docs(changelog): backfill Python v2.0.18 and TypeScript v3.1.6 SDK entries (#6918 )

docs: redirect deprecated OSS graph memory page to platform graph memory (#6914 )
fix(memory): escape delimiters when building the session scope key (#6892 )
2026-08-12 00:25:18 +05:30 · 2026-08-12 00:25:02 +05:30 · 2026-08-11 23:23:55 +05:30 · 2026-08-11 20:15:07 +05:30 · 2026-08-07 19:04:33 +05:30 · 2026-08-07 18:16:06 +05:30
52 changed files with 2574 additions and 347 deletions
@@ -95,6 +95,12 @@ client.project.update(
    custom_instructions="..."
 )

+# Separate extraction instructions for agent-scoped memories
+# (see /platform/features/custom-instructions)
+client.project.update(
+    agent_custom_instructions="..."
+)
+
 # Use the input language for memory storage and retrieval
 client.project.update(multilingual=True)

@@ -7,6 +7,23 @@ mode: "wide"
 <Tabs>
 <Tab title="Python">

+<Update label="2026-08-11" description="v2.0.18">
+
+**Bug Fixes:**
+- **Core:** Percent-escape `%`, `&`, and `=` in `user_id`, `agent_id`, and `run_id` when building the session scope key for the recent-conversation buffer, so the key stays unambiguous for ids containing those characters. Ordinary ids keep their existing key; an id already containing `%`, `&`, or `=` maps to a new key, so its buffer starts empty once and refills on the next `add()`. Stored memories are unaffected ([#6892](https://github.com/mem0ai/mem0/pull/6892))
+- **Vector Stores:** Raise `ValueError` when a PGVector `in`/`nin` filter value is not a list. A string value was previously iterated character by character into the generated `= ANY(...)` array (so `{"user_id": {"in": "alice"}}` matched `a`, `l`, `i`, `c`, `e`), and a non-iterable value raised a bare `TypeError` from deep inside filter building ([#6879](https://github.com/mem0ai/mem0/pull/6879))
+- **Vector Stores:** Reject `index_accuracy=0` in the Oracle AI Vector Search config. The range check sat behind a truthiness test, so `0` skipped validation entirely and was passed through to `WITH TARGET ACCURACY 0` instead of raising ([#6848](https://github.com/mem0ai/mem0/pull/6848))
+- **Vector Stores:** Close the Oracle connection or pool that Mem0 opened when initialization fails. A client version check, a database version check, or a `create_col()` error previously propagated with the connection still open, leaking it for the life of the process. A caller-supplied `client` is left untouched ([#6839](https://github.com/mem0ai/mem0/pull/6839))
+
+</Update>
+
+<Update label="2026-08-05" description="v2.0.17">
+
+**New Features:**
+- **Client:** Add `agent_custom_instructions` to `project.update()`/`update_project()` (sync and async) and to the `ProjectUpdateOptions` and `AddMemoryOptions` typed models. It sets a second extraction instruction set that applies only to agent-scoped memories: an add passing `agent_id` without `user_id` uses it, one passing both splits by attribution, and while it is unset `custom_instructions` continues to apply to every memory ([#6809](https://github.com/mem0ai/mem0/pull/6809))
+
+</Update>
+
 <Update label="2026-08-04" description="v2.0.16">

 **New Features:**
@@ -1189,6 +1206,31 @@ See the [OSS v2 to v3 migration guide](https://docs.mem0.ai/migration/oss-v2-to-

 <Tab title="TypeScript">

+<Update label="2026-08-11" description="v3.1.6">
+
+**New Features:**
+- **Vector Stores:** Add an Oracle AI Vector Search vector store (`oracledb`) to the OSS SDK, with pooled connections, `HNSW`/`IVF` indexes, an optional `indexAccuracy` target, JSON payload filtering, and six selectable distance metrics ([#6690](https://github.com/mem0ai/mem0/pull/6690))
+
+**Bug Fixes:**
+- **Core:** Percent-escape `%`, `&`, and `=` in `userId`, `agentId`, and `runId` when building the session scope key for the recent-conversation buffer, so the key stays unambiguous for ids containing those characters. Ordinary ids keep their existing key; an id already containing `%`, `&`, or `=` maps to a new key, so its buffer starts empty once and refills on the next `add()`. Stored memories are unaffected ([#6892](https://github.com/mem0ai/mem0/pull/6892))
+- **Vector Stores:** Close and clear the Oracle pool that Mem0 created when `initialize()` fails, and reset the cached init promise so the next call retries instead of replaying the rejection forever. A caller-supplied `client` is left untouched ([#6839](https://github.com/mem0ai/mem0/pull/6839))
+- **Vector Stores:** Validate Oracle `insert()` batches before touching the database: `ids` and `payloads` must match `vectors` in length, so a short array can no longer write rows with `undefined` ids or silently drop payloads ([#6839](https://github.com/mem0ai/mem0/pull/6839))
+
+**Improvements:**
+- **Vector Stores:** Cut Oracle round trips. `insert()` now sends the whole batch through one `executeMany()` with explicit `bindDefs` instead of one `INSERT` per vector, `list()` reads the total from a `COUNT(*) OVER ()` window in the same statement instead of issuing a second count query, and an unfiltered `search()` adds the `VECTOR_INDEX_TRANSFORM` hint so the vector index is used ([#6835](https://github.com/mem0ai/mem0/pull/6835))
+
+**Security:**
+- **Dependencies:** Patched 8 high and 18 medium severity dependency vulnerabilities across the pnpm workspace via `pnpm.overrides` (`undici`, `brace-expansion`, `ip-address`) ([#6847](https://github.com/mem0ai/mem0/pull/6847))
+
+</Update>
+
+<Update label="2026-08-05" description="v3.1.5">
+
+**New Features:**
+- **Client:** Add `agentCustomInstructions` to `PromptUpdatePayload`, `AddMemoryOptions`, and `ProjectResponse`. It sets a second extraction instruction set that applies only to agent-scoped memories: an add passing `agentId` without `userId` uses it, one passing both splits by attribution, and while it is unset `customInstructions` continues to apply to every memory ([#6809](https://github.com/mem0ai/mem0/pull/6809))
+
+</Update>
+
 <Update label="2026-08-04" description="v3.1.4">

 **New Features:**
@@ -2801,6 +2843,14 @@ Existing memories written by the previous versions are not rewritten. If your me

 <Tab title="n8n">

+<Update label="2026-08-05" description="n8n-nodes-mem0 v0.1.3">
+
+**Changes:**
+- **License changed to MIT:** The published `@mem0/n8n-nodes-mem0` package is now MIT (was Apache-2.0). n8n's Creator Portal requires verified community nodes to be MIT, and the failing license check was the blocker for verification. The rest of the mem0 repo stays Apache-2.0 ([#6804](https://github.com/mem0ai/mem0/pull/6804))
+- **Themed icons:** The node and credential icons now declare `{ light, dark }` variants instead of a single icon, clearing the remaining `icon-prefer-themed-variants` warnings from the Creator Portal scan. No functional changes ([#6804](https://github.com/mem0ai/mem0/pull/6804))
+
+</Update>
+
 <Update label="2026-08-04" description="n8n-nodes-mem0 v0.1.2">

 **Changes:**
@@ -3,17 +3,25 @@ title: "Oracle AI Vector Search"
 description: "Use Oracle Database AI Vector Search as a vector store in Mem0 for semantic and relational queries."
 ---

+{/* Copyright (c) 2026, Oracle and/or its affiliates. */}
+
 [Oracle AI Vector Search](https://www.oracle.com/database/ai-vector-search/) stores embeddings in an Oracle table using the native `VECTOR` data type, so you can combine semantic search over unstructured data with relational queries over business data in a single database.

 ### Requirements

 - Oracle Database 23.4 or later, with a user that can create tables and vector indexes
- The `python-oracledb` driver. In thick mode, Oracle Client 23.4 or later is also required.
+- The `python-oracledb` or `node-oracledb` driver. In thick mode, Oracle Client 23.4 or later is also required.

-```bash
+<CodeGroup>
+```bash Python
 pip install oracledb
 ```

+```bash TypeScript
+npm install oracledb
+```
+</CodeGroup>
+
 ### Usage

 <CodeGroup>
@@ -47,11 +55,58 @@ messages = [
 ]
 m.add(messages, user_id="alice", metadata={"category": "movies"})
 ```
+
+```typescript TypeScript
+import { Memory } from "mem0ai/oss";
+
+const config = {
+  vectorStore: {
+    provider: "oracledb",
+    config: {
+      collectionName: "mem0",
+      embeddingModelDims: 1536,
+      connectionParams: {
+        user: "mem0_user",
+        password: "your-password",
+        connectString: "localhost:1521/FREEPDB1",
+      },
+    },
+  },
+};
+
+const memory = new Memory(config);
+
+const messages = [
+  {
+    role: "user",
+    content: "I'm planning to watch a movie tonight. Any recommendations?",
+  },
+  {
+    role: "assistant",
+    content: "How about thriller movies? They can be quite engaging.",
+  },
+  {
+    role: "user",
+    content: "I'm not a big fan of thriller movies but I love sci-fi movies.",
+  },
+  {
+    role: "assistant",
+    content:
+      "Got it! I'll avoid thriller recommendations and suggest sci-fi movies in the future.",
+  },
+];
+
+await memory.add(messages, {
+  userId: "alice",
+  metadata: { category: "movies" },
+});
+```
 </CodeGroup>

-To reuse a connection or pool you already manage, pass it as `client` instead of `connection_params`:
+To reuse a connection or pool you already manage, pass it as `client` instead of the connection parameters:

-```python
+<CodeGroup>
+```python Python
 import oracledb

 pool = oracledb.create_pool(user="mem0_user", password="your-password", dsn="localhost:1521/FREEPDB1")
@@ -64,33 +119,52 @@ config = {
 }
 ```

+```typescript TypeScript
+import oracledb from "oracledb";
+
+const pool = await oracledb.createPool({
+  user: "mem0_user",
+  password: "your-password",
+  connectString: "localhost:1521/FREEPDB1",
+});
+
+const config = {
+  vectorStore: {
+    provider: "oracledb",
+    config: { client: pool },
+  },
+};
+```
+</CodeGroup>
+
 ### Config

 Here are the parameters available for configuring Oracle AI Vector Search:

-| Parameter | Description | Default Value |
-| --- | --- | --- |
-| `connection_params` | Connection settings passed to `python-oracledb`, such as `user`, `password` and `dsn`. See the [connection handling guide](https://python-oracledb.readthedocs.io/en/latest/user_guide/connection_handling.html). | `None` |
-| `use_connection_pool` | Create a connection pool from `connection_params` instead of a single connection | `True` |
-| `client` | An existing `oracledb.Connection` or `oracledb.ConnectionPool` to use instead of building one from `connection_params` | `None` |
-| `collection_name` | Name of the Oracle table that stores vectors and payloads | `mem0` |
-| `embedding_model_dims` | Dimension of your embedding vectors, must be greater than 0 | `1536` |
-| `distance_metric` | Distance function used for indexing and search: `COSINE`, `EUCLIDEAN`, `EUCLIDEAN_SQUARED`, `DOT`, `HAMMING` or `MANHATTAN` | `COSINE` |
-| `do_create_index` | Whether to create a vector index on the collection | `True` |
-| `index_type` | Vector index type: `HNSW` or `IVF` | `HNSW` |
-| `index_name` | Name of the vector index | `<collection_name>_VEC_IDX` |
-| `index_parameters` | Index tuning parameters. For `HNSW`: `neighbors`, `efconstruction`. For `IVF`: `neighbor partitions`, `samples_per_partition`, `min_vectors_per_partition`. | `None` |
-| `index_accuracy` | Target index accuracy from 1 to 100, applied as `WITH TARGET ACCURACY <n>` | `None` |
+| Python | TypeScript | Description | Default Value |
+| --- | --- | --- | --- |
+| `connection_params` | `connectionParams` | Connection settings passed to the Oracle driver, such as `user`, `password` and `dsn` (`connectString` in TypeScript). See the [Python](https://python-oracledb.readthedocs.io/en/latest/user_guide/connection_handling.html) or [Node.js](https://node-oracledb.readthedocs.io/en/latest/user_guide/connection_handling.html) connection handling guide. | `None` |
+| `use_connection_pool` | `useConnectionPool` | Create a connection pool from the connection parameters instead of a single connection | `True` |
+| `client` | `client` | An existing Oracle connection or pool to use instead of building one from the connection parameters | `None` |
+| `collection_name` | `collectionName` | Name of the Oracle table that stores vectors and payloads | `mem0` |
+| `embedding_model_dims` | `embeddingModelDims` | Dimension of your embedding vectors, must be greater than 0 | `1536` |
+| `distance_metric` | `distanceMetric` | Distance function used for indexing and search: `COSINE`, `EUCLIDEAN`, `EUCLIDEAN_SQUARED`, `DOT`, `HAMMING` or `MANHATTAN` | `COSINE` |
+| `do_create_index` | `doCreateIndex` | Whether to create a vector index on the collection | `True` |
+| `index_type` | `indexType` | Vector index type: `HNSW` or `IVF` | `HNSW` |
+| `index_name` | `indexName` | Name of the vector index | `<collection_name>_VEC_IDX` |
+| `index_parameters` | `indexParameters` | Index tuning parameters. For `HNSW`: `neighbors`, `efconstruction`. For `IVF`: `neighbor partitions`, `samples_per_partition`, `min_vectors_per_partition`. | `None` |
+| `index_accuracy` | `indexAccuracy` | Target index accuracy from 1 to 100, applied as `WITH TARGET ACCURACY <n>` | `None` |

 <Note>
-  When you pass a pre-built `client`, Mem0 uses it as-is and ignores `connection_params` and `use_connection_pool`. Mem0 does not close a client it did not create.
+  When you pass a pre-built `client`, Mem0 uses it as-is and ignores the connection parameters and pooling options. Mem0 does not close a client it did not create.
 </Note>

 ### Vector indexes

 Set the index type with `index_type` and tune it with `index_parameters`:

-```python
+<CodeGroup>
+```python Python
 config = {
    "vector_store": {
        "provider": "oracledb",
@@ -104,6 +178,25 @@ config = {
 }
 ```

+```typescript TypeScript
+const config = {
+  vectorStore: {
+    provider: "oracledb",
+    config: {
+      connectionParams: {
+        user: "mem0_user",
+        password: "your-password",
+        connectString: "localhost:1521/FREEPDB1",
+      },
+      indexType: "HNSW",
+      indexParameters: { neighbors: 32, efconstruction: 200 },
+      indexAccuracy: 95,
+    },
+  },
+};
+```
+</CodeGroup>
+
 For the full list of supported options, see the Oracle [`CREATE VECTOR INDEX`](https://docs.oracle.com/en/database/oracle/oracle-database/26/sqlrf/create-vector-index.html) reference.

 ### Search scores
@@ -121,14 +214,23 @@ Filters run against the JSON `payload` column and support:
 | Comparison | `{"score": {"gte": 0.5}}`, also `eq`, `ne`, `gt`, `lt`, `lte` |
 | Membership | `{"category": {"in": ["movies", "books"]}}`, also `nin` |
 | String matching | `{"title": {"contains": "sci-fi"}}`, also `icontains` for case-insensitive |
-| Logical groups | `{"AND": [...]}`, `{"OR": [...]}`, `{"NOT": [...]}` |
+| Logical groups | `{"AND": [...]}`, `{"OR": [...]}`, `{"NOT": [...]}`, also `$and`, `$or`, `$not` |

 Multiple fields at the top level are combined with `AND`:

