Skip to content

tasks.md

Latest commit

 

History

History
64 lines (40 loc) · 4.9 KB

tasks.md

File metadata and controls

64 lines (40 loc) · 4.9 KB
Raw

Tasks: Activity-Log Size-Based Retention Cap

Feature: 073-activity-size-retention | Spec: spec.md | Plan: plan.md Approach: Test-driven (tests written before implementation).

Phase 1: Setup

  • T001 Confirm branch 073-activity-size-retention is checked out and go build ./... is green before starting.

Phase 2: Foundational (config field)

  • T002 Add ActivityMaxSizeMB int field with json:"activity_max_size_mb,omitempty" and mapstructure:"activity-max-size-mb" to the config struct in internal/config/config.go, next to the existing Activity* fields.
  • T003 Set the default ActivityMaxSizeMB: 256 in DefaultConfig() in internal/config/config.go (alongside ActivityMaxRecords: 100000).

Phase 3: User Story 1 — Bound activity-log disk use (P1) [MVP]

Goal: A size cap removes oldest records until the log is within budget. Independent test: seed a bucket over budget, prune to size, assert remaining bytes ≤ budget and oldest were removed.

  • T004 [P] [US1] Write failing test TestPruneActivitiesToSize_RemovesOldestUntilUnderBudget in internal/storage/activity_size_test.go: save N activity records with known payload sizes (oldest→newest timestamps), call PruneActivitiesToSize(budget), assert summed remaining bytes ≤ budget and the removed records are the oldest.
  • T005 [US1] Implement func (m *Manager) PruneActivitiesToSize(maxBytes int64) (int, error) in internal/storage/activity.go: in one Update txn, forward-cursor the activity_records bucket computing total len(k)+len(v); if total ≤ maxBytes return 0; else delete oldest-first, subtracting each from total, stopping when total ≤ maxBytes; never delete the final (newest) remaining key. Return count deleted. Guard maxBytes <= 0 → no-op returns (0, nil).
  • T006 [P] [US1] Write test TestPruneActivitiesToSize_AlreadyUnderBudget_NoOp in internal/storage/activity_size_test.go: log under budget → returns 0, all records intact.

Phase 4: User Story 3 — Newest activity always retained (P1)

Goal: deletion never removes the newest record, even if a single record exceeds the budget. Independent test: one record larger than the budget → that newest record survives.

  • T007 [P] [US3] Write test TestPruneActivitiesToSize_KeepsNewestEvenIfOverBudget in internal/storage/activity_size_test.go: seed records where the newest single record exceeds maxBytes; assert the newest record is retained and older ones removed (log not emptied).

Phase 5: User Story 2 — Tune / disable the cap (P2)

Goal: cap of 0 disables size pruning; larger cap retains more. Independent test: PruneActivitiesToSize(0) deletes nothing.

  • T008 [P] [US2] Write test TestPruneActivitiesToSize_DisabledWhenZeroOrNegative in internal/storage/activity_size_test.go: PruneActivitiesToSize(0) and (-1) return (0, nil) with all records intact.

Phase 6: Retention wiring (runtime)

  • T009 [US1] Plumb the size budget into ActivityService: add a maxSizeBytes int64 field set from ActivityMaxSizeMB * 1024 * 1024 where the service is constructed, in internal/runtime/activity_service.go (find the constructor that already reads maxAge/maxRecords).
  • T010 [US1] In runRetentionCleanup (internal/runtime/activity_service.go), after the age and count prunes, call PruneActivitiesToSize(s.maxSizeBytes) when s.maxSizeBytes > 0; on success with deleted > 0 log Info("Pruned activity records to size budget", deleted, max_size_mb); on error log Error.
  • T011 [P] [US1] Write/extend a runtime test in internal/runtime/activity_service_test.go (TestActivityRetention_SizeCapRemovesOldest): construct the service with a small maxSizeBytes, seed an oversized log, run runRetentionCleanup, assert the log is within budget and newest records remain.

Phase 7: Polish & Docs

  • T012 [P] Document activity_max_size_mb (default 256, 0 = disabled, runs with age/count caps) in docs/configuration.md and the activity/retention docs.
  • T013 Run the full local verification: go test ./internal/storage/ ./internal/runtime/ ./internal/config/ and go build ./cmd/mcpproxy; ensure green.
  • T014 Run gofmt/goimports and ./scripts/run-linter.sh on changed files.

Dependencies & Order

  • Phase 2 (config field) blocks T009 (wiring reads the field).
  • T005 (implementation) unblocks all storage tests passing; tests T004/T006/T007/T008 are written first (TDD) and fail until T005.
  • Phase 6 (wiring) depends on T005 + Phase 2.
  • T013/T014 last.

Parallel Opportunities

  • T004, T006, T007, T008 are all in the same new test file — write together, but they only pass after T005.
  • T012 (docs) is independent and can be done anytime after the config field exists.

MVP

User Story 1 (T002–T006 + T009–T011) delivers the core: a default-on size cap that bounds the activity log. US2/US3 are small additions to the same method.