MemoryTax

I want to evaluate and, if it fits, integrate a technique from recent AI research into my project.

## Technique: MemoryTax
MemoryTax: when is an agent memory layer worth paying for? (Shanghai Jiao Tong)

Tests 7 ways to access a long conversation (full context, BM25, Mem0, Zep, Pi/Codex/Claude Code search) on LoCoMo with 3 models. Weak models gain from memory extraction, strong ones do as well or better with raw history; the model that writes the memory sets the ceiling (Zep 43-45% vs 87% when Opus builds and reads); Opus 5.5 + Pi reached 99% for $6.95 vs 95% for full context at $17.84. Directly actionable agent-memory cost guidance. Caveat: one conversation, one run, GPT-4o-mini judge. Overlaps Tab 38 (Jev Memory Compaction) and Tab 24 (Prune+Spill).

What the reference demo shows: 6-sub-tab demo: findings, writer-vs-reader matrix, harness cost table, break-even calculator, memory decision helper, implementation patterns.

## Reference implementation (Vector Ridge Labs, POC-053)
- Live demo: https://osoman2-gradio-pocs.hf.space/?poc=POC-053
- Source code: https://huggingface.co/spaces/osoman2/Gradio_pocs/tree/main/memorytax
Original source: AlphaSignal Sunday Deep Dive · 2026-10-25

## My project
<Describe your stack, the problem you want to solve and where this technique would sit. Example: "Python FastAPI RAG service over 50k PDFs, Postgres + pgvector, OpenAI embeddings, p95 latency budget 800 ms.">

## What I need from you
1. Read the reference source code and explain in at most 5 bullets how the technique works and what it would change compared to my current approach.
2. Decide whether it fits my case. If it does not, say why and stop there.
3. Propose the smallest viable integration: files or modules to touch, new dependencies, configuration.
4. Implement it behind a flag or as a separate function so it can be compared side by side with what I have today.
5. Add an evaluation script that measures retrieval quality (recall@k / nDCG on a labelled sample of my queries) and latency, before and after, on my own data.
6. List the risks (latency, cost, failure modes) and what to monitor once it is in production.

Constraints: keep my existing stack and conventions, avoid heavy dependencies unless the gain is clear, and reference the file in the demo source whenever you reuse its logic.