-```python
+<CodeGroup>
+```python Python
 m.search(
    "movie recommendations",
    user_id="alice",
    filters={"category": {"in": ["movies", "books"]}, "rating": {"gte": 4}},
 )
 ```
+
+```typescript TypeScript
+await memory.search("movie recommendations", {
+  userId: "alice",
+  filters: { category: { in: ["movies", "books"] }, rating: { gte: 4 } },
+});
+```
+</CodeGroup>
@@ -10,7 +10,7 @@ Mem0 includes built-in support for various popular databases. Memory can utilize
 See the list of supported vector databases below.

 <Note>
-  The following vector databases are supported in the Python implementation. The TypeScript implementation currently supports Qdrant, Redis, PGVector, Supabase, LangChain, Azure AI Search, Vectorize, Amazon S3 Vectors, Milvus, Neptune Analytics, and an in-memory store.
+  The following vector databases are supported in the Python implementation. The TypeScript implementation currently supports Qdrant, Redis, PGVector, Supabase, LangChain, Oracle AI Vector Search, Azure AI Search, Vectorize, Amazon S3 Vectors, Milvus, Neptune Analytics, and an in-memory store.
 </Note>

 <CardGroup cols={3}>
@@ -46,7 +46,7 @@ When new messages arrive, Mem0 extracts durable facts and stores them with the i
 1. **Context lookup.** Mem0 checks related existing memories so it can avoid storing the same fact again.
 2. **Fact extraction.** An LLM extracts preferences, decisions, plans, and other details your agent can reuse.
 3. **Deduplication and embedding.** Redundant facts are removed, then each memory is embedded for semantic search.
-4. **Entity linking.** When configured, Mem0 links people, places, organizations, and concepts across memories.
+4. **Entity extraction.** Mem0 pulls out the people, places, organizations, and concepts each memory mentions and stores them for entity matching at search time. On Platform these entities also become the nodes of [Graph Memory](/platform/features/graph-memory).

 The automatic extraction path is additive. If a user says, "I moved from Austin to Seattle," Mem0 can store the new fact without silently rewriting the old one. Use explicit `update` or `delete` operations when your application needs to correct or remove a memory.

@@ -61,7 +61,7 @@ When you call `search`, Mem0 ranks stored memories against your query and filter
 | **Entity** | Boosts memories linked to entities in the query | Questions about a person, project, or account |
 | **Temporal** | Scores candidates on time metadata extracted at write time against the query's temporal intent | Temporal questions ("when did...", current state, recency) |

-Platform retrieval fuses these signals in the managed service. OSS retrieval depends on your configured vector store, optional reranker, and graph store.
+Platform retrieval fuses these signals in the managed service, where the entity signal is powered by built-in [Graph Memory](/platform/features/graph-memory). OSS retrieval depends on your configured vector store and optional reranker, and boosts on entity overlap alone: it has no graph memory.

 <Note>
  Always scope searches with filters such as `user_id`, `agent_id`, or `run_id`. This keeps memories from different users, agents, or sessions from mixing.
@@ -75,7 +75,7 @@ Mem0 stores different parts of a memory in stores built for different lookup pat
 |---|---|---|
 | **SQL database** | Facts and metadata | The source of truth for each memory |
 | **Vector database** | Embeddings | Semantic similarity search |
-| **Entity or graph store** | Entities and relationships | Relationship-aware retrieval when graph memory is enabled |
+| **Entity store** | Entities extracted from memory text | Boosts memories sharing entities with the query. On Platform it also backs [Graph Memory](/platform/features/graph-memory) |

 On Mem0 Platform, these stores are managed for you. In OSS, you choose and operate the backing stores through your configuration.

@@ -1035,6 +1035,10 @@
      "source": "/features/graph-memory",
      "destination": "/platform/features/graph-memory"
    },
+    {
+      "source": "/open-source/features/graph-memory",
+      "destination": "/platform/features/graph-memory"
+    },
    {
      "source": "/features/:slug",
      "destination": "/platform/features/:slug"
@@ -177,14 +177,14 @@ If the user is on a pre-current major (Python < 2, TS < 3, or Platform `output_f
 - [Platform CLI](https://docs.mem0.ai/platform/cli) [Platform]: Use when the user wants to manage Platform memories from the terminal.
 - [Mem0 MCP Server](https://docs.mem0.ai/platform/mem0-mcp) [Platform]: Use when connecting memory to AI coding tools over MCP.
 - [Open Source Overview](https://docs.mem0.ai/open-source/overview) [OSS]: Use when the user needs full infra control and custom provider wiring.
- [Open Source Configuration](https://docs.mem0.ai/open-source/configuration) [OSS]: Use when configuring `Memory` - LLM, embedder, vector store, graph store.
+- [Open Source Configuration](https://docs.mem0.ai/open-source/configuration) [OSS]: Use when configuring `Memory` - LLM, embedder, vector store, reranker.
 - [Open Source Python Quickstart](https://docs.mem0.ai/open-source/python-quickstart) [OSS]: Use for the first self-hosted Python integration.
 - [Open Source Node.js Quickstart](https://docs.mem0.ai/open-source/node-quickstart) [OSS]: Use for the first self-hosted Node integration.
 - [Self-Hosted Setup](https://docs.mem0.ai/open-source/setup) [OSS]: Use when standing up the bundled REST server and dashboard via Docker Compose, including auth, API keys, and the setup wizard.

 ## Core Concepts

- [How Mem0 Works](https://docs.mem0.ai/core-concepts/how-it-works) [Both]: Use when explaining the end-to-end pipeline: extraction (ADD-only distillation), storage across vector/graph/history stores, and multi-signal retrieval.
+- [How Mem0 Works](https://docs.mem0.ai/core-concepts/how-it-works) [Both]: Use when explaining the end-to-end pipeline: extraction (ADD-only distillation), storage across vector/entity/history stores, and multi-signal retrieval.
 - [Memory Types](https://docs.mem0.ai/core-concepts/memory-types) [Both]: Use when explaining working, factual, episodic, and semantic memory distinctions.
 - [Memory Operations - Add](https://docs.mem0.ai/core-concepts/memory-operations/add) [Both]: Use when explaining how `add()` extracts facts, resolves conflicts, and writes to both stores.
 - [Memory Operations - Search](https://docs.mem0.ai/core-concepts/memory-operations/search) [Both]: Use when explaining how queries are processed and ranked.
@@ -1,6 +1,6 @@
 ---
 title: "Open Source: Migrating to the New Memory Algorithm"
-description: "Guide for self-hosted Mem0 users to upgrade to the new memory algorithm with ADD-only extraction, hybrid search, and entity linking."
+description: "Guide for self-hosted Mem0 users to upgrade to the new memory algorithm with ADD-only extraction, hybrid search, and entity-aware retrieval."
 icon: "arrow-right"
 iconType: "solid"
 ---
@@ -15,7 +15,8 @@ The new Mem0 release redesigns both extraction and retrieval, and cleans up the

 - **Extraction**: Single-pass ADD-only (one LLM call, no UPDATE/DELETE)
 - **Retrieval**: Multi-signal hybrid search (semantic + BM25 keyword + entity matching)
- **Entity linking**: Automatic entity extraction and cross-memory linking
+- **Entity matching**: Automatic entity extraction feeds a third scoring signal in hybrid search, boosting memories that share entities with the query
+- **Graph memory moved to Platform**: The external graph store integration is removed from OSS; graph memory is now a built-in, always-on [Mem0 Platform feature](/platform/features/graph-memory)
 - **SDK cleanup**: Deprecated parameters removed, naming conventions standardized
 - **API surface aligned with Platform**: Entity IDs now follow the same convention across OSS and Platform: top-level kwargs for `add()` / `delete_all()`, inside `filters` for `search()` / `get_all()`

@@ -37,7 +38,7 @@ These changes produce a **+20 point improvement on LoCoMo** (71.4 → 91.6) and
 | `add()` events | Returns `ADD`, `UPDATE`, `DELETE` | Returns `ADD` only | Update code expecting UPDATE/DELETE |
 | Custom extraction prompt | `custom_fact_extraction_prompt` | `custom_instructions` | Rename in config |
 | Custom update prompt | `custom_update_memory_prompt` | Deprecated | Use `custom_instructions` instead |
-| Graph memory | `enable_graph` + `graph_store` in config | Removed | Graph store support has been removed entirely |
+| Graph memory | `enable_graph` + `graph_store` in config | Removed | Graph memory is removed from OSS. It's a built-in, always-on [Mem0 Platform feature](/platform/features/graph-memory) |
 | Qdrant client | `>=1.9.1` | `>=1.12.0` | Update dependency |
 | Upstash client | `>=0.1.0` | `>=0.6.0` | Update dependency |

@@ -53,8 +54,8 @@ These changes produce a **+20 point improvement on LoCoMo** (71.4 → 91.6) and
 | `messages` in `add()` | Could be `null` / `undefined` | Required: throws on null/undefined | Always pass a string or array |
 | Payload key for lemmatized text | `text_lemmatized` (snake_case) | `textLemmatized` (camelCase) | TS-only internal field. If you share a vector store collection between Python and TS SDKs, lemma-based BM25 will not resolve across languages: keep collections language-scoped. |
 | Custom prompt | `customPrompt` | `customInstructions` | Rename in config |
-| Graph memory | `enableGraph` + `graphStore` in config | Removed | Graph store support has been removed entirely |
-| Default graph config | Neo4j default config applied | No default graph config | Graph store config is no longer used |
+| Graph memory | `enableGraph` + `graphStore` in config | Removed | Graph memory is removed from OSS. It's a built-in, always-on [Mem0 Platform feature](/platform/features/graph-memory) |
+| Default graph config | Neo4j default config applied | No default graph config | Graph store config is no longer read; OSS has no built-in graph config to fall back to |

 ### Python Client SDK

@@ -104,7 +105,7 @@ These changes produce a **+20 point improvement on LoCoMo** (71.4 → 91.6) and
 </Tabs>

 <Info>
-The Python `[nlp]` extra installs [spaCy](https://spacy.io/) for entity extraction and keyword lemmatization. Without it, Mem0 still works but falls back to semantic-only search (no entity linking, no BM25 lemmatization).
+The Python `[nlp]` extra installs [spaCy](https://spacy.io/) for entity extraction and keyword lemmatization. Without it, Mem0 still works but falls back to semantic-only search (no entity matching, no BM25 lemmatization).
 </Info>

 <Warning>
@@ -135,7 +136,7 @@ pip install fastembed
    config = {
        "custom_instructions": "Focus on user preferences",  # [OK] New name
        # custom_update_memory_prompt removed: use custom_instructions
-        # enable_graph and graph_store removed: graph store support has been removed
+        # enable_graph and graph_store removed: graph memory is now a Mem0 Platform feature
    }
    ```
  </Tab>
@@ -154,7 +155,7 @@ pip install fastembed
    // After
    const config = {
      customInstructions: "Focus on user preferences",  // [OK] New name
-      // enableGraph and graphStore removed: graph store support has been removed
+      // enableGraph and graphStore removed: graph memory is now a Mem0 Platform feature
    };
    ```
  </Tab>
@@ -320,35 +321,31 @@ pip install "qdrant-client>=1.12.0"
 pip install "upstash-vector>=0.6.0"
 ```

-### 6. Entity Store Setup
+### 6. Entity Matching Store Setup

-The new algorithm automatically creates a parallel entity store collection named `{your_collection}_entities`. No manual setup is required: it's created on first use.
+The new algorithm automatically creates a parallel collection named `{your_collection}_entities` to power entity matching, the third signal in hybrid search. No manual setup is required: it's created on first use. This is separate from and unrelated to graph memory, which is a Mem0 Platform feature.

 <Warning>
 Make sure your vector store user/credentials have permission to create new collections. If you're using a managed vector database with restricted permissions, pre-create the `{collection_name}_entities` collection with the same embedding dimensions as your main collection.
 </Warning>

-## Graph Memory: Now Built-In
+## Graph Memory: Platform Only

-External graph **store** support has been removed from the open-source SDK and replaced by **built-in graph memory** (entity linking), which runs natively with no external dependencies.
+Graph memory is removed from the open-source SDK. It is not being replaced by an OSS equivalent: graph memory is a **Mem0 Platform** feature, built in and always on, with no external graph database required. See [Graph Memory](/platform/features/graph-memory) for what it does on Platform.

-**What was removed:**
+**What was removed from OSS:**
 - `enable_graph` / `enableGraph` config flag
- `graph_store` / `graphStore` configuration block (Neo4j, Memgraph, Kuzu, Apache AGE, Neptune)
- All external graph store code paths (~4000 lines)
-
-**What replaces it:**
-
-Mem0 now builds the graph itself. It extracts entities (proper nouns, quoted text, compound noun phrases) from every memory during the add pipeline and stores them in a parallel collection (`{collection}_entities`) inside your existing vector store. Memories that share an entity are linked, and at search time entities from the query are matched against this collection to boost connected memories. The boost is folded into the combined `score` on each result.
+- `graph_store` / `graphStore` configuration block
+- All external graph store drivers (Neo4j, Memgraph, Kuzu, Apache AGE, Neptune) and their code paths (~4000 lines)

 **Migration:**
 - Remove `enable_graph` / `enableGraph` from your config
 - Remove the `graph_store` / `graphStore` block: it is no longer read
 - Uninstall external graph drivers (neo4j, memgraph, etc.) if you were using them only for Mem0
- No data migration is required. Built-in graph memory activates automatically on the next `add()` call.
+- If you need graph memory, use [Mem0 Platform](/platform/features/graph-memory) instead of self-hosted OSS

 <Warning>
-The old `relations` field on search results (populated by the external graph store) is no longer returned. Entity connections are now applied through retrieval ranking rather than exposed as a separate, directly traversable structure. If your application read or traversed the `relations` array, you will need to redesign that part against the new API.
+The old `relations` field on search results (populated by the external graph store) is no longer returned in OSS. OSS has no graph memory replacement, so there is nothing to populate this field with. If your application read or traversed the `relations` array, either move to Mem0 Platform to keep that data or redesign that part against the new OSS retrieval API.
 </Warning>

 ## How the New Algorithm Works
@@ -362,7 +359,7 @@ Input conversation
    → Batch embed extracted memories
    → Hash-based deduplication (MD5, prevents exact duplicates)
    → Batch insert into vector store
-    → Entity extraction + linking
+    → Entity extraction (for entity matching)
 ```

 The previous algorithm used two LLM calls: one to extract candidate facts, one to decide ADD/UPDATE/DELETE actions against existing memories. The new algorithm collapses this into a single call that only adds. The model spends its capacity on understanding the input rather than diffing against existing state.
@@ -375,7 +372,7 @@ Query
    → Parallel scoring:
        1. Semantic search (vector similarity)
        2. BM25 keyword search (normalized term matching)
-        3. Entity matching (entity graph boost)
+        3. Entity matching (entity overlap boost)
    → Score fusion → Top-K selection
 ```

@@ -408,8 +405,8 @@ The new features degrade gracefully when optional dependencies are missing:
 | Missing Dependency | Impact | Search Still Works? |
 |---|---|---|
 | spaCy (`mem0ai[nlp]`) | No entity extraction, no BM25 lemmatization | Yes (semantic-only) |
-| `fastembed` (Qdrant) | No BM25 keyword search | Yes (semantic + entity) |
-| Entity store unavailable | No entity boosting | Yes (semantic + BM25) |
+| `fastembed` (Qdrant) | No BM25 keyword search | Yes (semantic + entity matching) |
+| Entity matching store unavailable | No entity matching boost | Yes (semantic + BM25) |

 You always get semantic search. Hybrid search features layer on top when available.

@@ -449,13 +446,13 @@ These parameters have been removed across all SDKs. Remove them from your code:

 **Config:** `custom_update_memory_prompt` → deprecated, use `custom_instructions`

-**Config:** `enable_graph` + `graph_store` → removed (graph store support removed entirely)
+**Config:** `enable_graph` + `graph_store` → removed (graph memory is now a [Mem0 Platform feature](/platform/features/graph-memory))

 ### TypeScript OSS: Removed/renamed parameters

 **Config:** `customPrompt` → renamed to `customInstructions`

-**Config:** `enableGraph` + `graphStore` → removed (graph store support removed entirely)
+**Config:** `enableGraph` + `graphStore` → removed (graph memory is now a [Mem0 Platform feature](/platform/features/graph-memory))

 **search():** `limit` → renamed to `topK`

@@ -521,9 +518,9 @@ If spaCy is not installed at all, install the NLP extras:
 pip install "mem0ai[nlp]"
 ```

-### Entity store collection creation fails
+### Entity matching store collection creation fails

-The entity store tries to create a `{collection_name}_entities` collection automatically. If your vector database has restricted permissions, pre-create this collection with the same embedding dimensions as your main collection.
+The entity matching store tries to create a `{collection_name}_entities` collection automatically. If your vector database has restricted permissions, pre-create this collection with the same embedding dimensions as your main collection.

 ### Score values are different from before

@@ -4,7 +4,7 @@ description: "Configure Mem0 OSS in Python or TypeScript with your own LLM, embe
 icon: "sliders"
 ---

-Mem0 OSS works out of the box with OpenAI defaults. Point it at your own LLM, embedder, and vector store by passing a config when you create `Memory`. The Python SDK also supports a reranker and graph memory.
+Mem0 OSS works out of the box with OpenAI defaults. Point it at your own LLM, embedder, vector store, and reranker by passing a config when you create `Memory`.

 <Info>
  **Prerequisites**
@@ -90,11 +90,11 @@ Set your provider keys as environment variables:

 ```bash
 export OPENAI_API_KEY="..."
-export COHERE_API_KEY="..."   # Python reranker only
+export COHERE_API_KEY="..."   # Cohere reranker only
 ```

 <Note>
-  The TypeScript OSS SDK configures the LLM, embedder, vector store, and history store. Reranker and graph memory are Python-only today.
+  The TypeScript OSS SDK configures the LLM, embedder, vector store, history store, and reranker. Graph memory is not part of OSS in either language: it is a built-in [Mem0 Platform feature](/platform/features/graph-memory).
 </Note>

 Prefer a config file? Load YAML into Python's `from_config`:
@@ -2784,6 +2784,10 @@
 										"type": "string",
 										"description": "Project-level instructions that guide extraction for this call."
 									},
+									"agent_custom_instructions": {
+										"type": "string",
+										"description": "Extraction instructions for agent-scoped memories, overriding the project-level setting for this call. Applied when `agent_id` is sent without `user_id`; when both are sent it governs the assistant-attributed memories while `custom_instructions` governs the rest."
+									},
 									"custom_categories": {
 										"type": "array",
 										"description": "Category catalog for this call. Replaces the project-level list rather than merging with it. Omit to fall back to the project list, then the default catalog.",
@@ -5805,6 +5809,11 @@
 										},
 										"description": "Custom instructions for memory processing in this project"
 									},
+									"agent_custom_instructions": {
+										"type": "string",
+										"nullable": true,
+										"description": "Extraction instructions for agent-scoped memories. Falls back to `custom_instructions` when unset. Send an empty string to clear it."
+									},
 									"custom_categories": {
 										"type": "array",
 										"items": {
@@ -7431,6 +7440,12 @@
 						"type": "string",
 						"nullable": true
 					},
+					"agent_custom_instructions": {
+						"description": "Extraction instructions that apply only to agent-scoped memories. Used when `agent_id` is sent without `user_id`; when both are sent it governs the assistant-attributed memories while `custom_instructions` governs the rest. Falls back to `custom_instructions` when unset.",
+						"title": "Agent custom instructions",
+						"type": "string",
+						"nullable": true
+					},
 					"immutable": {
 						"description": "Whether the memory is immutable.",
 						"title": "Immutable",
@@ -95,6 +95,100 @@ Exclude:
 - [Irrelevant information]
 ```

