Pewarta Proxy

📋 Server Request & Proxy Logs

Loading active proxy settings...
Developed by Khairul Adha (adhexa.id)
Mengambil log server...

⚙️ Pengaturan Proxy Target

CodeBurn
Getting StartedGuidesProvidersReferenceCommunity

Optimize

Scan your sessions for waste patterns and get ready-to-paste fixes.

codeburn optimize output

Usage

codeburn optimize                # last 30 days (default)
codeburn optimize -p today       # today only
codeburn optimize -p week        # last 7 days
codeburn optimize --provider claude

Or press o in the dashboard when a finding count appears. Press b to return.

What It Detects

DetectorWhat it catches
Junk readsReads into node_modules, .git, dist
Duplicate readsRe-reads of the same file in a session
MCP coverageMCP servers with many tools but low usage
Unused MCPConfigured but never invoked servers
Bloated CLAUDE.mdOversized context files (with @-import expansion)
Low read:edit ratioEditing without reading leads to retries
Low cache hit rateSessions with a prompt cache hit rate under 80%
Cache bloatHigh cache_creation_input_tokens
Ghost agentsDefined but never-invoked agents
Ghost skillsDefined but never-invoked skills
Ghost commandsDefined but never-invoked slash commands
Bash bloatShell output limit above recommended 15K chars
Low-worth sessionsSessions with cost but no edits or delivery
Context bloatInput:output ratio above 25:1
Session outliersSessions costing 2x+ the project average

Low Cache Hit Rate

A session is flagged when less than 80% of its input came from the prompt cache. The hit rate is cache reads divided by all input: uncached input, cache reads and cache writes. Only sessions with at least 5 calls and 75K input tokens count, and sessions with no cache tokens at all are skipped, because their cache use was not recorded. The finding lists the sessions with the most missed cache reads first. Low rates usually come from switching models or tools mid-session, editing instruction files mid-session, or long pauses that let the cache expire.

Coaching Notes

Alongside the ranked findings, optimize surfaces up to three workflow coaching notes drawn from session history: a high rate of user corrections, an unusually long median time to first edit, and files repeatedly reworked across sessions.

Applying Fixes

Config-class findings (an environment variable, a shell config change, a mv to archive an unused item) can be applied directly instead of copy-pasted:

codeburn optimize --apply              # interactively apply config-class fixes
codeburn optimize --apply --yes        # apply every appliable fix without prompting
codeburn optimize --apply --dry-run    # print the plan without changing anything
codeburn optimize --apply --only id1,id2   # restrict to specific finding ids

Every applied fix is backed up and journaled. Review what was applied with codeburn act list, and roll any of it back with codeburn act undo.

Health Grade

Findings are ranked by urgency (impact weighted against observed waste) and rolled up into an A to F setup health grade. Repeat runs classify each finding as new, improving, or resolved against a 48-hour recent window.

Fix Actions

Each finding includes a ready-to-paste fix:

  • A CLAUDE.md line to add
  • An environment variable to set
  • A mv command to archive unused items
  • A shell config change
ModelsCompare