Scan a project directory to detect languages, frameworks, structure, and dependencies. Returns a project fingerprint used for context-aware coding assistance and skill tracking.

Get git repository context including recent commits, branch info, diff stats, and activity patterns. Powers daily summaries of what the user built.

Run project diagnostics to capture compiler errors, lint warnings, and test failures. Tracks error patterns over time to identify struggle points and learning opportunities.

Read a file's contents with language detection, line numbers, and metadata. Supports reading specific line ranges.

Get aggregated learning progress: error patterns, language activity, skill tree progress, streaks, and personalized recommendations.

Generate an LLM-powered daily coding summary with pet reaction. Aggregates session events, computes stats, and produces a natural language summary.

Real-time diagnostic watcher — monitors your project for errors as you code. Start the watcher, query current errors, or get a full scan.

Actions:

• "start": Begin watching a project directory for file changes

• "stop": Stop the watcher

• "status": Get current watcher state and recent errors

• "scan": Run a one-time full diagnostic scan (no persistent watching)

Works with: TypeScript, Swift, Python, ESLint, and any language with a CLI linter. Platform-agnostic: works with any editor that saves to disk.

Attempt to automatically fix a coding error detected by the diagnostic watcher.

Byte will try a pattern-based fix first (fast, no AI needed). If the error is too complex, Byte returns a structured AI fix request that can be sent to Claude for an intelligent fix.

Modes:

• "suggest": Show the fix without applying it (default, safe)

• "apply": Apply the fix directly to the file (use with care)

• "apply_all": Fix all auto-fixable errors in the project