+## Agent Custom Instructions
+
+`custom_instructions` applies to every memory your project extracts, no matter whose it is. But what is worth remembering about an agent is rarely what is worth remembering about a user: an agent's useful memories are things like which tools fail, which retry strategies work, and how a given environment behaves, not personal preferences.
+
+`agent_custom_instructions` is an optional second set of extraction rules that applies only to agent-scoped memories.
+
+Available from Python SDK `v2.0.17` and TypeScript SDK `v3.1.5`. Upgrade first if you are on an earlier release.
+
+Set it on the project alongside `custom_instructions`:
+
+<CodeGroup>
+```python Python
+client.project.update(
+    custom_instructions="Extract the user's preferences, goals, and constraints.",
+    agent_custom_instructions=(
+        "Extract operational lessons for the agent:\n"
+        "- Tools that failed and the error returned\n"
+        "- Retry or fallback strategies that worked\n"
+        "- Environment quirks worth recalling on the next run\n\n"
+        "Exclude: user preferences, personal details."
+    ),
+)
+```
+
+```javascript JavaScript
+await client.updateProject({
+  customInstructions: "Extract the user's preferences, goals, and constraints.",
+  agentCustomInstructions: `Extract operational lessons for the agent:
+- Tools that failed and the error returned
+- Retry or fallback strategies that worked
+- Environment quirks worth recalling on the next run
+
+Exclude: user preferences, personal details.`,
+});
+```
+</CodeGroup>
+
+### Which instructions apply
+
+Which set governs an `add()` call depends on the entity IDs you pass with it:
+
+| The add call passes | Instructions applied |
+|---------------------|----------------------|
+| `user_id` only | `custom_instructions` |
+| `agent_id` only | `agent_custom_instructions` |
+| `user_id` **and** `agent_id` | `agent_custom_instructions` govern the memories attributed to the assistant; `custom_instructions` govern the rest |
+
+`agent_custom_instructions` is unset by default. While it is unset, `custom_instructions` applies to every memory, so projects that don't set it behave exactly as they did before.
+
+### Overriding for a single call
+
+Both fields are also accepted per request, overriding the project setting for that `add()` only:
+
+<CodeGroup>
+```python Python
+client.add(
+    messages,
+    filters={"agent_id": "support-agent"},
+    agent_custom_instructions="Only remember which tools errored and why.",
+)
+```
+
+```javascript JavaScript
+await client.add(messages, {
+  agentId: "support-agent",
+  agentCustomInstructions: "Only remember which tools errored and why.",
+});
+```
+</CodeGroup>
+
+### Reading and clearing
+
+<CodeGroup>
+```python Python
+# Read the current value
+response = client.project.get(fields=["agent_custom_instructions"])
+print(response["agent_custom_instructions"])
+
+# Clear it; agent memories fall back to custom_instructions
+client.project.update(agent_custom_instructions="")
+```
+
+```javascript JavaScript
+// Read the current value
+const response = await client.getProject({
+  fields: ["agentCustomInstructions"],
+});
+console.log(response.agentCustomInstructions);
+
+// Clear it; agent memories fall back to customInstructions
+await client.updateProject({ agentCustomInstructions: "" });
+```
+</CodeGroup>
+
 ## Real-World Examples

 <Tabs>
@@ -1,201 +1,21 @@
-                                 Apache License
-                           Version 2.0, January 2004
-                        http://www.apache.org/licenses/
+MIT License

-   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+Copyright (c) 2023-2026 Taranjeet Singh

-   1. Definitions.
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:

-      "License" shall mean the terms and conditions for use, reproduction,
-      and distribution as defined by Sections 1 through 9 of this document.
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.

-      "Licensor" shall mean the copyright owner or entity authorized by
-      the copyright owner that is granting the License.
-
-      "Legal Entity" shall mean the union of the acting entity and all
-      other entities that control, are controlled by, or are under common
-      control with that entity. For the purposes of this definition,
-      "control" means (i) the power, direct or indirect, to cause the
-      direction or management of such entity, whether by contract or
-      otherwise, or (ii) ownership of fifty percent (50%) or more of the
-      outstanding shares, or (iii) beneficial ownership of such entity.
-
-      "You" (or "Your") shall mean an individual or Legal Entity
-      exercising permissions granted by this License.
-
-      "Source" form shall mean the preferred form for making modifications,
-      including but not limited to software source code, documentation
-      source, and configuration files.
-
-      "Object" form shall mean any form resulting from mechanical
-      transformation or translation of a Source form, including but
-      not limited to compiled object code, generated documentation,
-      and conversions to other media types.
-
-      "Work" shall mean the work of authorship, whether in Source or
-      Object form, made available under the License, as indicated by a
-      copyright notice that is included in or attached to the work
-      (an example is provided in the Appendix below).
-
-      "Derivative Works" shall mean any work, whether in Source or Object
-      form, that is based on (or derived from) the Work and for which the
-      editorial revisions, annotations, elaborations, or other modifications
-      represent, as a whole, an original work of authorship. For the purposes
-      of this License, Derivative Works shall not include works that remain
-      separable from, or merely link (or bind by name) to the interfaces of,
-      the Work and Derivative Works thereof.
-
-      "Contribution" shall mean any work of authorship, including
-      the original version of the Work and any modifications or additions
-      to that Work or Derivative Works thereof, that is intentionally
-      submitted to Licensor for inclusion in the Work by the copyright owner
-      or by an individual or Legal Entity authorized to submit on behalf of
-      the copyright owner. For the purposes of this definition, "submitted"
-      means any form of electronic, verbal, or written communication sent
-      to the Licensor or its representatives, including but not limited to
-      communication on electronic mailing lists, source code control systems,
-      and issue tracking systems that are managed by, or on behalf of, the
-      Licensor for the purpose of discussing and improving the Work, but
-      excluding communication that is conspicuously marked or otherwise
-      designated in writing by the copyright owner as "Not a Contribution."
-
-      "Contributor" shall mean Licensor and any individual or Legal Entity
-      on behalf of whom a Contribution has been received by Licensor and
-      subsequently incorporated within the Work.
-
-   2. Grant of Copyright License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      copyright license to reproduce, prepare Derivative Works of,
-      publicly display, publicly perform, sublicense, and distribute the
-      Work and such Derivative Works in Source or Object form.
-
-   3. Grant of Patent License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      (except as stated in this section) patent license to make, have made,
-      use, offer to sell, sell, import, and otherwise transfer the Work,
-      where such license applies only to those patent claims licensable
-      by such Contributor that are necessarily infringed by their
-      Contribution(s) alone or by combination of their Contribution(s)
-      with the Work to which such Contribution(s) was submitted. If You
-      institute patent litigation against any entity (including a
-      cross-claim or counterclaim in a lawsuit) alleging that the Work
-      or a Contribution incorporated within the Work constitutes direct
-      or contributory patent infringement, then any patent licenses
-      granted to You under this License for that Work shall terminate
-      as of the date such litigation is filed.
-
-   4. Redistribution. You may reproduce and distribute copies of the
-      Work or Derivative Works thereof in any medium, with or without
-      modifications, and in Source or Object form, provided that You
-      meet the following conditions:
-
-      (a) You must give any other recipients of the Work or
-          Derivative Works a copy of this License; and
-
-      (b) You must cause any modified files to carry prominent notices
-          stating that You changed the files; and
-
-      (c) You must retain, in the Source form of any Derivative Works
-          that You distribute, all copyright, patent, trademark, and
-          attribution notices from the Source form of the Work,
-          excluding those notices that do not pertain to any part of
-          the Derivative Works; and
-
-      (d) If the Work includes a "NOTICE" text file as part of its
-          distribution, then any Derivative Works that You distribute must
-          include a readable copy of the attribution notices contained
-          within such NOTICE file, excluding those notices that do not
-          pertain to any part of the Derivative Works, in at least one
-          of the following places: within a NOTICE text file distributed
-          as part of the Derivative Works; within the Source form or
-          documentation, if provided along with the Derivative Works; or,
-          within a display generated by the Derivative Works, if and
-          wherever such third-party notices normally appear. The contents
-          of the NOTICE file are for informational purposes only and
-          do not modify the License. You may add Your own attribution
-          notices within Derivative Works that You distribute, alongside
-          or as an addendum to the NOTICE text from the Work, provided
-          that such additional attribution notices cannot be construed
-          as modifying the License.
-
-      You may add Your own copyright statement to Your modifications and
-      may provide additional or different license terms and conditions
-      for use, reproduction, or distribution of Your modifications, or
-      for any such Derivative Works as a whole, provided Your use,
-      reproduction, and distribution of the Work otherwise complies with
-      the conditions stated in this License.
-
-   5. Submission of Contributions. Unless You explicitly state otherwise,
-      any Contribution intentionally submitted for inclusion in the Work
-      by You to the Licensor shall be under the terms and conditions of
-      this License, without any additional terms or conditions.
-      Notwithstanding the above, nothing herein shall supersede or modify
-      the terms of any separate license agreement you may have executed
-      with Licensor regarding such Contributions.
-
-   6. Trademarks. This License does not grant permission to use the trade
-      names, trademarks, service marks, or product names of the Licensor,
-      except as required for reasonable and customary use in describing the
-      origin of the Work and reproducing the content of the NOTICE file.
-
-   7. Disclaimer of Warranty. Unless required by applicable law or
-      agreed to in writing, Licensor provides the Work (and each
-      Contributor provides its Contributions) on an "AS IS" BASIS,
-      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
-      implied, including, without limitation, any warranties or conditions
-      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
-      PARTICULAR PURPOSE. You are solely responsible for determining the
-      appropriateness of using or redistributing the Work and assume any
-      risks associated with Your exercise of permissions under this License.
-
-   8. Limitation of Liability. In no event and under no legal theory,
-      whether in tort (including negligence), contract, or otherwise,
-      unless required by applicable law (such as deliberate and grossly
-      negligent acts) or agreed to in writing, shall any Contributor be
-      liable to You for damages, including any direct, indirect, special,
-      incidental, or consequential damages of any character arising as a
-      result of this License or out of the use or inability to use the
-      Work (including but not limited to damages for loss of goodwill,
-      work stoppage, computer failure or malfunction, or any and all
-      other commercial damages or losses), even if such Contributor
-      has been advised of the possibility of such damages.
-
-   9. Accepting Warranty or Additional Liability. While redistributing
-      the Work or Derivative Works thereof, You may choose to offer,
-      and charge a fee for, acceptance of support, warranty, indemnity,
-      or other liability obligations and/or rights consistent with this
-      License. However, in accepting such obligations, You may act only
-      on Your own behalf and on Your sole responsibility, not on behalf
-      of any other Contributor, and only if You agree to indemnify,
-      defend, and hold each Contributor harmless for any liability
-      incurred by, or claims asserted against, such Contributor by reason
-      of your accepting any such warranty or additional liability.
-
-   END OF TERMS AND CONDITIONS
-
-   APPENDIX: How to apply the Apache License to your work.
-
-      To apply the Apache License to your work, attach the following
-      boilerplate notice, with the fields enclosed by brackets "[]"
-      replaced with your own identifying information. (Don't include
-      the brackets!)  The text should be enclosed in the appropriate
-      comment syntax for the file format. We also recommend that a
-      file or class name and description of purpose be included on the
-      same "printed page" as the copyright notice for easier
-      identification within third-party archives.
-
-   Copyright [2026] [Taranjeet Singh]
-
-   Licensed under the Apache License, Version 2.0 (the "License");
-   you may not use this file except in compliance with the License.
-   You may obtain a copy of the License at
-
-       http://www.apache.org/licenses/LICENSE-2.0
-
-   Unless required by applicable law or agreed to in writing, software
-   distributed under the License is distributed on an "AS IS" BASIS,
-   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-   See the License for the specific language governing permissions and
-   limitations under the License.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
@@ -11,7 +11,7 @@ export class Mem0Api implements ICredentialType {

 	displayName = 'Mem0 API';

-	icon: Icon = 'file:mem0.svg';
+	icon: Icon = { light: 'file:mem0.svg', dark: 'file:mem0.svg' };

 	documentationUrl = 'https://docs.mem0.ai/platform/quickstart';

@@ -21,7 +21,7 @@ export class Mem0 implements INodeType {
 	description: INodeTypeDescription = {
 		displayName: 'Mem0',
 		name: 'mem0',
-		icon: 'file:mem0.svg',
+		icon: { light: 'file:mem0.svg', dark: 'file:mem0.svg' },
 		group: ['transform'],
 		version: 1,
 		subtitle: '={{$parameter["operation"] + ": " + $parameter["resource"]}}',
@@ -1,6 +1,6 @@
 {
  "name": "@mem0/n8n-nodes-mem0",
-  "version": "0.1.2",
+  "version": "0.1.3",
  "description": "n8n community node for Mem0 — the memory layer for AI agents. Add, search, get, update, and delete long-term memories.",
  "keywords": [
    "n8n-community-node-package",
@@ -10,7 +10,7 @@
    "agents",
    "llm"
  ],
-  "license": "Apache-2.0",
+  "license": "MIT",
  "homepage": "https://mem0.ai",
  "author": {
    "name": "Mem0",
@@ -72,7 +72,7 @@
      "@qdrant/js-client-rest": "^1.18.0",
      "uuid@<11.1.1": ">=11.1.1",
      "esbuild": ">=0.28.1",
-      "undici@<6.27.0": ">=6.27.0 <8.0.0",
+      "undici@<7.29.0": ">=7.29.0 <8.0.0",
      "axios@<1.18.0": ">=1.18.0 <2.0.0",
      "postcss@<8.5.18": ">=8.5.18 <9.0.0",
      "mongoose@>=9.0.0 <9.7.2": ">=9.7.2 <10.0.0"
@@ -13,7 +13,7 @@ overrides:
  '@qdrant/js-client-rest': ^1.18.0
  uuid@<11.1.1: '>=11.1.1'
  esbuild: '>=0.28.1'
-  undici@<6.27.0: '>=6.27.0 <8.0.0'
+  undici@<7.29.0: '>=7.29.0 <8.0.0'
  axios@<1.18.0: '>=1.18.0 <2.0.0'
  postcss@<8.5.18: '>=8.5.18 <9.0.0'
  mongoose@>=9.0.0 <9.7.2: '>=9.7.2 <10.0.0'
@@ -2036,8 +2036,8 @@ packages:
  undici-types@6.21.0:
    resolution: {integrity: sha512-iwDZqg0QAGrg9Rav5H4n0M64c3mkR59cJ6wQp+7C4nI0gsmExaedaYLNO44eT4AtBBwjbTiGPMlt2Md0T9H9JQ==}

-  undici@7.28.0:
-    resolution: {integrity: sha512-cRZYrTDwWznlnRiPjggAGxZXanty6M8RV1ff8Wm4LWXBp7/IG8v5DnOm74DtUBp9OONpK75YlPnIjQqX0dBDtA==}
+  undici@7.29.0:
+    resolution: {integrity: sha512-IDxfleLmmbSskfWSUATiN1nfn2rDuvnMOqb5CWR92iIfojA0Ud+ulOAAEQ57LPr9rWmsreUyf5lwyao+7GNNVw==}
    engines: {node: '>=20.18.1'}

  util-deprecate@1.0.2:
@@ -2547,7 +2547,7 @@ snapshots:
    dependencies:
      '@qdrant/openapi-typescript-fetch': 1.2.6
      typescript: 5.9.3
-      undici: 7.28.0
+      undici: 7.29.0

  '@qdrant/openapi-typescript-fetch@1.2.6': {}

@@ -4121,7 +4121,7 @@ snapshots:

  undici-types@6.21.0: {}

-  undici@7.28.0: {}
+  undici@7.29.0: {}

  util-deprecate@1.0.2: {}

@@ -21,7 +21,7 @@ overrides:
  "@qdrant/js-client-rest": "^1.18.0"
  "uuid@<11.1.1": ">=11.1.1"
  "esbuild": ">=0.28.1"
-  "undici@<6.27.0": ">=6.27.0 <8.0.0"
+  "undici@<7.29.0": ">=7.29.0 <8.0.0"
  "axios@<1.18.0": ">=1.18.0 <2.0.0"
  "postcss@<8.5.18": ">=8.5.18 <9.0.0"
  "mongoose@>=9.0.0 <9.7.2": ">=9.7.2 <10.0.0"
@@ -73,8 +73,8 @@
      "form-data@<4.0.6": ">=4.0.6",
      "uuid@<11.1.1": ">=11.1.1",
      "esbuild": ">=0.28.1",
-      "undici@<6.27.0": ">=6.27.0 <8.0.0",
-      "undici@>=8.0.0 <8.5.0": ">=8.5.0",
+      "undici@<7.29.0": ">=7.29.0 <8.0.0",
+      "undici@>=8.0.0 <8.9.0": ">=8.9.0 <9.0.0",
      "axios@<1.18.0": ">=1.18.0 <2.0.0",
      "brace-expansion@>=3.0.0 <5.0.8": ">=5.0.8 <6.0.0",
      "postcss@<8.5.18": ">=8.5.18 <9.0.0",
@@ -8,8 +8,8 @@ overrides:
  form-data@<4.0.6: '>=4.0.6'
  uuid@<11.1.1: '>=11.1.1'
  esbuild: '>=0.28.1'
-  undici@<6.27.0: '>=6.27.0 <8.0.0'
-  undici@>=8.0.0 <8.5.0: '>=8.5.0'
+  undici@<7.29.0: '>=7.29.0 <8.0.0'
+  undici@>=8.0.0 <8.9.0: '>=8.9.0 <9.0.0'
  axios@<1.18.0: '>=1.18.0 <2.0.0'
  brace-expansion@>=3.0.0 <5.0.8: '>=5.0.8 <6.0.0'
  postcss@<8.5.18: '>=8.5.18 <9.0.0'
@@ -2354,12 +2354,12 @@ packages:
  undici-types@7.24.6:
    resolution: {integrity: sha512-WRNW+sJgj5OBN4/0JpHFqtqzhpbnV0GuB+OozA9gCL7a993SmU+1JBZCzLNxYsbMfIeDL+lTsphD5jN5N+n0zg==}

