ByteRover

• Before starting any task — run brv query to surface relevant rules, preferences, and past decisions before your agent touches a single file.
• After implementing a meaningful change — run brv curate to persist new patterns, architectural choices, or decisions so future sessions inherit that knowledge.
• On team projects — use cloud sync (brv push / brv pull) so every agent or developer shares the same curated knowledge base.
• When the agent needs to recall past actions or capabilities — query stored memory instead of relying on context-window history that may have scrolled out of reach.

• The information you need is already present in the current context window — skip brv query to avoid redundant LLM calls.
• The fact is transient, task-specific, or amounts to general world knowledge — don't pollute the knowledge base with ephemeral data using brv curate.

All knowledge is saved as Markdown files inside .brv/context-tree/ in your project directory. Files are fully human-readable, diffable, and can be committed to version control alongside your code.

The built-in ByteRover provider powers brv query and brv curate out of the box — just run brv providers connect byterover. You can also swap in OpenAI, Anthropic, Google, or other providers using your own API key.

brv query synthesizes answers from stored Markdown context using an LLM. brv curate categorizes and structures new knowledge — including content from up to 5 source files — and writes it back to the knowledge base.

Every brv curate run generates a log entry with a unique ID. You can review recent curations, filter by time or status, or inspect full file-operation details for any past entry using brv curate view.

Authenticate with a ByteRover API key to push and pull your knowledge base to a shared cloud space. Teams can organize knowledge into named spaces within teams, and switching spaces automatically pulls the relevant context.

No data is sent to ByteRover servers unless you explicitly run brv push. All query and curate operations stay local — only the configured LLM provider receives context for processing.

Run brv query "What are the authentication patterns in this project?" before the agent writes any code. Previously curated architectural decisions surface immediately, preventing the agent from reinventing or contradicting established patterns.

After migrating from REST to GraphQL, run brv curate "GraphQL is now the API layer; REST endpoints deprecated" -f src/api/schema.ts so every future agent session inherits that context automatically.

Connect a shared ByteRover cloud space, push curated conventions (naming rules, deployment steps, code style decisions), and let every team member's agent pull the same knowledge base before starting work.

Use brv curate view --since 7d --status completed to review a week's worth of curations, verify accuracy, and catch any low-quality entries before they influence future agent behavior.