aws-samples / aws-samples/sample-autonomous-cloud-coding-agents
feat(memory): temporal decay for memory entries
- Dominant language
- TypeScript
- Stars
- 143
- Forks
- 46
- Avg merge
- 3d 9h
- Merged PRs (30d)
- 20
Description
**Context:** ROADMAP.md → Memory security → Temporal decay
---
## Component
Agent (Python runtime)
## Describe the feature
Configurable **per-entry TTL** with faster decay for unverified content. Expired entries excluded from retrieval (soft-delete or archive).
## Use case
Repo conventions change; year-old review-derived rules may be wrong. Unverified single-source memories should not persist at full weight indefinitely.
## Proposed solution
1. TTL fields on write; decay curve configurable per source type.
2. Background sweeper or lazy expiry at read time.
3. Operator override to pin high-trust entries (no decay).
4. Metrics: `memory_entries_expired_total`.
## Other information
- Design context: `docs/design/MEMORY.md`, `docs/design/SECURITY.md`.
- [ ] This might be a breaking change
Contributor guide
Research direction
Read docs/design/MEMORY.md and docs/design/SECURITY.md first, then trace the Agent (Python runtime) memory write and retrieval entry points. Define how per-entry TTLs, source-based decay, lazy or background expiry, pinning, and the memory_entries_expired_total metric fit together; done means the design covers each proposed behavior and its breaking-change impact.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai, backend
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100