Skip to content

mem0

v2.0.12 Security

This release includes 4 security fixes for security teams reviewing exposed deployments.

Published 13d AI Agents & Assistants
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →
This release patches 4 known CVEs

Topics

agents ai ai-agents application chatbots chatgpt
+7 more
genai llm long-term-memory memory memory-management python state-management

Affected surfaces

rce_ssrf

ReleasePort's take

Moderate signal
editorial:auto 12d

The v2.0.12 release patches SQL and Cypher injection flaws in PGVector, Azure MySQL, and Neptune providers; it also validates Elasticsearch filter keys to stop term query injection.

Why it matters: Injection vulnerabilities (severity 90 for vector stores, severity 85 for Elasticsearch) can let attackers execute arbitrary queries. Patching v2.0.12 eliminates these high‑impact risks.

Summary

AI summary

Updates v2.0.12, OSS, and CVE-2026-4372 across a mixed release.

Changes in this release

Security Critical

Fixes SQL and Cypher injection vulnerabilities in PGVector, Azure MySQL, and Neptune providers.

Fixes SQL and Cypher injection vulnerabilities in PGVector, Azure MySQL, and Neptune providers.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Security High

Validates Elasticsearch filter keys and values to prevent term query injection.

Validates Elasticsearch filter keys and values to prevent term query injection.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Feature Low

Accepts `text` argument in `Memory.update()` and `AsyncMemory.update()`, deprecating `data`.

Accepts `text` argument in `Memory.update()` and `AsyncMemory.update()`, deprecating `data`.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Feature Low

Changes Together default LLM model to `MiniMaxAI/MiniMax-M3`.

Changes Together default LLM model to `MiniMaxAI/MiniMax-M3`.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Feature Low

Changes xAI default model to `grok-4.3`.

Changes xAI default model to `grok-4.3`.

Source: granite4.1:30b@2026-07-14-audit

Confidence: low

 —
Feature Low

Updates Together default embedding model to `intfloat/multilingual-e5-large-instruct` at 1024 dimensions, with guidance for existing vectors.

Updates Together default embedding model to `intfloat/multilingual-e5-large-instruct` at 1024 dimensions, with guidance for existing vectors.

Source: granite4.1:30b@2026-07-14-audit

Confidence: low

 —
Feature Low

Changes Cohere default rerank model to `rerank-v3.5`.

Changes Cohere default rerank model to `rerank-v3.5`.

Source: granite4.1:30b@2026-07-14-audit

Confidence: low

 —
Dependency High

Requires transformers>=5.3.0 to remediate GHSA-29pf-2h5f-8g72 (CVE-2026-4372).

Requires transformers>=5.3.0 to remediate GHSA-29pf-2h5f-8g72 (CVE-2026-4372).

Source: llm_adapter@2026-07-14

Confidence: high

 —
Dependency Medium

Makes `langchain-core` optional for default async procedural memory path.

Makes `langchain-core` optional for default async procedural memory path.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Bugfix Medium

Coerces non‑string entity IDs (user_id, agent_id, run_id) instead of crashing on .strip().

Coerces non‑string entity IDs (user_id, agent_id, run_id) instead of crashing on .strip().

Source: llm_adapter@2026-07-14

Confidence: high

 —
Bugfix Medium

Encodes dynamic URL path segments to prevent malformed requests with special characters.

Encodes dynamic URL path segments to prevent malformed requests with special characters.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Bugfix Medium

Skips `temperature` and `top_p` for newer Anthropic models that reject sampling parameters.

Skips `temperature` and `top_p` for newer Anthropic models that reject sampling parameters.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Bugfix Medium

Restores previous payload on Neptune Analytics vector upsert failure to keep payload and embedding in sync.

Restores previous payload on Neptune Analytics vector upsert failure to keep payload and embedding in sync.

Source: llm_adapter@2026-07-14

Confidence: high

 —
Full changelog

Mem0 Python SDK (v2.0.12)

New Features:

  • Memory (OSS): Accept text in Memory.update() and AsyncMemory.update(). data still works but is now deprecated, so prefer text in new code (#6044)

Bug Fixes:

  • Core: Coerce non-string entity IDs (user_id, agent_id, run_id) instead of crashing on .strip(), so passing an integer ID no longer raises AttributeError (#6206)
  • Core: Stop requiring langchain-core for the default async procedural memory path. The optional dependency is now only imported when you pass a custom LangChain LLM, matching the sync behavior (#6209)
  • Client: Encode dynamic URL path segments so IDs containing special characters no longer produce malformed requests (#5963)
  • LLMs: Skip temperature and top_p for newer Anthropic models that reject sampling parameters. Detection is automatic per model family and version, and the new enable_sampling_parameters config flag overrides it (#6211)
  • Vector Stores: Stop writing internal OutputData model fields as properties on Weaviate update() (#6149)
  • Vector Stores: Improve wildcard search handling in Milvus (#6187)
  • Vector Stores: Keep env-resolved Upstash Vector credentials after config validation. An env-var-only config previously passed validation and then failed to build (#5811)
  • Vector Stores: Restore the previous payload when a Neptune Analytics vector upsert fails inside update(), so a partial write can no longer leave the payload and embedding out of sync (#5824)

Changes:

  • LLMs: The Together default model is now MiniMaxAI/MiniMax-M3 (was mistralai/Mixtral-8x7B-Instruct-v0.1) (#6049)
  • LLMs: The xAI default model is now grok-4.3 (was grok-2-latest) (#6115)
  • Embeddings: The Together default embedding model is now intfloat/multilingual-e5-large-instruct at 1024 dimensions (was togethercomputer/m2-bert-80M-8k-retrieval at 768). If you use the Together embedder without pinning model, existing vectors were written at the old dimension: either re-embed them, or pin model and embedding_dims to the old values (#5989)
  • Rerankers: The Cohere default rerank model is now rerank-v3.5 (was rerank-english-v3.0) (#6055)

Security:

  • Vector Stores: Fix SQL and Cypher injection vulnerabilities in the PGVector, Azure MySQL, and Neptune providers (#4878)
  • Vector Stores: Validate Elasticsearch filter keys and values to prevent term query injection (#5980)
  • Dependencies: Require transformers>=5.3.0 to remediate GHSA-29pf-2h5f-8g72 (CVE-2026-4372) (#6110)

Breaking Changes

  • `data` argument in `Memory.update()` and `AsyncMemory.update()` is deprecated; prefer `text`.

Security Fixes

  • CVE-2026-4372 (GHSA-29pf-2h5f-8g72) — Require transformers>=5.3.0 to remediate vulnerability.
  • Vector Stores: Fix SQL injection vulnerabilities in PGVector and Azure MySQL providers.
  • Vector Stores: Fix Cypher injection vulnerabilities in Neptune provider.
  • Vector Stores: Validate Elasticsearch filter keys/values to prevent term query injection.
View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track mem0

Get notified when new releases ship.

Sign up free

About mem0

Universal memory layer for AI Agents

All releases →

Related context

Beta — feedback welcome: [email protected]