-  undici@7.28.0:
-    resolution: {integrity: sha512-cRZYrTDwWznlnRiPjggAGxZXanty6M8RV1ff8Wm4LWXBp7/IG8v5DnOm74DtUBp9OONpK75YlPnIjQqX0dBDtA==}
+  undici@7.29.0:
+    resolution: {integrity: sha512-IDxfleLmmbSskfWSUATiN1nfn2rDuvnMOqb5CWR92iIfojA0Ud+ulOAAEQ57LPr9rWmsreUyf5lwyao+7GNNVw==}
    engines: {node: '>=20.18.1'}

-  undici@8.5.0:
-    resolution: {integrity: sha512-xamtWoB1EshgjpmlXd7GGm2VfdDtw1+rD8uhry8pSNW3If6S8E0m2T2+orSKeZXEn/aPJMviCpDBA65WJt8zhg==}
+  undici@8.10.0:
+    resolution: {integrity: sha512-HvltHd7avK13QIw/oLe4qoOLyoVSoafqJ2jYOrtMRBkbYT31eiBQ8O0ehRKZiEZCMEyLFQNIADpgCWC5fALvYQ==}
    engines: {node: '>=22.19.0'}

  util-deprecate@1.0.2:
@@ -2942,7 +2942,7 @@ snapshots:
      minimatch: 10.2.5
      proper-lockfile: 4.1.2
      typebox: 1.1.38
-      undici: 8.5.0
+      undici: 8.10.0
      yaml: 2.9.0
    optionalDependencies:
      '@mariozechner/clipboard': 0.3.9
@@ -3203,7 +3203,7 @@ snapshots:
    dependencies:
      '@qdrant/openapi-typescript-fetch': 1.2.6
      typescript: 6.0.3
-      undici: 7.28.0
+      undici: 7.29.0

  '@qdrant/openapi-typescript-fetch@1.2.6': {}

@@ -4895,9 +4895,9 @@ snapshots:

  undici-types@7.24.6: {}

-  undici@7.28.0: {}
+  undici@7.29.0: {}

-  undici@8.5.0: {}
+  undici@8.10.0: {}

  util-deprecate@1.0.2: {}

@@ -5,8 +5,8 @@ overrides:
  "form-data@<4.0.6": ">=4.0.6"
  "uuid@<11.1.1": ">=11.1.1"
  "esbuild": ">=0.28.1"
-  "undici@<6.27.0": ">=6.27.0 <8.0.0"
-  "undici@>=8.0.0 <8.5.0": ">=8.5.0"
+  "undici@<7.29.0": ">=7.29.0 <8.0.0"
+  "undici@>=8.0.0 <8.9.0": ">=8.9.0 <9.0.0"
  "axios@<1.18.0": ">=1.18.0 <2.0.0"
  "brace-expansion@>=3.0.0 <5.0.8": ">=5.0.8 <6.0.0"
  "postcss@<8.5.18": ">=8.5.18 <9.0.0"
@@ -41,10 +41,17 @@
  ],
  "jest": {
    "testEnvironment": "node",
-    "testMatch": ["**/test/**/*.test.ts"],
-    "setupFiles": ["<rootDir>/test/setup.ts"],
+    "testMatch": [
+      "**/test/**/*.test.ts"
+    ],
+    "setupFiles": [
+      "<rootDir>/test/setup.ts"
+    ],
    "transform": {
-      "^.+\\.tsx?$": ["ts-jest", {}]
+      "^.+\\.tsx?$": [
+        "ts-jest",
+        {}
+      ]
    }
  },
  "dependencies": {
@@ -62,8 +69,9 @@
      "form-data@<4.0.6": ">=4.0.6",
      "uuid@<11.1.1": ">=11.1.1",
      "esbuild": ">=0.28.1",
-      "undici@<6.27.0": ">=6.27.0 <8.0.0",
-      "undici@>=8.0.0 <8.5.0": ">=8.5.0"
+      "undici@<7.29.0": ">=7.29.0 <8.0.0",
+      "undici@>=8.0.0 <8.9.0": ">=8.9.0 <9.0.0",
+      "brace-expansion@<1.1.16": ">=1.1.16 <2.0.0"
    }
  }
 }
@@ -8,8 +8,9 @@ overrides:
  form-data@<4.0.6: '>=4.0.6'
  uuid@<11.1.1: '>=11.1.1'
  esbuild: '>=0.28.1'
-  undici@<6.27.0: '>=6.27.0 <8.0.0'
-  undici@>=8.0.0 <8.5.0: '>=8.5.0'
+  undici@<7.29.0: '>=7.29.0 <8.0.0'
+  undici@>=8.0.0 <8.9.0: '>=8.9.0 <9.0.0'
+  brace-expansion@<1.1.16: '>=1.1.16 <2.0.0'

 importers:

@@ -413,8 +414,8 @@ packages:
    engines: {node: '>=6.0.0'}
    hasBin: true

-  brace-expansion@1.1.15:
-    resolution: {integrity: sha512-EwOCDEex4quD37XhqM3omwtMoJjr//isUZz1JopUNWms+4Z2ViyM/k1YIRePpoVNnQhENnxtFjLaxNHrT7xIUg==}
+  brace-expansion@1.1.18:
+    resolution: {integrity: sha512-Edep/X9fGqVNmzKBVsDYIOtD+z1tuezV70LBjdCst9Tqu76lsnvRiZ6oTic1n+/BIwX6QDGAO94PN4N2SADvtw==}

  braces@3.0.3:
    resolution: {integrity: sha512-yQbXgO/OSZVD2IsiLlro+7Hf6Q18EJrKSEsdoMzKePKXct3gvD8oLcOQdIzGupr5Fj+EDe8gO/lxc1BzfMpxvA==}
@@ -530,6 +531,7 @@ packages:

  crypto-js@4.2.0:
    resolution: {integrity: sha512-KALDyEYgpY+Rlob/iriUtjV6d5Eq+Y191A5g4UqLAi8CyGP9N1+FdVbkc1SxKc2r4YAYqG8JzO2KGL+AizD70Q==}
+    deprecated: Active development of CryptoJS has been discontinued. This library is no longer maintained.

  debug@4.4.3:
    resolution: {integrity: sha512-RGwwWnwQvkVfavKVt22FGLw+xYSdzARwm0ru6DhTVA3umU5hZc28V3kO4stgYryrTlLpuvgI9GiijltAjNbcqA==}
@@ -1995,7 +1997,7 @@ snapshots:

  baseline-browser-mapping@2.10.41: {}

-  brace-expansion@1.1.15:
+  brace-expansion@1.1.18:
    dependencies:
      balanced-match: 1.0.2
      concat-map: 0.0.1
@@ -2792,7 +2794,7 @@ snapshots:

  minimatch@3.1.5:
    dependencies:
-      brace-expansion: 1.1.15
+      brace-expansion: 1.1.18

  minimist@1.2.8: {}

@@ -5,5 +5,6 @@ overrides:
  "form-data@<4.0.6": ">=4.0.6"
  "uuid@<11.1.1": ">=11.1.1"
  "esbuild": ">=0.28.1"
-  "undici@<6.27.0": ">=6.27.0 <8.0.0"
-  "undici@>=8.0.0 <8.5.0": ">=8.5.0"
+  "undici@<7.29.0": ">=7.29.0 <8.0.0"
+  "undici@>=8.0.0 <8.9.0": ">=8.9.0 <9.0.0"
+  "brace-expansion@<1.1.16": ">=1.1.16 <2.0.0"
@@ -1,6 +1,6 @@
 {
  "name": "mem0ai",
-  "version": "3.1.4",
+  "version": "3.1.6",
  "description": "The Memory Layer For Your AI Apps",
  "main": "./dist/index.js",
  "module": "./dist/index.mjs",
@@ -86,6 +86,7 @@
  "license": "Apache-2.0",
  "devDependencies": {
    "@types/better-sqlite3": "^7.6.13",
+    "@types/oracledb": "^7.0.1",
    "@types/node": "^22.7.6",
    "@types/uuid": "^9.0.8",
    "dotenv": "^16.4.5",
@@ -139,6 +140,7 @@
    "mongodb": "^7.0.0",
    "weaviate-client": "^3.0.0",
    "ollama": "^0.5.14",
+    "oracledb": "^6.5.0 || ^7.0.0",
    "pg": "8.11.3",
    "redis": "^4.6.13",
    "@elastic/elasticsearch": "^9.0.0",
@@ -252,6 +254,9 @@
    },
    "iovalkey": {
      "optional": true
+    },
+    "oracledb": {
+      "optional": true
    }
  },
  "engines": {
@@ -287,18 +292,19 @@
      "glob@>=10.2.0 <10.5.0": "^10.5.0",
      "@modelcontextprotocol/sdk": "^1.25.4",
      "esbuild": ">=0.28.1",
-      "undici@<6.27.0": ">=6.27.0 <8.0.0",
+      "undici@<7.29.0": ">=7.29.0 <8.0.0",
      "@aws-sdk/client-bedrock-runtime": "3.967.0",
      "@aws-sdk/client-neptune-graph": "3.966.0",
      "axios@<1.18.0": ">=1.18.0 <2.0.0",
      "brace-expansion@<1.1.16": ">=1.1.16 <2.0.0",
-      "brace-expansion@>=2.0.0 <2.1.2": ">=2.1.2 <3.0.0",
+      "brace-expansion@>=2.0.0 <2.1.4": ">=2.1.4 <3.0.0",
      "brace-expansion@>=3.0.0 <5.0.8": ">=5.0.8 <6.0.0",
      "fast-xml-parser@>=5.0.0 <5.10.1": ">=5.10.1 <6.0.0",
      "tar@>=7.0.0 <7.5.21": ">=7.5.21 <8.0.0",
      "thrift@<0.23.0": "^0.23.0",
      "mongoose@>=9.0.0 <9.7.2": ">=9.7.2 <10.0.0",
-      "protobufjs@<7.6.5": ">=7.6.5 <8.0.0"
+      "protobufjs@<7.6.5": ">=7.6.5 <8.0.0",
+      "ip-address@<10.3.1": ">=10.3.1 <11.0.0"
    }
  }
 }
@@ -25,18 +25,19 @@ overrides:
  glob@>=10.2.0 <10.5.0: ^10.5.0
  '@modelcontextprotocol/sdk': ^1.25.4
  esbuild: '>=0.28.1'
-  undici@<6.27.0: '>=6.27.0 <8.0.0'
+  undici@<7.29.0: '>=7.29.0 <8.0.0'
  '@aws-sdk/client-bedrock-runtime': 3.967.0
  '@aws-sdk/client-neptune-graph': 3.966.0
  axios@<1.18.0: '>=1.18.0 <2.0.0'
  brace-expansion@<1.1.16: '>=1.1.16 <2.0.0'
-  brace-expansion@>=2.0.0 <2.1.2: '>=2.1.2 <3.0.0'
+  brace-expansion@>=2.0.0 <2.1.4: '>=2.1.4 <3.0.0'
  brace-expansion@>=3.0.0 <5.0.8: '>=5.0.8 <6.0.0'
  fast-xml-parser@>=5.0.0 <5.10.1: '>=5.10.1 <6.0.0'
  tar@>=7.0.0 <7.5.21: '>=7.5.21 <8.0.0'
  thrift@<0.23.0: ^0.23.0
  mongoose@>=9.0.0 <9.7.2: '>=9.7.2 <10.0.0'
  protobufjs@<7.6.5: '>=7.6.5 <8.0.0'
+  ip-address@<10.3.1: '>=10.3.1 <11.0.0'

 importers:

@@ -153,6 +154,9 @@ importers:
      openai:
        specifier: ^4.93.0
        version: 4.104.0(ws@5.2.5)(zod@3.25.76)
+      oracledb:
+        specifier: ^6.5.0 || ^7.0.0
+        version: 7.0.1
      pg:
        specifier: 8.11.3
        version: 8.11.3
@@ -181,6 +185,9 @@ importers:
      '@types/node':
        specifier: ^22.7.6
        version: 22.19.21
+      '@types/oracledb':
+        specifier: ^7.0.1
+        version: 7.0.1
      '@types/uuid':
        specifier: ^9.0.8
        version: 9.0.8
@@ -1911,6 +1918,9 @@ packages:
  '@types/normalize-package-data@2.4.4':
    resolution: {integrity: sha512-37i+OaWTh9qeK4LSHPsyRC7NahnGotNuZvjLSgcPzblpHB3rrCJxAOgI5gCdKm7coonsaX1Of0ILiTcnZjbfxA==}

+  '@types/oracledb@7.0.1':
+    resolution: {integrity: sha512-0A6m9YE4yu73KXehr5D6cbyALUzZoANGY4bG5cAPQIpJoJG4eMVPbR1Vau8ycnC25V/+F4Au1pdFSy8ODOAD0w==}
+
  '@types/pad-left@2.1.1':
    resolution: {integrity: sha512-Xd22WCRBydkGSApl5Bw0PhAOHKSVjNL3E3AwzKaps96IMraPqy5BvZIsBVK6JLwdybUzjHnuWVwpDd0JjTfHXA==}

@@ -2156,8 +2166,8 @@ packages:
  brace-expansion@1.1.16:
    resolution: {integrity: sha512-IDw48K2/2kRkg9LdJxurvq3lV3aBgq0REY89duEqFRthjlPdXHKMj7EnQOXVckxzgisinf3nHfrcE2FufFLXMw==}

-  brace-expansion@2.1.2:
-    resolution: {integrity: sha512-w5JZcKgdhDOgOwm8H+KgbosopHMuGcl6qbulwjtz3SM7I7P3yW1eAjzMPLrIE+NQ9vjgANKHWeMHnrT0OXW1oA==}
+  brace-expansion@2.1.4:
+    resolution: {integrity: sha512-hGfVzPxthbf3+2yjg/RBs60cB0FhqBS/zvdV/4wn4/BmN0bNMMHPc4V/BbFieqf1TKAGGAHnY4eSjajCl0f2Xg==}

  brace-expansion@5.0.8:
    resolution: {integrity: sha512-JZyDyq3D4AUifKTPOB7DELf6XsB3WdPuNxCtob1vFXPsSXhdAiHBWJ/tJ8HAc9aH84BK+5JFZLNkJKx3G9kzQg==}
@@ -3018,8 +3028,8 @@ packages:
    resolution: {integrity: sha512-4rTJX6Q5wTYEvxboXi8DsEiUo+OvqJGtLYOSGm37KpdRXsG5XJjbVtYKGJpPSWP+QT7rWscA4vsrdmzbEbenpw==}
    engines: {node: '>=18.12.0'}

-  ip-address@10.2.0:
-    resolution: {integrity: sha512-/+S6j4E9AHvW9SWMSEY9Xfy66O5PWvVEJ08O0y5JGyEKQpojb0K0GKpz/v5HJ/G0vi3D2sjGK78119oXZeE0qA==}
+  ip-address@10.4.0:
+    resolution: {integrity: sha512-oSK96Grm3aP6OrS263xVxbNDGVL7rzBtYdpGqlDG8iQdoenDoTs/nkki+DflYbAEE8Xl6o5YxhxlrKvI3nqKXQ==}
    engines: {node: '>= 12'}

  is-arrayish@0.2.1:
@@ -3796,6 +3806,10 @@ packages:
  openid-client@5.7.1:
    resolution: {integrity: sha512-jDBPgSVfTnkIh71Hg9pRvtJc6wTwqjRkN88+gCFtYWrlP4Yx2Dsrow8uPi3qLr/aeymPF3o2+dS+wOpglK04ew==}

+  oracledb@7.0.1:
+    resolution: {integrity: sha512-xlM0Ceh6A5stQLAdEfKf3pgCSkbOjQLo2ZPEi3+kXklz+KbZD3fLi/nsTSbQeZZNFBSFDNxdn1Ek3+bxG40M8w==}
+    engines: {node: '>=14.17'}
+
  p-finally@1.0.0:
    resolution: {integrity: sha512-LICb2p9CB7FS+0eR1oqWnHhp0FljGLZCWBE9aix0Uye9W8LTQPwMTYVGWQWIw9RdQiDg4+epXQODwIYJtSJaow==}
    engines: {node: '>=4'}
@@ -4613,8 +4627,8 @@ packages:
  undici-types@7.24.6:
    resolution: {integrity: sha512-WRNW+sJgj5OBN4/0JpHFqtqzhpbnV0GuB+OozA9gCL7a993SmU+1JBZCzLNxYsbMfIeDL+lTsphD5jN5N+n0zg==}

-  undici@7.28.0:
-    resolution: {integrity: sha512-cRZYrTDwWznlnRiPjggAGxZXanty6M8RV1ff8Wm4LWXBp7/IG8v5DnOm74DtUBp9OONpK75YlPnIjQqX0dBDtA==}
+  undici@7.29.0:
+    resolution: {integrity: sha512-IDxfleLmmbSskfWSUATiN1nfn2rDuvnMOqb5CWR92iIfojA0Ud+ulOAAEQ57LPr9rWmsreUyf5lwyao+7GNNVw==}
    engines: {node: '>=20.18.1'}

  update-browserslist-db@1.2.3:
@@ -6142,7 +6156,7 @@ snapshots:
      ms: 2.1.3
      secure-json-parse: 4.1.0
      tslib: 2.8.1
-      undici: 7.28.0
+      undici: 7.29.0
    transitivePeerDependencies:
      - supports-color

@@ -6698,7 +6712,7 @@ snapshots:
    dependencies:
      '@qdrant/openapi-typescript-fetch': 1.2.6
      typescript: 5.5.4
-      undici: 7.28.0
+      undici: 7.29.0

