EX MCP Server

# Phase 3: Archive-first Cleanup Plan This plan follows the Cleanup Prompt. All removals are staged into `archive/` first, and can be restored. ## Guiding principles - Backwards compatible: no schema changes - Archive-first: do not delete code immediately - Reversible: provide a restore manifest and script - Small, safe batches ## Archive directories - `archive/consensus_legacy/` - `archive/assessment_legacy/` - `archive/schemas_legacy/` - `archive/templates_legacy/` - `archive/config_legacy/` - `archive/error_legacy/` - `archive/tools_legacy/` - `archive/monitoring_legacy/` ## Candidates to archive (first pass) - Consensus legacy duplicates (older variants of consensus schema/templates) - Assessment infrastructure superseded by agentic expert-analysis path - Duplicated schema builders and prompt templates - Redundant error utilities replaced by ResilientErrorHandler - Unused/verbose tool description duplication and dead branches ## Process per batch 1) Dry-run move (list) with cleanup script 2) Archive files into `archive/<category>/` 3) Commit with message `chore(cleanup): archive <batch_name>` 4) Optional: after validation, run delete mode to remove archived files ## Restore process - Use the manifest entries for a batch to restore files back to their original paths ## Next steps - Author the manifest JSON - Implement the cleanup script with dry-run, archive, delete, and restore modes - Prepare the initial batch (consensus legacy and assessment infra) ## Initial batches to populate - consensus_legacy: - assessments/json/*consensus.json error snapshots related to legacy required findings - assessment_legacy: - assessments/json/*.kimi.json and *.glm.json artifacts from earlier analysis runs - schemas_templates: - any duplicated/obsolete schema builders or templates once identified ## Final plan and task tree (derived from 5 docs) - Consensus legacy cleanup - [x] Inventory errors snapshots under assessments/json/*consensus.json - [x] Add to manifest (consensus_legacy) - [ ] Archive and optionally delete originals after backup - Assessment infra cleanup - [x] Inventory *.kimi.json and *.glm.json under assessments/json - [x] Add to manifest (assessment_legacy) - [ ] Archive and optionally delete originals after backup - Schemas/templates cleanup - [ ] Inventory duplicated/obsolete schema builders and templates - [ ] Add to manifest (schemas_templates) - [ ] Archive and optionally delete originals after backup - Error/config legacy cleanup - [ ] Identify and add obsolete error utilities and config entries replaced by flags - [ ] Add to manifest (error_legacy, config_legacy) - [ ] Archive and optionally delete originals after backup - Monitoring/log artifacts cleanup - [x] Inventory log artifacts - [x] Add to manifest (monitoring_legacy) - [ ] Archive and optionally delete originals after backup Notes: - Feature flags remain default OFF; no schema/API changes - Archive-first process, restore supported per batch ## Execution results (current run) - Commands executed - `python scripts/cleanup_phase3.py dry-run` - `python scripts/cleanup_phase3.py archive` - Batch outcomes - consensus_legacy: Present=0, Missing=20 - Rationale: these files had already been archived earlier (now live under `archive/consensus_legacy/`), so the script correctly marked them as missing at their old locations - assessment_legacy: Present=40, Missing=0 → archived to `archive/assessment_legacy/` - schemas_templates: Present=0, Missing=0 (no items populated yet) - monitoring_legacy: Present=8, Missing=0 → archived to `archive/monitoring_legacy/` - Sample archive logs (abbrev.) - `[ARCHIVE 21/40] assessments/json/activity.glm.json -> archive/assessment_legacy/assessments/json/activity.glm.json` - `[ARCHIVE 8/8] logs/ws_shim.log -> archive/monitoring_legacy/logs/ws_shim.log` - Deletion step: not executed yet (archive-only run) ## Current checklist status - Consensus legacy cleanup - [x] Inventory errors snapshots under assessments/json/*consensus.json - [x] Add to manifest (consensus_legacy) - [x] Archived previously (script now reports Missing at old paths) - [ ] Optional: run delete for any lingering originals (none expected) - Assessment infra cleanup - [x] Inventory *.kimi.json and *.glm.json under assessments/json - [x] Add to manifest (assessment_legacy) - [x] Archived to `archive/assessment_legacy/` - [ ] Optional: run `archive --delete` to remove originals - Schemas/templates cleanup - [ ] Inventory duplicated/obsolete schema builders and templates - [ ] Add to manifest (schemas_templates) - [ ] Archive and optionally delete originals after backup - Error/config legacy cleanup - [ ] Identify and add obsolete error utilities and config entries replaced by flags - [ ] Add to manifest (error_legacy, config_legacy) - [ ] Archive and optionally delete originals after backup - Monitoring/log artifacts cleanup - [x] Inventory log artifacts - [x] Add to manifest (monitoring_legacy) - [x] Archived to `archive/monitoring_legacy/` - [ ] Optional: run `archive --delete` to remove originals ## Restore validation plan - Run: `python scripts/cleanup_phase3.py restore --batch consensus_legacy` - Verify files restored to original locations - Re-archive or remove restored samples after validation ## Rollback and notes - Archive-first mapping is reversible per batch - No schema/API changes; all feature flags remain default OFF - The manifest retains consensus_legacy entries to allow restore, even though originals are absent at old paths ## Post-archive deletion (executed) - Deleted originals for archived batches: - assessment_legacy: removed 40 files from assessments/json/*.kimi.json and *.glm.json - monitoring_legacy: removed 8 files from logs/* - consensus_legacy: unchanged (already archived earlier; originals absent at old paths) ## Restore note - Manifest retains entries for consensus_legacy and other batches; use: - `python scripts/cleanup_phase3.py restore --batch consensus_legacy` - After restore validation, re-archive or delete restored samples to maintain consistency ## Monitoring logs note (Windows) - Some logs may be immediately re-created by the running server/shim after deletion (e.g., mcp_activity.log, mcp_server.log, ws_daemon.health.json, ws_shim.log) - To fully remove: stop the processes, then re-run `python scripts/cleanup_phase3.py archive --delete` - The script now summarizes deletions with a "Deferred(Locked)" count to flag these cases