "Auth uses JWT 24h expiry, tokens in httpOnly cookies" -f src/auth.ts # Bad - too vague brv curate "Fixed auth" ``` **Note:** Context argument must come before `-f` flags. Max 5 files. ## Best Practices 1. **Break down large contexts** - Run multiple `brv curate` commands for complex topics rather than one massive context. Smaller chunks are easier to retrieve and update. 2. **Let ByteRover read files** - Don't read files yourself before curating. Use `-f` flags to let ByteRover read them directly: ```bash # Good - ByteRover reads the files brv curate "Auth implementation details" -f src/auth.ts -f src/middleware/jwt.ts # Wasteful - reading files twice # [agent reads files] then brv curate "..." -f same-files ``` 3. **Be specific in queries** - Queries block your workflow. Use precise questions to get faster, more relevant results: ```bash # Good - specific brv query "What validation library is used for API request schemas?" # Bad - vague, slow brv query "How is validation done?" ``` 4. **Signal outdated context** - When curating updates that replace existing knowledge, explicitly tell ByteRover to clean up: ```bash brv curate "OUTDATED: Previous auth used sessions. NEW: Now uses JWT with refresh tokens. Clean up old session-based auth context." -f src/auth.ts ``` 5. **Specify structure expectations** - Guide ByteRover on how to organize the knowledge: ```bash # Specify topics/domains brv curate "Create separate topics for: 1) JWT validation, 2) refresh token flow, 3) logout handling" -f src/auth.ts # Specify detail level brv curate "Document the error handling patterns in detail (at least 30 lines covering all error types)" -f src/errors/ ``` ## Prerequisites Run `brv status` first. If errors occur, the agent cannot fix them—instruct the user to take action in their brv terminal. See [TROUBLESHOOTING.md](TROUBLESHOOTING.md) for details. --- **See also:** [WORKFLOWS.md](WORKFLOWS.md) for detailed patterns and examples, [TROUBLESHOOTING.md](TROUBLESHOOTING.md) for error handling
Enhance your coding prompts.
Right where you code.
For clearer instructions, faster output, and better
coding results.