  '@qdrant/openapi-typescript-fetch@1.2.6': {}

@@ -7170,7 +7184,7 @@ snapshots:
  '@turbopuffer/turbopuffer@2.5.0':
    dependencies:
      pako: 2.2.0
-      undici: 7.28.0
+      undici: 7.29.0

  '@types/babel__core@7.20.5':
    dependencies:
@@ -7247,6 +7261,10 @@ snapshots:

  '@types/normalize-package-data@2.4.4': {}

+  '@types/oracledb@7.0.1':
+    dependencies:
+      '@types/node': 22.19.21
+
  '@types/pad-left@2.1.1': {}

  '@types/pg@8.11.0':
@@ -7525,7 +7543,7 @@ snapshots:
      balanced-match: 1.0.2
      concat-map: 0.0.1

-  brace-expansion@2.1.2:
+  brace-expansion@2.1.4:
    dependencies:
      balanced-match: 1.0.2

@@ -8431,7 +8449,7 @@ snapshots:
    transitivePeerDependencies:
      - supports-color

-  ip-address@10.2.0: {}
+  ip-address@10.4.0: {}

  is-arrayish@0.2.1: {}

@@ -9059,7 +9077,7 @@ snapshots:

  minimatch@9.0.9:
    dependencies:
-      brace-expansion: 2.1.2
+      brace-expansion: 2.1.4

  minimist@1.2.8: {}

@@ -9349,6 +9367,8 @@ snapshots:
      object-hash: 2.2.0
      oidc-token-hash: 5.2.0

+  oracledb@7.0.1: {}
+
  p-finally@1.0.0: {}

  p-limit@2.3.0:
@@ -9867,7 +9887,7 @@ snapshots:

  socks@2.8.9:
    dependencies:
-      ip-address: 10.2.0
+      ip-address: 10.4.0
      smart-buffer: 4.2.0

  source-map-support@0.5.13:
@@ -10213,7 +10233,7 @@ snapshots:

  undici-types@7.24.6: {}

-  undici@7.28.0: {}
+  undici@7.29.0: {}

  update-browserslist-db@1.2.3(browserslist@4.28.2):
    dependencies:
@@ -25,13 +25,14 @@ overrides:
  "glob@>=10.2.0 <10.5.0": "^10.5.0"
  "@modelcontextprotocol/sdk": "^1.25.4"
  "esbuild": ">=0.28.1"
-  "undici@<6.27.0": ">=6.27.0 <8.0.0"
+  "undici@<7.29.0": ">=7.29.0 <8.0.0"
  "axios@<1.18.0": ">=1.18.0 <2.0.0"
  "brace-expansion@<1.1.16": ">=1.1.16 <2.0.0"
-  "brace-expansion@>=2.0.0 <2.1.2": ">=2.1.2 <3.0.0"
+  "brace-expansion@>=2.0.0 <2.1.4": ">=2.1.4 <3.0.0"
  "brace-expansion@>=3.0.0 <5.0.8": ">=5.0.8 <6.0.0"
  "fast-xml-parser@>=5.0.0 <5.10.1": ">=5.10.1 <6.0.0"
  "tar@>=7.0.0 <7.5.21": ">=7.5.21 <8.0.0"
  "thrift@<0.23.0": "^0.23.0"
  "mongoose@>=9.0.0 <9.7.2": ">=9.7.2 <10.0.0"
  "protobufjs@<7.6.5": ">=7.6.5 <8.0.0"
+  "ip-address@<10.3.1": ">=10.3.1 <11.0.0"
@@ -81,6 +81,8 @@ class APIError extends Error {
 interface ClientOptions {
  apiKey: string;
  host?: string;
+  /** Max cached identities per process. Defaults to 50. */
+  identityCacheMax?: number;
 }

 interface ClientIdentity {
@@ -90,7 +92,7 @@ interface ClientIdentity {
 }

 // Shares one ping per (host, api key) across clients; FIFO-capped.
-const IDENTITY_CACHE_MAX = 50;
+const IDENTITY_CACHE_MAX_DEFAULT = 50;
 const identityByCredentials = new Map<string, Promise<ClientIdentity>>();

 export default class MemoryClient {
@@ -102,6 +104,7 @@ export default class MemoryClient {
  client: any;
  telemetryId: string;
  private initialized: Promise<void>;
+  private identityCacheMax: number;

  _validateApiKey(): any {
    if (!this.apiKey) {
@@ -120,6 +123,8 @@ export default class MemoryClient {
    this.host = options.host || "https://api.mem0.ai";
    this.organizationId = null;
    this.projectId = null;
+    this.identityCacheMax =
+      options.identityCacheMax ?? IDENTITY_CACHE_MAX_DEFAULT;

    this.headers = {
      Authorization: `Token ${this.apiKey}`,
@@ -136,7 +141,7 @@ export default class MemoryClient {

    this.telemetryId = "";

-    // Requests never wait on this; only telemetry does.
+    // Memory requests never wait on this; telemetry and _awaitIdentity do.
    this.initialized = this._resolveIdentity();
  }

@@ -146,7 +151,7 @@ export default class MemoryClient {
    let shared = identityByCredentials.get(credentials);
    if (!shared) {
      shared = this._initializeClient();
-      if (identityByCredentials.size >= IDENTITY_CACHE_MAX) {
+      if (identityByCredentials.size >= this.identityCacheMax) {
        identityByCredentials.delete(
          identityByCredentials.keys().next().value!,
        );
@@ -165,6 +170,11 @@ export default class MemoryClient {
    });
  }

+  // Blocks until the ping has populated organizationId/projectId.
+  private async _awaitIdentity(): Promise<void> {
+    await this.initialized;
+  }
+
  private async _initializeClient(): Promise<ClientIdentity> {
    try {
      await this.ping();
@@ -616,6 +626,7 @@ export default class MemoryClient {
    const payloadKeys = Object.keys(options || {});
    this._captureEvent("get_project", [payloadKeys]);
    const { fields } = options;
+    await this._awaitIdentity();

    if (!(this.organizationId && this.projectId)) {
      throw new Error(
@@ -639,6 +650,7 @@ export default class MemoryClient {
    prompts: PromptUpdatePayload,
  ): Promise<Record<string, any>> {
    this._captureEvent("update_project", []);
+    await this._awaitIdentity();
    if (!(this.organizationId && this.projectId)) {
      throw new Error(
        "organizationId and projectId must be set to update instructions or categories",
@@ -659,7 +671,11 @@ export default class MemoryClient {
  // WebHooks
  async getWebhooks(data?: { projectId?: string }): Promise<Array<Webhook>> {
    this._captureEvent("get_webhooks", []);
+    if (!data?.projectId) await this._awaitIdentity();
    const project_id = data?.projectId || this.projectId;
+    if (!project_id) {
+      throw new Error("projectId must be set to access webhooks");
+    }
    const response = await this._fetchWithErrorHandling(
      `${this.host}/api/v1/webhooks/projects/${project_id}/`,
      {
@@ -671,6 +687,10 @@ export default class MemoryClient {

  async createWebhook(webhook: WebhookCreatePayload): Promise<Webhook> {
    this._captureEvent("create_webhook", []);
+    await this._awaitIdentity();
+    if (!this.projectId) {
+      throw new Error("projectId must be set to create a webhook");
+    }
    const body = {
      name: webhook.name,
      url: webhook.url,
@@ -12,6 +12,7 @@ export interface AddMemoryOptions extends EntityOptions {
  infer?: boolean;
  customCategories?: custom_categories[];
  customInstructions?: string;
+  agentCustomInstructions?: string;
  timestamp?: number;
  expirationDate?: string;
  structuredDataSchema?: Record<string, any>;
@@ -60,6 +61,7 @@ export interface ProjectOptions {

 export interface PromptUpdatePayload {
  customInstructions?: string;
+  agentCustomInstructions?: string;
  customCategories?: custom_categories[];
  version?: string;
  memoryDepth?: string | null;
@@ -174,6 +176,7 @@ export interface PaginatedMemories {

 export interface ProjectResponse {
  customInstructions?: string;
+  agentCustomInstructions?: string;
  // The API returns category objects (`[{ "<name>": "<description>" }]`),
  // not bare strings (see issue #5738).
  customCategories?: custom_categories[];
@@ -74,6 +74,35 @@ describe("MemoryClient - add()", () => {
    expect(getFetchBody(call!).expiration_date).toBe("2030-01-31");
  });

+  test("serializes agentCustomInstructions as agent_custom_instructions", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/v3/memories/add/", { status: 200, body: [createMockMemory()] });
+    const mock = setupMockFetch(extra);
+
+    const client = new MemoryClient({ apiKey: TEST_API_KEY });
+    await client.add([{ role: "user", content: "test" }], {
+      agentId: "a1",
+      agentCustomInstructions: "Remember tool failures",
+    });
+
+    const call = findFetchCall(mock, "/v3/memories/add/", "POST");
+    expect(getFetchBody(call!).agent_custom_instructions).toBe(
+      "Remember tool failures",
+    );
+  });
+
+  test("omits agent_custom_instructions when it is not passed", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/v3/memories/add/", { status: 200, body: [createMockMemory()] });
+    const mock = setupMockFetch(extra);
+
+    const client = new MemoryClient({ apiKey: TEST_API_KEY });
+    await client.add([{ role: "user", content: "test" }], { userId: "u1" });
+
+    const call = findFetchCall(mock, "/v3/memories/add/", "POST");
+    expect(getFetchBody(call!)).not.toHaveProperty("agent_custom_instructions");
+  });
+
  test("throws an error when given an empty messages array", async () => {
    setupMockFetch();

@@ -0,0 +1,228 @@
+/**
+ * MemoryClient unit tests — org/project identity resolution ordering.
+ *
+ * Each test calls the method under test as the first awaited operation on a
+ * fresh client. Any earlier await on the instance resolves identity and voids
+ * the test.
+ */
+import { MemoryClient } from "../mem0";
+import { TEST_ORG_ID, TEST_PROJECT_ID } from "./helpers";
+import { setupMockFetch, installConsoleSuppression } from "./setup";
+
+installConsoleSuppression();
+
+// Distinct key per client keeps each test off the module-scope identity cache.
+let keySeq = 0;
+const freshClient = () =>
+  new MemoryClient({ apiKey: `test-api-key-identity-${keySeq++}` });
+
+// Selects by path; the ping and PostHog calls also land in the mock.
+const findUrlOrNone = (mock: jest.Mock, needle: string): string | undefined =>
+  mock.mock.calls
+    .map((c: [string, RequestInit]) => c[0])
+    .find((u: string) => u.includes(needle));
+
+const findUrl = (mock: jest.Mock, needle: string): string => {
+  const url = findUrlOrNone(mock, needle);
+  expect(url).toBeDefined();
+  return url as string;
+};
+
+describe("MemoryClient - project-scoped URLs on a fresh client", () => {
+  test("getProject uses the resolved org and project ids", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", {
+      status: 200,
+      body: { custom_instructions: "Be helpful" },
+    });
+    const mock = setupMockFetch(extra);
+
+    await freshClient().getProject({ fields: ["custom_instructions"] });
+
+    const url = findUrl(mock, "/api/v1/orgs/organizations/");
+    expect(url).toContain(`/api/v1/orgs/organizations/${TEST_ORG_ID}/`);
+    expect(url).toContain(`/projects/${TEST_PROJECT_ID}/`);
+  });
+
+  test("updateProject uses the resolved org and project ids", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", {
+      status: 200,
+      body: { custom_instructions: "Updated" },
+    });
+    const mock = setupMockFetch(extra);
+
+    await freshClient().updateProject({ customInstructions: "Updated" });
+
+    const url = findUrl(mock, "/api/v1/orgs/organizations/");
+    expect(url).toContain(`/api/v1/orgs/organizations/${TEST_ORG_ID}/`);
+    expect(url).toContain(`/projects/${TEST_PROJECT_ID}/`);
+  });
+
+  test("getWebhooks targets the resolved project, not null", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/webhooks/projects/", { status: 200, body: [] });
+    const mock = setupMockFetch(extra);
+
+    await freshClient().getWebhooks();
+
+    const url = findUrl(mock, "/api/v1/webhooks/projects/");
+    expect(url).toContain(`/api/v1/webhooks/projects/${TEST_PROJECT_ID}/`);
+    expect(url).not.toContain("/projects/null/");
+    expect(url).not.toContain("/projects/undefined/");
+  });
+
+  test("createWebhook targets the resolved project, not null", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/webhooks/projects/", {
+      status: 200,
+      body: { webhook_id: "wh_1" },
+    });
+    const mock = setupMockFetch(extra);
+
+    await freshClient().createWebhook({
+      name: "hook",
+      url: "https://example.com/hook",
+      eventTypes: ["memory_add"],
+    });
+
+    const url = findUrl(mock, "/api/v1/webhooks/projects/");
+    expect(url).toContain(`/api/v1/webhooks/projects/${TEST_PROJECT_ID}/`);
+    expect(url).not.toContain("/projects/null/");
+    expect(url).not.toContain("/projects/undefined/");
+  });
+
+  test("an explicit projectId is honored without waiting on identity", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/webhooks/projects/", { status: 200, body: [] });
+    const mock = setupMockFetch(extra);
+
+    await freshClient().getWebhooks({ projectId: "proj_explicit_789" });
+
+    expect(findUrl(mock, "/api/v1/webhooks/projects/")).toContain(
+      "/api/v1/webhooks/projects/proj_explicit_789/",
+    );
+  });
+});
+
+describe("MemoryClient - memory endpoints do not wait on identity", () => {
+  test("search completes while the ping is still pending", async () => {
+    // The ping never settles.
+    const mock = jest.fn((url: string) => {
+      if (url.includes("/v1/ping/")) return new Promise(() => {});
+      return Promise.resolve({
+        ok: true,
+        status: 200,
+        json: () => Promise.resolve({ results: [] }),
+        text: () => Promise.resolve(""),
+      });
+    });
+    global.fetch = mock as unknown as typeof global.fetch;
+
+    await expect(
+      freshClient().search("query", { filters: { user_id: "alice" } }),
+    ).resolves.toBeDefined();
+
+    expect(findUrl(mock, "/v3/memories/search/")).toBeDefined();
+  });
+});
+
+describe("MemoryClient - identity cache overflow", () => {
+  const pingCount = (mock: jest.Mock) =>
+    mock.mock.calls.filter((c: [string, RequestInit]) =>
+      c[0].includes("/v1/ping/"),
+    ).length;
+
+  // The cache is module scope, so each test needs its own module registry.
+  const isolatedClient = (): typeof MemoryClient => {
+    jest.resetModules();
+    // eslint-disable-next-line @typescript-eslint/no-require-imports
+    return require("../mem0").MemoryClient;
+  };
+
+  const projectOk = () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", { status: 200, body: {} });
+    return setupMockFetch(extra);
+  };
+
+  test("a reused credential pair pings once", async () => {
+    const mock = projectOk();
+    const Client = isolatedClient();
+
+    const opts = { apiKey: "cache-warm-key", identityCacheMax: 1 };
+    await new Client(opts).getProject({ fields: [] });
+    await new Client(opts).getProject({ fields: [] });
+
+    expect(pingCount(mock)).toBe(1);
+  });
+
+  test("an evicted credential pair re-pings and still resolves identity", async () => {
+    const mock = projectOk();
+    const Client = isolatedClient();
+    const opts = (apiKey: string) => ({ apiKey, identityCacheMax: 1 });
+
+    await new Client(opts("key-a")).getProject({ fields: [] });
+    // Evicts key-a.
+    await new Client(opts("key-b")).getProject({ fields: [] });
+    expect(pingCount(mock)).toBe(2);
+
+    await new Client(opts("key-a")).getProject({ fields: [] });
+
+    // Eviction costs a ping; it never yields an unresolved identity.
+    expect(pingCount(mock)).toBe(3);
+    expect(findUrl(mock, "/api/v1/orgs/organizations/")).toContain(
+      `/api/v1/orgs/organizations/${TEST_ORG_ID}/`,
+    );
+  });
+});
+
+describe("MemoryClient - unresolved identity", () => {
+  const pingFails = () => {
+    const mock = jest.fn((url: string) => {
+      if (url.includes("/v1/ping/")) {
+        return Promise.resolve({
+          ok: false,
+          status: 500,
+          text: () => Promise.resolve("boom"),
+          json: () => Promise.resolve({}),
+        });
+      }
+      return Promise.resolve({
+        ok: true,
+        status: 200,
+        json: () => Promise.resolve([]),
+        text: () => Promise.resolve(""),
+      });
+    });
+    global.fetch = mock as unknown as typeof global.fetch;
+    return mock;
+  };
+
+  test("getProject reports the unset ids", async () => {
+    pingFails();
+    await expect(freshClient().getProject({ fields: [] })).rejects.toThrow(
+      "organizationId and projectId must be set",
+    );
+  });
+
+  test("getWebhooks reports the unset project instead of requesting null", async () => {
+    const mock = pingFails();
+    await expect(freshClient().getWebhooks()).rejects.toThrow(
+      "projectId must be set",
+    );
+    expect(findUrlOrNone(mock, "/api/v1/webhooks/")).toBeUndefined();
+  });
+
+  test("createWebhook reports the unset project instead of requesting null", async () => {
+    const mock = pingFails();
+    await expect(
+      freshClient().createWebhook({
+        name: "hook",
+        url: "https://example.com/hook",
+        eventTypes: ["memory_add"],
+      }),
+    ).rejects.toThrow("projectId must be set");
+    expect(findUrlOrNone(mock, "/api/v1/webhooks/")).toBeUndefined();
+  });
+});
@@ -96,6 +96,84 @@ describe("MemoryClient - updateProject()", () => {
      "Updated instructions",
    );
  });
+
+  test("camelCases agentCustomInstructions into agent_custom_instructions", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", {
+      status: 200,
+      body: { message: "Updated" },
+    });
+    const mock = setupMockFetch(extra);
+
+    const client = new MemoryClient({ apiKey: TEST_API_KEY });
+    await client.ping();
+    await client.updateProject({
+      agentCustomInstructions: "Remember tool failures",
+    });
+
+    const call = findFetchCall(mock, "/api/v1/orgs/organizations/", "PATCH");
+    expect(getFetchBody(call!).agent_custom_instructions).toBe(
+      "Remember tool failures",
+    );
+  });
+
+  test("sends both instruction sets in one PATCH", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", {
+      status: 200,
+      body: { message: "Updated" },
+    });
+    const mock = setupMockFetch(extra);
+
+    const client = new MemoryClient({ apiKey: TEST_API_KEY });
+    await client.ping();
+    await client.updateProject({
+      customInstructions: "Remember user preferences",
+      agentCustomInstructions: "Remember tool failures",
+    });
+
+    const body = getFetchBody(
+      findFetchCall(mock, "/api/v1/orgs/organizations/", "PATCH")!,
+    );
+    expect(body.custom_instructions).toBe("Remember user preferences");
+    expect(body.agent_custom_instructions).toBe("Remember tool failures");
+  });
+
+  test("an empty string round-trips, so the field can be cleared", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", {
+      status: 200,
+      body: { message: "Updated" },
+    });
+    const mock = setupMockFetch(extra);
+
+    const client = new MemoryClient({ apiKey: TEST_API_KEY });
+    await client.ping();
+    await client.updateProject({ agentCustomInstructions: "" });
+
+    const body = getFetchBody(
+      findFetchCall(mock, "/api/v1/orgs/organizations/", "PATCH")!,
+    );
+    expect(body).toHaveProperty("agent_custom_instructions", "");
+  });
+
+  test("omits agent_custom_instructions when it is not passed", async () => {
+    const extra = new Map<string, { status: number; body: unknown }>();
+    extra.set("/api/v1/orgs/organizations/", {
+      status: 200,
+      body: { message: "Updated" },
+    });
+    const mock = setupMockFetch(extra);
+
+    const client = new MemoryClient({ apiKey: TEST_API_KEY });
+    await client.ping();
+    await client.updateProject({ customInstructions: "Be helpful" });
+
+    const body = getFetchBody(
+      findFetchCall(mock, "/api/v1/orgs/organizations/", "PATCH")!,
+    );
+    expect(body).not.toHaveProperty("agent_custom_instructions");
+  });
 });

