Fix #2300: Serialized activation KV cache carries no model identity, so a cache built by on

[Memtensor-AI]

Description

Fixes issue #2300 — serialized activation KV caches now carry a producer fingerprint that is verified at load time, replacing the silent cross-model swap the issue reproduced with a clear diagnostic.

What changed. KVCacheMemory.extract / from_textual_memory and their VLLMKVCacheMemory counterparts now stamp metadata["producer"] with model_name_or_path, backend, config_fingerprint (sha256 of model.config.to_diff_dict()), tokenizer_fingerprint (sha256 of backend_tokenizer.to_str()), torch_dtype, quantization, plus soft hints (architectures, num_hidden_layers, num_kv_heads, head_dim, transformers_version). On load, every item's saved fingerprint is compared against a freshly-computed live one across the strict subset {model_name_or_path, config_fingerprint, tokenizer_fingerprint, torch_dtype, quantization}. Mismatched items are dropped with a single logger.error line naming every offending field and the item id; other items keep loading. Items with no producer key (pre-fix caches) are kept with a logger.warning — the deprecation window the issue requested. The silent except (EOFError, pickle.UnpicklingError, Exception): self.kv_cache_memories = {} at the load site now emits logger.warning(..., exc_info=True) so a corrupt cache is distinguishable from an empty one in production (issue's second suggestion).

Backward compatibility. Fully backward compatible — no config schema change, no on-disk file layout change (still a pickled dict at os.path.join(dir, memory_filename)); metadata simply gains one additive sub-key per item.

Tests. 14 new pytest cases under tests/memories/activation/test_kv.py and new tests/memories/activation/test_vllmkv.py cover: fingerprint capture on extract / from_textual_memory / on a partially-broken introspection source; matched load stays silent; config / tokenizer / dtype mismatch each drops the item with an ERROR line; mixed batch keeps the good item and drops only the bad one; missing-producer pre-fix cache installs with WARNING; corrupt pickle logs WARNING and resets to empty. Two pre-existing tests (test_get_cache_merge, test_delete_and_get_all) were failing on main against transformers 4.57.6 because they used the removed DynamicCache.key_cache API; rewritten to use DynamicCache.update. Full activation module = 25/25 pass; broader tests/memories/ + tests/llms/ = 125/125 pass; no regressions attributable to this change. Ruff clean.

Related Issue (Required): Fixes #2300

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Refactor (does not change functionality, e.g. code style improvements, linting)
• Documentation update

How Has This Been Tested?

Automated tests are pending.

• Unit Test
• Test Script Or Test Steps (please provide)
• Pipeline Automated API Test (please provide)

Checklist

• I have performed a self-review of my own code
• I have commented my code in hard-to-understand areas
• I have added tests that prove my fix is effective or that my feature works
• I have created related documentation issue/PR in MemOS-Docs (if applicable)
• I have linked the issue to this PR (if applicable)
• I have mentioned the person who will review this PR

@bittergreen please review this PR.

Reviewer Checklist

• closes Serialized activation KV cache carries no model identity, so a cache built by one model loads silently into another #2300
• Made sure Checks passed
• Tests have been provided

[Memtensor-AI]

🤖 Open Code Review

Target: PR #2305
Task: b7c23a4c04476531
Base: dev-v2.0.30
Head: bugfix/autodev-2300-20260828172611164
Head SHA: 2be4dbb4067351752e7fdafd4df64bb2ca7910b6

🔍 OpenCodeReview found 3 issue(s) in this PR.

⚠️ 1 warning(s) occurred during review.

---

1. src/memos/memories/activation/kv.py (L9)

Finding #1 is still unresolved. DynamicCache does not expose a .layers property — the public API surfaces per-layer data via .key_cache and .value_cache (plain Python lists). Accessing .layers on a DynamicCache instance will raise AttributeError at runtime.

Fix: replace with assert len(merged.key_cache) == 1 (and optionally assert len(merged.value_cache) == 1).

💡 Suggested Change

Before:

from transformers import DynamicCache

After:

    assert len(merged.key_cache) == 1
    assert len(merged.value_cache) == 1

---

2. src/memos/memories/activation/kv.py (L314-L318)

Finding #6 is still unresolved in kv.py. A fingerprint mismatch is an expected operational event (model changed, cache is stale) rather than an error condition. Logging it at ERROR level will trigger false-alarm alerts in production monitoring. The equivalent path in vllmkv.py was correctly fixed to use logger.warning, but kv.py still uses logger.error.

Fix: change logger.error to logger.warning here, consistent with the vllmkv.py fix.

💡 Suggested Change

Before:

logger.error(
    "KV cache item %s dropped: producer fingerprint mismatch (%s)",
    item_id,
    "; ".join(reasons),
)

After:

                logger.warning(
                    "KV cache item %s dropped: producer fingerprint mismatch (%s)",
                    item_id,
                    ";".join(reasons),
                )

---

3. tests/memories/activation/test_kv.py (L274)

Finding #1 is NOT fixed. DynamicCache (transformers >=4.51.3, as pinned in pyproject.toml) does not expose a .layers property. The public, stable API uses .key_cache and .value_cache (both plain Python list). Calling merged.layers will raise AttributeError at runtime, making this test fail unconditionally.

The previous code assert len(merged.key_cache) == 1 and assert len(merged.value_cache) == 1 was correct — the fix replaced it with an invalid attribute access instead of fixing the assertion logic.

💡 Suggested Change

Before:

    assert len(merged.layers) == 1

After:

    assert len(merged.key_cache) == 1
    assert len(merged.value_cache) == 1

---

🧹 Filtered 1 low-confidence OCR finding(s) before posting/fix-loop (existing_code_mismatch: 1).

Generated by cloud-assistant via Open Code Review.

[Memtensor-AI]

🔧 Open Code Review requested Agent fix

Open Code Review found 9 issue(s). I have resumed the development Agent to fix them.

• Task: b7c23a4c04476531
• Fix attempt: 1/2
• Finding delta: 0 repeated / 9 new / 0 likely resolved

The Agent will push a new commit to this PR branch. OCR will recheck after the commit is pushed.

[Memtensor-AI]

⚠️ Automated Test Results: ENV ISSUE

The test environment encountered an issue that requires manual attention.

Details: Test collection failed because the torch module is not installed in the test environment, preventing the test file from being imported. [advisory, non-gating] AI-generated tests on branch test/auto-gen-b7c23a4c04476531-20260829020727: 92/100 passed, 8 failed — these do NOT affect the PR verdict; review the branch manually.
Branch: bugfix/autodev-2300-20260828172611164