3.0.0 (2025-12-23)

Code Refactoring

• remove server-side caps, give AI full limit control (#21) (d0be1b9)

BREAKING CHANGES

• Configuration schema changed from max*/caps to defaults-only

Changes:

• Config schema: maxResults/maxLogLines → defaultLimit/defaultLogLines/
  defaultMetricDataPoints
• Environment variables: MCP_MAX_* → MCP_DEFAULT_* pattern
• All tools: Removed Math.min() caps, AI has full control when specifying
  limits
• Metrics tool: Added pointLimit parameter for AI control over timeseries
  resolution
• Tests: Updated to verify AI control instead of cap enforcement
  (903/903 passing)
• README: Documented limit philosophy - env vars are defaults, NOT caps

The environment variables set what value is used when the AI doesn't
specify a limit. They do NOT cap what the AI can request.

Default values:

• defaultLimit: 50 (general tools)
• defaultLogLines: 200 (logs tool)
• defaultMetricDataPoints: 1000 (metrics timeseries)
• defaultTimeRangeHours: 24 (time range fallback)

• refactor: add input validation to prevent invalid limit values

• Add .min(1) validation to all limit/pageSize/count parameters across 16 tool files
• Prevents invalid values (0 or negative) while maintaining AI full control over maximums
• Extract SAMPLE_DIVERSITY_MULTIPLIER constant in logs.ts for clarity
• Remove 25k token limit documentation (Claude Code CLI specific, not universal MCP)

This improves input validation without restricting AI capabilities.