 // ─── feedback() ─────────────────────────────────────────
@@ -51,6 +51,7 @@ export * from "./vector_stores/milvus";
 export * from "./vector_stores/mongodb";
 export * from "./vector_stores/opensearch";
 export * from "./vector_stores/weaviate";
+export * from "./vector_stores/oracledb";
 export * from "./rerankers/base";
 export * from "./rerankers/cohere";
 export * from "./rerankers/llm";
@@ -556,11 +556,18 @@ export class Memory {
    }
  }

+  private escapeScopeValue(val: unknown): string {
+    return String(val)
+      .replace(/%/g, "%25")
+      .replace(/&/g, "%26")
+      .replace(/=/g, "%3D");
+  }
+
  private buildSessionScope(filters: SearchFilters): string {
    const parts: string[] = [];
    for (const key of ["agent_id", "run_id", "user_id"].sort()) {
      const val = (filters as any)[key];
-      if (val) parts.push(`${key}=${val}`);
+      if (val) parts.push(`${key}=${this.escapeScopeValue(val)}`);
    }
    return parts.join("&");
  }
@@ -71,6 +71,7 @@ import { TurbopufferDB } from "../vector_stores/turbopuffer";
 import { Milvus } from "../vector_stores/milvus";
 import { MongoDB } from "../vector_stores/mongodb";
 import { WeaviateDB } from "../vector_stores/weaviate";
+import { OracleAIVectorSearch } from "../vector_stores/oracledb";

 export class EmbedderFactory {
  static create(provider: string, config: EmbeddingConfig): Embedder {
@@ -205,6 +206,8 @@ export class VectorStoreFactory {
        return new MongoDB(config as any);
      case "weaviate":
        return new WeaviateDB(config as any);
+      case "oracledb":
+        return new OracleAIVectorSearch(config as any);
      default:
        throw new Error(`Unsupported vector store provider: ${provider}`);
    }
@@ -0,0 +1,764 @@
+import type { BindParameters, Connection, Pool } from "oracledb";
+import { v4 as uuidv4 } from "uuid";
+import { VectorStore } from "./base";
+import { SearchFilters, VectorStoreConfig, VectorStoreResult } from "../types";
+import { loadPeer } from "../utils/load_peer";
+
+const DISTANCE_METRICS = [
+  "COSINE",
+  "EUCLIDEAN",
+  "EUCLIDEAN_SQUARED",
+  "DOT",
+  "HAMMING",
+  "MANHATTAN",
+] as const;
+
+type DistanceMetric = (typeof DISTANCE_METRICS)[number];
+type IndexType = "HNSW" | "IVF";
+
+const SCORE_FROM_DISTANCE: Record<DistanceMetric, (d: number) => number> = {
+  COSINE: (d) => Math.max(0, Math.min(1, 1 - d)),
+  EUCLIDEAN: (d) => 1 / (1 + Math.max(0, d)),
+  EUCLIDEAN_SQUARED: (d) => 1 / (1 + Math.sqrt(Math.max(0, d))),
+  HAMMING: (d) => 1 / (1 + Math.max(0, d)),
+  MANHATTAN: (d) => 1 / (1 + Math.max(0, d)),
+  DOT: (d) => -d,
+};
+
+const INDEX_PARAMETER_RANGES: Record<
+  IndexType,
+  Record<string, [number, number]>
+> = {
+  HNSW: {
+    neighbors: [2, 2048],
+    efconstruction: [1, 65535],
+  },
+  IVF: {
+    "neighbor partitions": [1, 10_000_000],
+    samples_per_partition: [1, Number.MAX_SAFE_INTEGER],
+    min_vectors_per_partition: [0, Number.MAX_SAFE_INTEGER],
+  },
+};
+
+const IDENTIFIER_RE = /^(?:"[^"]+"|[^".]+)(?:\.(?:"[^"]+"|[^".]+))*$/;
+const METADATA_KEY_RE = /^[a-zA-Z0-9_.[\],\s*]+$/;
+
+export function quoteIdentifier(name: string): string {
+  const trimmed = name.trim();
+  if (!IDENTIFIER_RE.test(trimmed)) {
+    throw new Error(`Identifier name ${name} is not valid.`);
+  }
+  return [...trimmed.matchAll(/"([^"]+)"|([^".]+)/g)]
+    .map((m) => `"${m[1] ?? m[2]}"`)
+    .join(".");
+}
+
+function jsonPath(metadataKey: string): string {
+  if (!METADATA_KEY_RE.test(metadataKey)) {
+    throw new Error(
+      `Invalid metadata key '${metadataKey}'. Only letters, numbers, underscores, ` +
+        `nesting via '.', and array wildcards '[*]' are allowed.`,
+    );
+  }
+  return metadataKey
+    .split(".")
+    .map((part) =>
+      part.endsWith("[*]") ? `."${part.slice(0, -3)}"[*]` : `."${part}"`,
+    )
+    .join("");
+}
+
+const COMPARISON_OPERATORS: Record<string, string> = {
+  eq: "==",
+  ne: "!=",
+  gt: ">",
+  gte: ">=",
+  lt: "<",
+  lte: "<=",
+};
+
+const FIELD_OPERATORS = new Set([
+  ...Object.keys(COMPARISON_OPERATORS),
+  "in",
+  "nin",
+  "contains",
+  "icontains",
+]);
+
+const LOGICAL_OPERATORS: Record<string, "and" | "or" | "not"> = {
+  $and: "and",
+  $or: "or",
+  $not: "not",
+  AND: "and",
+  OR: "or",
+  NOT: "not",
+};
+
+function isScalar(value: any): boolean {
+  return value === null || (typeof value !== "object" && !Array.isArray(value));
+}
+
+function bindFilterValue(
+  value: any,
+  binds: Record<string, any>,
+): [string, string] {
+  const name = `f_${Object.keys(binds).length}`;
+  binds[name] = value;
+  return [`$${name}`, `:${name} AS "${name}"`];
+}
+
+function jsonExists(
+  path: string,
+  predicate: string,
+  passings: string[],
+): string {
+  const passingClause =
+    passings.length > 0 ? ` PASSING ${passings.join(", ")}` : "";
+  return `JSON_EXISTS(payload, '$${path}?(${predicate})'${passingClause})`;
+}
+
+function buildFieldCondition(
+  metadataKey: string,
+  value: any,
+  binds: Record<string, any>,
+): string {
+  const path = jsonPath(metadataKey);
+
+  if (value === "*") {
+    return `JSON_EXISTS(payload, '$${path}')`;
+  }
+
+  if (isScalar(value)) {
+    if (value === null) {
+      return jsonExists(path, "@ == null", []);
+    }
+    const [variable, passing] = bindFilterValue(value, binds);
+    return jsonExists(path, `@ == ${variable}`, [passing]);
+  }
+
+  if (Array.isArray(value)) {
+    throw new Error(
+      `Oracle filter for field '${metadataKey}' must be a scalar or an operator object`,
+    );
+  }
+
+  const operators = Object.entries(value);
+  if (operators.length === 0) {
+    throw new Error(
+      `Operator filter for field '${metadataKey}' must not be empty`,
+    );
+  }
+
+  const unsupported = operators
+    .map(([op]) => op)
+    .filter((op) => !FIELD_OPERATORS.has(op));
+  if (unsupported.length > 0) {
+    throw new Error(
+      `Unsupported Oracle filter operator(s) for field '${metadataKey}': ${unsupported.sort().join(", ")}`,
+    );
+  }
+
+  const predicates: string[] = [];
+  const passings: string[] = [];
+  const additionalClauses: string[] = [];
+
+  for (const [operator, operand] of operators) {
+    if (operator in COMPARISON_OPERATORS) {
+      if (!isScalar(operand)) {
+        throw new Error(
+          `Oracle filter operator '${operator}' requires a scalar value`,
+        );
+      }
+      if (operand === null) {
+        if (operator !== "eq" && operator !== "ne") {
+          throw new Error(
+            `Oracle filter operator '${operator}' does not support null`,
+          );
+        }
+        predicates.push(`@ ${COMPARISON_OPERATORS[operator]} null`);
+        continue;
+      }
+      const [variable, passing] = bindFilterValue(operand, binds);
+      predicates.push(`@ ${COMPARISON_OPERATORS[operator]} ${variable}`);
+      passings.push(passing);
+      continue;
+    }
+
+    if (operator === "in" || operator === "nin") {
+      if (!Array.isArray(operand) || operand.length === 0) {
+        throw new Error(
+          `Oracle filter operator '${operator}' requires a non-empty array`,
+        );
+      }
+
+      const variables: string[] = [];
+      const listPassings: string[] = [];
+      for (const item of operand) {
+        if (!isScalar(item)) {
+          throw new Error(
+            `Oracle filter operator '${operator}' requires scalar values`,
+          );
+        }
+        if (item === null) {
+          variables.push("null");
+          continue;
+        }
+        const [variable, passing] = bindFilterValue(item, binds);
+        variables.push(variable);
+        listPassings.push(passing);
+      }
+
+      const membership = jsonExists(
+        path,
+        `@ in (${variables.join(", ")})`,
+        listPassings,
+      );
+      additionalClauses.push(
+        operator === "in" ? membership : `NOT (${membership})`,
+      );
+      continue;
+    }
+
+    if (typeof operand !== "string") {
+      throw new Error(
+        `Oracle filter operator '${operator}' requires a string value`,
+      );
+    }
+
+    if (operator === "contains") {
+      const [variable, passing] = bindFilterValue(operand, binds);
+      predicates.push(`@ has substring ${variable}`);
+      passings.push(passing);
+    } else {
+      const [variable, passing] = bindFilterValue(operand.toLowerCase(), binds);
+      predicates.push(`@.lower() has substring ${variable}`);
+      passings.push(passing);
+    }
+  }
+
+  const clauses = [...additionalClauses];
+  if (predicates.length > 0) {
+    clauses.unshift(jsonExists(path, predicates.join(" && "), passings));
+  }
+
+  return clauses.length === 1 ? clauses[0] : `(${clauses.join(" AND ")})`;
+}
+
+export function buildFilterGroup(
+  filters: Record<string, any>,
+  binds: Record<string, any>,
+): string {
+  const entries = Object.entries(filters ?? {});
+  if (entries.length === 0) {
+    throw new Error("Oracle filter groups must be non-empty objects");
+  }
+
+  const clauses: string[] = [];
+  for (const [key, value] of entries) {
+    const logicalOperator = LOGICAL_OPERATORS[key];
+    if (logicalOperator) {
+      if (!Array.isArray(value) || value.length === 0) {
+        throw new Error(
+          `Logical filter operator '${key}' requires a non-empty array`,
+        );
+      }
+      const nested = value.map((condition) =>
+        buildFilterGroup(condition, binds),
+      );
+      if (logicalOperator === "not") {
+        clauses.push(`NOT (${nested.join(" OR ")})`);
+      } else {
+        clauses.push(
+          `(${nested.join(logicalOperator === "and" ? " AND " : " OR ")})`,
+        );
+      }
+      continue;
+    }
+
+    if (key.startsWith("$")) {
+      throw new Error(`Unsupported Oracle logical filter operator: ${key}`);
+    }
+
+    clauses.push(buildFieldCondition(key, value, binds));
+  }
+
+  return clauses.length === 1 ? clauses[0] : `(${clauses.join(" AND ")})`;
+}
+
+export function buildWhereClause(
+  filters?: SearchFilters,
+): [string, Record<string, any>] {
+  if (!filters || Object.keys(filters).length === 0) {
+    return ["", {}];
+  }
+  const binds: Record<string, any> = {};
+  return [`WHERE ${buildFilterGroup(filters, binds)}`, binds];
+}
+
+interface OracleDBConfig extends VectorStoreConfig {
+  connectionParams?: Record<string, any>;
+  useConnectionPool?: boolean;
+  client?: Connection | Pool;
+  collectionName?: string;
+  embeddingModelDims?: number;
+  distanceMetric?: DistanceMetric;
+  doCreateIndex?: boolean;
+  indexType?: IndexType;
+  indexName?: string;
+  indexParameters?: Record<string, number>;
+  indexAccuracy?: number;
+}
+
+export class OracleAIVectorSearch implements VectorStore {
+  private readonly collectionName: string;
+  private readonly indexName: string;
+  private readonly embeddingModelDims: number;
+  private readonly distanceMetric: DistanceMetric;
+  private readonly indexType: IndexType;
+  private readonly indexParameters: Record<string, number>;
+  private readonly indexAccuracy?: number;
+  private readonly doCreateIndex: boolean;
+  private readonly config: OracleDBConfig;
+  private oracledb: any;
+  private client?: Connection | Pool;
+  private ownsClient = false;
+  private _initPromise?: Promise<void>;
+
+  constructor(config: OracleDBConfig) {
+    if (!config.connectionParams && !config.client) {
+      throw new Error(
+        "Must provide at least one of `connectionParams` and `client`",
+      );
+    }
+
+    this.collectionName = quoteIdentifier(config.collectionName || "mem0");
+    this.indexName = quoteIdentifier(
+      config.indexName || `${config.collectionName || "mem0"}_VEC_IDX`,
+    );
+
+    this.embeddingModelDims = config.embeddingModelDims ?? 1536;
+    if (
+      !Number.isInteger(this.embeddingModelDims) ||
+      this.embeddingModelDims <= 0
+    ) {
+      throw new Error("`embeddingModelDims` must be a positive integer");
+    }
+
+    const distanceMetric = (config.distanceMetric ??
+      "COSINE") as string as DistanceMetric;
+    this.distanceMetric = distanceMetric.toUpperCase() as DistanceMetric;
+    if (!DISTANCE_METRICS.includes(this.distanceMetric)) {
+      throw new Error(`Unsupported distance metric: ${config.distanceMetric}`);
+    }
+
+    const indexType = (config.indexType ?? "HNSW") as string;
+    this.indexType = indexType.toUpperCase() as IndexType;
+    if (this.indexType !== "HNSW" && this.indexType !== "IVF") {
+      throw new Error(`Unsupported index type: ${config.indexType}`);
+    }
+
+    this.indexAccuracy = config.indexAccuracy;
+    if (
+      this.indexAccuracy !== undefined &&
+      (!Number.isInteger(this.indexAccuracy) ||
+        this.indexAccuracy <= 0 ||
+        this.indexAccuracy > 100)
+    ) {
+      throw new Error("`indexAccuracy` must be an integer between 1 and 100");
+    }
+
+    this.indexParameters = this.validateIndexParameters(config.indexParameters);
+    this.doCreateIndex = config.doCreateIndex ?? true;
+    this.config = config;
+  }
+
+  private validateIndexParameters(
+    parameters?: Record<string, number>,
+  ): Record<string, number> {
+    if (!parameters) return {};
+
+    const allowed = INDEX_PARAMETER_RANGES[this.indexType];
+    const validated: Record<string, number> = {};
+
+    for (const [key, value] of Object.entries(parameters)) {
+      const range = allowed[key];
+      if (!range) {
+        throw new Error(
+          `Unsupported ${this.indexType} index parameter '${key}'. ` +
+            `Allowed: ${Object.keys(allowed).join(", ")}`,
+        );
+      }
+      if (!Number.isInteger(value) || value < range[0] || value > range[1]) {
+        throw new Error(
+          `Index parameter '${key}' must be an integer between ${range[0]} and ${range[1]}`,
+        );
+      }
+      validated[key] = value;
+    }
+
+    return validated;
+  }
+
+  async initialize(): Promise<void> {
+    if (!this._initPromise) {
+      this._initPromise = this._doInitialize().catch(async (error) => {
+        if (this.ownsClient && this.client) {
+          await Promise.resolve(this.client.close()).catch(() => {});
+          this.client = undefined;
+          this.ownsClient = false;
+        }
+        this._initPromise = undefined;
+        throw error;
+      });
+    }
+    return this._initPromise;
+  }
+
+  private async _doInitialize(): Promise<void> {
+    const sdk = await loadPeer(
+      "oracledb",
+      "Oracle AI Vector Search",
+      () => import("oracledb"),
+    );
+    this.oracledb = sdk.default ?? sdk;
+
+    if (this.config.client) {
+      this.client = this.config.client;
+    } else if (this.config.useConnectionPool ?? true) {
+      this.client = await this.oracledb.createPool({
+        poolMin: 1,
+        poolMax: 4,
+        ...this.config.connectionParams,
+      });
+      this.ownsClient = true;
+    } else {
+      this.client = await this.oracledb.getConnection(
+        this.config.connectionParams,
+      );
+      this.ownsClient = true;
+    }
+
+    await this.assertVectorSupport();
+    await this.createCol();
+  }
+
+  private isPool(client: Connection | Pool): client is Pool {
+    return typeof (client as Pool).getConnection === "function";
+  }
+
+  private async withConnection<T>(
+    fn: (connection: Connection) => Promise<T>,
+    commit = false,
+  ): Promise<T> {
+    const client = this.client!;
+
+    if (!this.isPool(client)) {
+      const connection = client as Connection;
+      try {
+        const result = await fn(connection);
+        if (commit) await connection.commit();
+        return result;
+      } catch (err) {
+        await connection.rollback();
+        throw err;
+      }
+    }
+
+    const connection = await client.getConnection();
+    try {
+      const result = await fn(connection);
+      if (commit) await connection.commit();
+      return result;
+    } catch (err) {
+      await connection.rollback();
+      throw err;
+    } finally {
+      await connection.close();
+    }
+  }
+
+  private async assertVectorSupport(): Promise<void> {
+    if (!this.oracledb.thin) {
+      const [major, minor] = [
+        Math.floor(this.oracledb.oracleClientVersion / 100000000),
+        Math.floor(this.oracledb.oracleClientVersion / 100000) % 100,
+      ];
+      if (major < 23 || (major === 23 && minor < 4)) {
+        throw new Error(
+          `Oracle DB client driver version ${this.oracledb.oracleClientVersionString} ` +
+            "not supported, must be >=23.4 for vector support",
+        );
+      }
+    }
+
+    const version = await this.withConnection(
+      async (connection) => connection.oracleServerVersionString,
+    );
+    const [major, minor] = version.split(".").map(Number);
+    if (major < 23 || (major === 23 && minor < 4)) {
+      throw new Error(
+        `Oracle DB version ${version} not supported, must be >=23.4 for vector support`,
+      );
+    }
+  }
+
+  private createIndexDdl(): string {
+    const accuracy = this.indexAccuracy
+      ? `WITH TARGET ACCURACY ${this.indexAccuracy}`
+      : "";
+
+    const parameterEntries = Object.entries(this.indexParameters);
+    const parameters =
+      parameterEntries.length > 0
+        ? `PARAMETERS (${[
+            `type ${this.indexType}`,
+            ...parameterEntries.map(([key, value]) => `${key} ${value}`),
+          ].join(", ")})`
+        : "";
+
+    const organization =
+      this.indexType === "HNSW"
+        ? "INMEMORY NEIGHBOR GRAPH"
+        : "NEIGHBOR PARTITIONS";
+
+    return (
+      `CREATE VECTOR INDEX IF NOT EXISTS ${this.indexName} ON ${this.collectionName} (vector) ` +
+      `ORGANIZATION ${organization} DISTANCE ${this.distanceMetric} ${accuracy} ${parameters}`
+    );
+  }
+
+  private async createCol(): Promise<void> {
+    await this.withConnection(async (connection) => {
+      await connection.execute(`
+        CREATE TABLE IF NOT EXISTS ${this.collectionName} (
+          id VARCHAR2(36) PRIMARY KEY,
+          vector VECTOR(${this.embeddingModelDims}),
+          payload JSON
+        )
+      `);
+
+      await connection.execute(`
+        CREATE TABLE IF NOT EXISTS memory_migrations (
+          id NUMBER PRIMARY KEY,
+          user_id VARCHAR2(255) NOT NULL
+        )
+      `);
+
+      if (this.doCreateIndex) {
+        await connection.execute(this.createIndexDdl());
+      }
+    }, true);
+  }
+
+  private loadPayload(value: any): Record<string, any> {
+    if (value === null || value === undefined) return {};
+    if (typeof value === "string") return JSON.parse(value);
+    if (Buffer.isBuffer(value)) return JSON.parse(value.toString("utf-8"));
+    return value;
+  }
+
+  private vectorBind(vector: number[]) {
+    return {
+      type: this.oracledb.DB_TYPE_VECTOR,
+      val: new Float32Array(vector),
+    };
+  }
+
+  private payloadBind(payload: Record<string, any>) {
+    return { type: this.oracledb.DB_TYPE_JSON, val: payload };
+  }
+
+  async insert(
+    vectors: number[][],
+    ids: string[],
+    payloads: Record<string, any>[],
+  ): Promise<void> {
+    if (ids.length !== vectors.length) {
+      throw new Error("ids and vectors must have the same length");
+    }
+    if (payloads.length !== vectors.length) {
+      throw new Error("payloads and vectors must have the same length");
+    }
+
+    if (vectors.length === 0) return;
+
+    await this.initialize();
+
+    await this.withConnection(async (connection) => {
+      await connection.executeMany(
+        `INSERT INTO ${this.collectionName} (id, vector, payload) VALUES (:id, :vector, :payload)`,
+        vectors.map((vector, i) => ({
+          id: ids[i],
+          vector: new Float32Array(vector),
+          payload: payloads[i] ?? {},
+        })) as BindParameters[],
+        {
+          bindDefs: {
+            id: { type: this.oracledb.DB_TYPE_VARCHAR, maxSize: 36 },
+            vector: { type: this.oracledb.DB_TYPE_VECTOR },
+            payload: { type: this.oracledb.DB_TYPE_JSON },
+          },
+        },
+      );
+    }, true);
+  }
+
+  async search(
+    query: number[],
+    topK: number = 5,
+    filters?: SearchFilters,
+  ): Promise<VectorStoreResult[]> {
+    await this.initialize();
+
+    const [whereClause, filterBinds] = buildWhereClause(filters);
+    const hasFilter = whereClause.length > 0;
+    const selectClause = hasFilter
+      ? "SELECT"
+      : `SELECT /*+ VECTOR_INDEX_TRANSFORM(${this.collectionName}) */`;
+    const sql =
+      `${selectClause} id, payload, VECTOR_DISTANCE(vector, :query_vec, ${this.distanceMetric}) distance ` +
+      `FROM ${this.collectionName} ${whereClause} ORDER BY distance FETCH APPROX FIRST :max_rows ROWS ONLY`;
+
+    const rows = await this.withConnection(async (connection) => {
+      const result = await connection.execute<any[]>(sql, {
+        query_vec: this.vectorBind(query),
+        max_rows: topK,
+        ...filterBinds,
+      });
+      return result.rows ?? [];
+    });
+
+    return rows.map((row) => ({
+      id: row[0],
+      payload: this.loadPayload(row[1]),
+      score: SCORE_FROM_DISTANCE[this.distanceMetric](Number(row[2])),
+    }));
+  }
+
+  async get(vectorId: string): Promise<VectorStoreResult | null> {
+    await this.initialize();
+
+    const rows = await this.withConnection(async (connection) => {
+      const result = await connection.execute<any[]>(
+        `SELECT id, payload FROM ${this.collectionName} WHERE id = :vector_id`,
+        { vector_id: vectorId },
+      );
+      return result.rows ?? [];
+    });
+
+    if (rows.length === 0) return null;
+    return { id: rows[0][0], payload: this.loadPayload(rows[0][1]) };
+  }
+
+  async update(
+    vectorId: string,
+    vector: number[],
+    payload: Record<string, any>,
+  ): Promise<void> {
+    await this.initialize();
+
+    const assignments: string[] = [];
+    const binds: Record<string, any> = { vector_id: vectorId };
+
+    if (vector) {
+      assignments.push("vector = :vector");
+      binds.vector = this.vectorBind(vector);
+    }
+    if (payload) {
+      assignments.push("payload = :payload");
+      binds.payload = this.payloadBind(payload);
+    }
+    if (assignments.length === 0) return;
+
+    await this.withConnection(
+      (connection) =>
+        connection.execute(
+          `UPDATE ${this.collectionName} SET ${assignments.join(", ")} WHERE id = :vector_id`,
+          binds,
+        ),
+      true,
+    );
+  }
+
+  async delete(vectorId: string): Promise<void> {
+    await this.initialize();
+
+    await this.withConnection(
+      (connection) =>
+        connection.execute(
+          `DELETE FROM ${this.collectionName} WHERE id = :vector_id`,
+          { vector_id: vectorId },
+        ),
+      true,
+    );
+  }
+
+  async deleteCol(): Promise<void> {
+    await this.initialize();
+
+    await this.withConnection(
+      (connection) =>
+        connection.execute(`DROP TABLE ${this.collectionName} PURGE`),
+      true,
+    );
+  }
+
+  async list(
+    filters?: SearchFilters,
+    topK: number = 100,
+  ): Promise<[VectorStoreResult[], number]> {
+    await this.initialize();
+
+    const [whereClause, filterBinds] = buildWhereClause(filters);
+
+    return this.withConnection(async (connection) => {
+      const listResult = await connection.execute<any[]>(
+        `SELECT id, payload, COUNT(*) OVER () total FROM ${this.collectionName} ${whereClause} FETCH FIRST :max_rows ROWS ONLY`,
+        { ...filterBinds, max_rows: topK },
+      );
+
+      const rows = listResult.rows ?? [];
+      const results = rows.map((row) => ({
+        id: row[0],
+        payload: this.loadPayload(row[1]),
+      }));
+
+      return [results, Number(rows[0]?.[2] ?? 0)];
+    });
+  }
+
+  async getUserId(): Promise<string> {
+    await this.initialize();
+
+    const rows = await this.withConnection(async (connection) => {
+      const result = await connection.execute<any[]>(
+        "SELECT user_id FROM memory_migrations WHERE id = 1",
+      );
+      return result.rows ?? [];
+    });
+
+    if (rows.length > 0) return rows[0][0];
+
+    const generatedUserId = uuidv4();
+    await this.setUserId(generatedUserId);
+    return generatedUserId;
+  }
+
+  async setUserId(userId: string): Promise<void> {
+    await this.initialize();
+
+    await this.withConnection(async (connection) => {
+      await connection.execute("DELETE FROM memory_migrations WHERE id = 1");
+      await connection.execute(
+        "INSERT INTO memory_migrations (id, user_id) VALUES (1, :user_id)",
+        { user_id: userId },
+      );
+    }, true);
+  }
+
+  async close(): Promise<void> {
+    if (this.client && this.ownsClient) {
+      await this.client.close();
+    }
+  }
+}
@@ -0,0 +1,475 @@
+/// <reference types="jest" />
+/** Oracle AI Vector Search filter, config and SQL tests. The driver is mocked, so no database is needed. */
+const DB_TYPE_VECTOR = { name: "DB_TYPE_VECTOR" };
+const DB_TYPE_JSON = { name: "DB_TYPE_JSON" };
+const DB_TYPE_VARCHAR = { name: "DB_TYPE_VARCHAR" };
+
+const mockCreatePool = jest.fn();
+
+jest.mock(
+  "oracledb",
+  () => ({
+    thin: true,
+    DB_TYPE_VECTOR,
+    DB_TYPE_JSON,
+    DB_TYPE_VARCHAR,
+    createPool: mockCreatePool,
+  }),
+  { virtual: true },
+);
+
+import {
+  OracleAIVectorSearch,
+  buildWhereClause,
+  quoteIdentifier,
+} from "../src/vector_stores/oracledb";
+
+type Call = { sql: string; binds: any; options?: any };
+
+function fakeConnection(calls: Call[], resultsBySql: Array<any[][]>) {
+  let selectIndex = 0;
+  return {
+    oracleServerVersionString: "23.4.0.24.05",
+    async execute(sql: string, binds: any = {}) {
+      calls.push({ sql: sql.replace(/\s+/g, " ").trim(), binds });
+      if (/^\s*SELECT/i.test(sql)) {
+        return { rows: resultsBySql[selectIndex++] ?? [] };
+      }
+      return { rows: [] };
+    },
+    async executeMany(sql: string, binds: any[], options: any = {}) {
+      calls.push({ sql: sql.replace(/\s+/g, " ").trim(), binds, options });
+      return { rows: [] };
+    },
+    async commit() {},
+    async rollback() {},
+    async close() {},
+  };
+}
+
+function makeStore(
+  calls: Call[],
+  results: Array<any[][]> = [],
+  overrides = {},
+) {
+  return new OracleAIVectorSearch({
+    client: fakeConnection(calls, results) as any,
+    collectionName: "mem0",
+    embeddingModelDims: 3,
+    ...overrides,
+  } as any);
+}
+
+describe("quoteIdentifier", () => {
+  it("quotes a bare name", () => {
+    expect(quoteIdentifier("mem0")).toBe('"mem0"');
+  });
+
+  it("quotes each segment of a schema-qualified name", () => {
+    expect(quoteIdentifier("app.mem0")).toBe('"app"."mem0"');
+  });
+
+  it("preserves already-quoted segments", () => {
+    expect(quoteIdentifier('"App"."Mem0"')).toBe('"App"."Mem0"');
+  });
+
+  it("rejects a name that would break out of the quoting", () => {
+    expect(() => quoteIdentifier('mem0" (x); DROP TABLE t--')).toThrow(
+      /is not valid/,
+    );
+  });
+});
+
+describe("buildWhereClause", () => {
+  it("returns no clause for empty filters", () => {
+    expect(buildWhereClause(undefined)).toEqual(["", {}]);
+    expect(buildWhereClause({})).toEqual(["", {}]);
+  });
+
+  it("binds a scalar equality instead of inlining it", () => {
+    const [clause, binds] = buildWhereClause({ user_id: "alice" });
+    expect(clause).toBe(
+      `WHERE JSON_EXISTS(payload, '$."user_id"?(@ == $f_0)' PASSING :f_0 AS "f_0")`,
+    );
+    expect(binds).toEqual({ f_0: "alice" });
+  });
+
+  it("ANDs multiple fields", () => {
+    const [clause, binds] = buildWhereClause({
+      user_id: "alice",
+      agent_id: "bot",
+    });
+    expect(clause.startsWith("WHERE (")).toBe(true);
+    expect(clause).toContain(" AND ");
+    expect(binds).toEqual({ f_0: "alice", f_1: "bot" });
+  });
+
+  it("applies every operator in a compound range filter", () => {
+    const [clause, binds] = buildWhereClause({ age: { gte: 10, lte: 20 } });
+    expect(clause).toContain("@ >= $f_0 && @ <= $f_1");
+    expect(binds).toEqual({ f_0: 10, f_1: 20 });
+  });
+
+  it("builds an existence check for the wildcard filter", () => {
+    const [clause, binds] = buildWhereClause({ user_id: "*" });
+    expect(clause).toBe(`WHERE JSON_EXISTS(payload, '$."user_id"')`);
+    expect(binds).toEqual({});
+  });
+
+  it("builds membership for in and negates it for nin", () => {
+    const [inClause] = buildWhereClause({ user_id: { in: ["a", "b"] } });
+    expect(inClause).toContain("@ in ($f_0, $f_1)");
+    expect(inClause).not.toContain("NOT (");
+
+    const [ninClause] = buildWhereClause({ user_id: { nin: ["a"] } });
+    expect(ninClause).toContain("NOT (");
+  });
+
+  it("lowercases the operand for icontains", () => {
+    const [clause, binds] = buildWhereClause({ data: { icontains: "SciFi" } });
+    expect(clause).toContain("@.lower() has substring $f_0");
+    expect(binds).toEqual({ f_0: "scifi" });
+  });
+
+  it("ORs the branches of a $or group", () => {
+    const [clause, binds] = buildWhereClause({
+      $or: [{ user_id: "alice" }, { agent_id: "bot" }],
+    });
+    expect(clause).toContain(" OR ");
+    expect(binds).toEqual({ f_0: "alice", f_1: "bot" });
+  });
+
+  it("negates a $not group", () => {
+    const [clause] = buildWhereClause({ $not: [{ user_id: "alice" }] });
+    expect(clause.startsWith("WHERE NOT (")).toBe(true);
+  });
+
+  it("nests logical groups", () => {
+    const [clause, binds] = buildWhereClause({
+      user_id: "alice",
+      $or: [{ agent_id: "bot" }, { run_id: "r1" }],
+    });
+    expect(clause).toContain(" AND ");
+    expect(clause).toContain(" OR ");
+    expect(Object.keys(binds)).toEqual(["f_0", "f_1", "f_2"]);
+  });
+
+  it("compares against JSON null without a bind", () => {
+    const [clause, binds] = buildWhereClause({ agent_id: null });
+    expect(clause).toBe(
+      `WHERE JSON_EXISTS(payload, '$."agent_id"?(@ == null)')`,
+    );
+    expect(binds).toEqual({});
+  });
+
+  it("rejects a metadata key that could escape the JSON path", () => {
+    expect(() => buildWhereClause({ 'a"?(1==1))--': "x" })).toThrow(
+      /Invalid metadata key/,
+    );
+  });
+
+  it("rejects an unsupported field operator", () => {
+    expect(() => buildWhereClause({ age: { regex: "^a" } })).toThrow(
+      /Unsupported Oracle filter operator/,
+    );
+  });
+
+  it("rejects an unsupported logical operator", () => {
+    expect(() => buildWhereClause({ $nor: [{ a: 1 }] })).toThrow(
+      /Unsupported Oracle logical filter operator/,
+    );
+  });
+
+  it("rejects an empty in list", () => {
+    expect(() => buildWhereClause({ user_id: { in: [] } })).toThrow(
+      /non-empty array/,
+    );
+  });
+
+  it("rejects a non-scalar comparison operand", () => {
+    expect(() => buildWhereClause({ age: { gt: [1] } })).toThrow(
+      /requires a scalar value/,
+    );
+  });
+});
+
+describe("OracleAIVectorSearch config validation", () => {
+  it("requires connectionParams or client", () => {
+    expect(() => new OracleAIVectorSearch({} as any)).toThrow(
+      /connectionParams.*client/,
+    );
+  });
+
+  it("rejects an unsupported distance metric", () => {
+    expect(() => makeStore([], [], { distanceMetric: "JACCARD" })).toThrow(
+      /Unsupported distance metric/,
+    );
+  });
+
+  it("rejects a non-positive embedding dimension", () => {
+    expect(() => makeStore([], [], { embeddingModelDims: 0 })).toThrow(
+      /positive integer/,
+    );
+  });
+
+  it("rejects an out-of-range index accuracy", () => {
+    expect(() => makeStore([], [], { indexAccuracy: 101 })).toThrow(
+      /between 1 and 100/,
+    );
+  });
+
+  it("rejects an index parameter that does not belong to the index type", () => {
+    expect(() =>
+      makeStore([], [], {
+        indexType: "HNSW",
+        indexParameters: { samples_per_partition: 10 },
+      }),
+    ).toThrow(/Unsupported HNSW index parameter/);
+  });
+
+  it("rejects an index parameter outside its allowed range", () => {
+    expect(() =>
+      makeStore([], [], { indexParameters: { neighbors: 1 } }),
+    ).toThrow(/between 2 and 2048/);
+  });
+});
+
+describe("OracleAIVectorSearch SQL", () => {
+  it("creates the table and a vector index on initialize", async () => {
+    const calls: Call[] = [];
+    await makeStore(calls, [], {
+      indexParameters: { neighbors: 32, efconstruction: 200 },
+      indexAccuracy: 95,
+    }).initialize();
+
+    const ddl = calls.map((c) => c.sql).join("\n");
+    expect(ddl).toContain(
+      'CREATE TABLE IF NOT EXISTS "mem0" ( id VARCHAR2(36) PRIMARY KEY, vector VECTOR(3), payload JSON )',
+    );
+    expect(ddl).toContain(
+      'CREATE VECTOR INDEX IF NOT EXISTS "mem0_VEC_IDX" ON "mem0" (vector) ORGANIZATION INMEMORY NEIGHBOR GRAPH DISTANCE COSINE WITH TARGET ACCURACY 95 PARAMETERS (type HNSW, neighbors 32, efconstruction 200)',
+    );
+  });
+
+  it("skips index creation when doCreateIndex is false", async () => {
+    const calls: Call[] = [];
+    await makeStore(calls, [], { doCreateIndex: false }).initialize();
+    expect(calls.map((c) => c.sql).join("\n")).not.toContain(
+      "CREATE VECTOR INDEX",
+    );
+  });
+
+  it("binds vectors as DB_TYPE_VECTOR and payloads as DB_TYPE_JSON on insert", async () => {
+    const calls: Call[] = [];
+    await makeStore(calls).insert([[1, 2, 3]], ["id-1"], [{ data: "hello" }]);
+
+    const insert = calls.find((c) => c.sql.startsWith("INSERT INTO"))!;
+    expect(insert.binds).toEqual([
+      {
+        id: "id-1",
+        vector: new Float32Array([1, 2, 3]),
+        payload: { data: "hello" },
+      },
+    ]);
+    expect(insert.options.bindDefs.id).toEqual({
+      type: DB_TYPE_VARCHAR,
+      maxSize: 36,
+    });
+    expect(insert.options.bindDefs.vector.type).toBe(DB_TYPE_VECTOR);
+    expect(insert.options.bindDefs.payload.type).toBe(DB_TYPE_JSON);
+  });
+
+  it("issues a single executeMany call with one bind row per vector on a multi-row insert", async () => {
+    const calls: Call[] = [];
+    await makeStore(calls).insert(
+      [
+        [1, 2, 3],
+        [4, 5, 6],
+      ],
+      ["id-1", "id-2"],
+      [{ a: 1 }, { b: 2 }],
+    );
+
+    const inserts = calls.filter((c) => c.sql.startsWith("INSERT INTO"));
+    expect(inserts).toHaveLength(1);
+    expect(inserts[0].binds).toHaveLength(2);
+    expect(inserts[0].binds[0].id).toBe("id-1");
+    expect(inserts[0].binds[1].id).toBe("id-2");
+  });
+
+  it("issues no insert statement when inserting an empty batch", async () => {
+    const calls: Call[] = [];
+    await makeStore(calls).insert([], [], []);
+    expect(calls.filter((c) => c.sql.startsWith("INSERT INTO"))).toHaveLength(
+      0,
+    );
+  });
+
+  it("rejects insert batches whose IDs or payloads do not match vectors", async () => {
+    const store = makeStore([]);
+
+    await expect(store.insert([[1, 2, 3]], [], [{}])).rejects.toThrow(
+      "ids and vectors must have the same length",
+    );
+    await expect(store.insert([[1, 2, 3]], ["id-1"], [])).rejects.toThrow(
+      "payloads and vectors must have the same length",
+    );
+  });
+
+  it("converts cosine distance to a similarity score", async () => {
+    const calls: Call[] = [];
+    const store = makeStore(calls, [[["id-1", { data: "hello" }, 0.25]]]);
+    const results = await store.search([1, 2, 3], 5);
+
+    expect(results).toEqual([
+      { id: "id-1", payload: { data: "hello" }, score: 0.75 },
+    ]);
+    const select = calls.find((c) => c.sql.includes("VECTOR_DISTANCE"))!;
+    expect(select.sql).toContain(
+      "VECTOR_DISTANCE(vector, :query_vec, COSINE) distance",
+    );
+    expect(select.sql).toContain("FETCH APPROX FIRST :max_rows ROWS ONLY");
+    expect(select.binds.max_rows).toBe(5);
+  });
+
+  it("inverts the sign of a DOT distance", async () => {
+    const store = makeStore([], [[["id-1", {}, -0.4]]], {
+      distanceMetric: "DOT",
+    });
+    const [result] = await store.search([1, 2, 3]);
+    expect(result.score).toBeCloseTo(0.4);
+  });
+
+  it("adds the VECTOR_INDEX_TRANSFORM hint when searching without filters", async () => {
+    const calls: Call[] = [];
+    const store = makeStore(calls, [[["id-1", {}, 0.1]]]);
+    await store.search([1, 2, 3], 5);
+
+    const select = calls.find((c) => c.sql.includes("VECTOR_DISTANCE"))!;
+    expect(select.sql).toContain('/*+ VECTOR_INDEX_TRANSFORM("mem0") */');
+  });
+
+  it("omits the VECTOR_INDEX_TRANSFORM hint when searching with filters", async () => {
+    const calls: Call[] = [];
+    const store = makeStore(calls, [[["id-1", {}, 0.1]]]);
+    await store.search([1, 2, 3], 5, { user_id: "alice" });
+
+    const select = calls.find((c) => c.sql.includes("VECTOR_DISTANCE"))!;
+    expect(select.sql).not.toContain("VECTOR_INDEX_TRANSFORM");
+  });
+
+  it("parses a payload returned as a JSON string", async () => {
+    const store = makeStore([], [[["id-1", '{"data":"hello"}']]]);
+    expect(await store.get("id-1")).toEqual({
+      id: "id-1",
+      payload: { data: "hello" },
+    });
+  });
+
+  it("returns null when get finds no row", async () => {
+    expect(await makeStore([], [[]]).get("missing")).toBeNull();
+  });
+
+  it("generates and persists a UUID user id when none is stored", async () => {
+    const calls: Call[] = [];
+    const userId = await makeStore(calls, [[]]).getUserId();
+
+    expect(userId).toMatch(
+      /^[0-9a-f]{8}-[0-9a-f]{4}-4[0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$/,
+    );
+    const insert = calls.find((c) =>
+      c.sql.startsWith("INSERT INTO memory_migrations"),
+    )!;
+    expect(insert.binds).toEqual({ user_id: userId });
+  });
+
+  it("returns the stored user id when one exists", async () => {
+    expect(await makeStore([], [[["alice"]]]).getUserId()).toBe("alice");
+  });
+
+  it("returns rows and the total count from list", async () => {
+    const calls: Call[] = [];
+    const store = makeStore(calls, [[["id-1", { data: "hello" }, 7]]]);
+    const [results, count] = await store.list({ user_id: "alice" }, 10);
+
+    expect(results).toEqual([{ id: "id-1", payload: { data: "hello" } }]);
+    expect(count).toBe(7);
+
+    const selects = calls.filter((c) => /^SELECT/i.test(c.sql));
+    expect(selects).toHaveLength(1);
+    const [list] = selects;
+    expect(list.sql).toContain(
+      "SELECT id, payload, COUNT(*) OVER () total FROM",
+    );
+    expect(list.sql).toContain("WHERE JSON_EXISTS(payload,");
+    expect(list.binds).toEqual({ f_0: "alice", max_rows: 10 });
+  });
+
+  it("returns a total of 0 from list when no rows match", async () => {
+    const store = makeStore([], [[]]);
+    const [results, count] = await store.list();
+    expect(results).toEqual([]);
+    expect(count).toBe(0);
+  });
+});
+
+describe("OracleAIVectorSearch initialization failure", () => {
+  beforeEach(() => mockCreatePool.mockReset());
+
+  function fakePool(serverVersion: string) {
+    const connection = {
+      ...fakeConnection([], []),
+      oracleServerVersionString: serverVersion,
+    };
+    return {
+      close: jest.fn(async () => {}),
+      getConnection: jest.fn(async () => connection),
+    };
+  }
+
+  function makePoolStore() {
+    return new OracleAIVectorSearch({
+      connectionParams: { user: "u", password: "p", connectString: "d" },
+      collectionName: "mem0",
+      embeddingModelDims: 3,
+    } as any);
+  }
+
+  it("closes the pool it owns when initialization fails", async () => {
+    const pool = fakePool("23.3.0.24.05");
+    mockCreatePool.mockResolvedValue(pool);
+
+    await expect(makePoolStore().initialize()).rejects.toThrow(
+      "Oracle DB version 23.3.0.24.05 not supported",
+    );
+    expect(pool.close).toHaveBeenCalledTimes(1);
+  });
+
+  it("does not cache the rejection, so a later attempt can succeed", async () => {
+    const failing = fakePool("23.3.0.24.05");
+    const working = fakePool("23.4.0.24.05");
+    mockCreatePool
+      .mockResolvedValueOnce(failing)
+      .mockResolvedValueOnce(working);
+
+    const store = makePoolStore();
+    await expect(store.initialize()).rejects.toThrow("not supported");
+    await expect(store.initialize()).resolves.toBeUndefined();
+    expect(mockCreatePool).toHaveBeenCalledTimes(2);
+  });
+
+  it("leaves a caller-supplied client open when initialization fails", async () => {
+    const client = fakeConnection([], []);
+    client.oracleServerVersionString = "23.3.0.24.05";
+    const close = jest.spyOn(client, "close");
+
+    const store = new OracleAIVectorSearch({
+      client: client as any,
+      collectionName: "mem0",
+      embeddingModelDims: 3,
+    } as any);
+
+    await expect(store.initialize()).rejects.toThrow("not supported");
+    expect(close).not.toHaveBeenCalled();
+  });
+});
@@ -0,0 +1,152 @@
+/**
+ * Unit tests for Memory.buildSessionScope. The recent-conversation buffer key
+ * derived from user_id/agent_id/run_id must be unique per id combination.
+ */
+/// <reference types="jest" />
+jest.mock("../src/utils/factory", () => {
+  const { MemoryVectorStore } = jest.requireActual(
+    "../src/vector_stores/memory",
+  );
+  const { SQLiteManager } = jest.requireActual("../src/storage/SQLiteManager");
+  const testEmbedding = new Array(1536).fill(0.1);
+  class MockEmbedder {
+    embeddingDims = 1536;
+    async embed(): Promise<number[]> {
+      return testEmbedding;
+    }
+    async embedBatch(texts: string[]): Promise<number[][]> {
+      return texts.map(() => testEmbedding);
+    }
+  }
+  class MockLLM {
+    async generateResponse() {
+      return JSON.stringify({ memory: [] });
+    }
+  }
+  return {
+    __esModule: true,
+    EmbedderFactory: { create: jest.fn(() => new MockEmbedder()) },
+    LLMFactory: { create: jest.fn(() => new MockLLM()) },
+    VectorStoreFactory: {
+      create: jest.fn(
+        () => new MemoryVectorStore({ collectionName: "t", dimension: 1536 }),
+      ),
+    },
+    HistoryManagerFactory: {
+      create: jest.fn(() => new SQLiteManager(":memory:")),
+    },
+  };
+});
+
+import { Memory } from "../src/memory";
+import { SearchFilters } from "../src/types";
+
+function scopeOf(memory: Memory, filters: SearchFilters): string {
+  return (memory as any).buildSessionScope(filters);
+}
+
+function cartesian(values: string[], length: number): string[][] {
+  if (length === 0) return [[]];
+  const rest = cartesian(values, length - 1);
+  const result: string[][] = [];
+  for (const value of values) {
+    for (const tail of rest) {
+      result.push([value, ...tail]);
+    }
+  }
+  return result;
+}
+
+describe("Memory session scope key", () => {
+  let memory: Memory;
+
+  beforeAll(async () => {
+    memory = new Memory();
+    await (memory as any)._initPromise;
+  });
+
+  it("keeps the unchanged key format for ids without delimiter characters", () => {
+    expect(
+      scopeOf(memory, { user_id: "550e8400-e29b-41d4-a716-446655440000" }),
+    ).toBe("user_id=550e8400-e29b-41d4-a716-446655440000");
+    expect(scopeOf(memory, { agent_id: "agent.assistant:v2" })).toBe(
+      "agent_id=agent.assistant:v2",
+    );
+    expect(scopeOf(memory, { run_id: "12345" })).toBe("run_id=12345");
+    expect(
+      scopeOf(memory, { user_id: "user@example.com", agent_id: "support-bot" }),
+    ).toBe("agent_id=support-bot&user_id=user@example.com");
+  });
+
+  it("no longer collides an id embedding the join syntax with the equivalent split filters", () => {
+    const collapsedRun = scopeOf(memory, { run_id: "proj-x&user_id=u1" });
+    const splitRun = scopeOf(memory, { user_id: "u1", run_id: "proj-x" });
+    expect(collapsedRun).not.toBe(splitRun);
+
+    const collapsedAgent = scopeOf(memory, { run_id: "proj-y&agent_id=a1" });
+    const splitAgent = scopeOf(memory, { agent_id: "a1", run_id: "proj-y" });
+    expect(collapsedAgent).not.toBe(splitAgent);
+  });
+
+  it("maps every distinct filter combination of delimiter-heavy ids to a distinct key", () => {
+    const keys: string[] = ["user_id", "agent_id", "run_id"];
+    const values = [
+      "u1",
+      "r1",
+      "a1",
+      "%",
+      "&",
+      "=",
+      "a==",
+      "a1&run_id=r1",
+      "a1&user_id=u1",
+      "r1&user_id=u1",
+      "a1&run_id=r1&user_id=u1",
+    ];
+    const seen = new Map<string, SearchFilters>();
+
+    for (let mask = 1; mask < 1 << keys.length; mask++) {
+      const subset = keys.filter((_key, i) => mask & (1 << i));
+      for (const combo of cartesian(values, subset.length)) {
+        const filters: SearchFilters = {};
+        subset.forEach((key, i) => (filters[key] = combo[i]));
+        const scope = scopeOf(memory, filters);
+        if (seen.has(scope)) {
+          expect(seen.get(scope)).toEqual(filters);
+        } else {
+          seen.set(scope, filters);
+        }
+      }
+    }
+  });
+
+  it("gives ids containing delimiter characters a new key format", () => {
+    expect(scopeOf(memory, { user_id: "dXNlcl9pZDE=" })).toBe(
+      "user_id=dXNlcl9pZDE%3D",
+    );
+    expect(scopeOf(memory, { agent_id: "x&y" })).toBe("agent_id=x%26y");
+    expect(scopeOf(memory, { run_id: "50% off" })).toBe("run_id=50%25 off");
+  });
+
+  it("routes the add pipeline through the builder", async () => {
+    const db = (memory as any).db;
+    const spy = jest.spyOn(db, "getLastMessages");
+    await memory.add([{ role: "user", content: "hello" }], {
+      runId: "proj-x&user_id=u1",
+    });
+    expect(spy).toHaveBeenCalledWith("run_id=proj-x%26user_id%3Du1", 10);
+    spy.mockRestore();
+  });
+
+  it("pins the exact key strings shared with the Python test suite", () => {
+    expect(
+      scopeOf(memory, { user_id: "550e8400-e29b-41d4-a716-446655440000" }),
+    ).toBe("user_id=550e8400-e29b-41d4-a716-446655440000");
+    expect(
+      scopeOf(memory, { user_id: "u1", agent_id: "a1", run_id: "r1" }),
+    ).toBe("agent_id=a1&run_id=r1&user_id=u1");
+    expect(scopeOf(memory, { run_id: "proj-x&user_id=u1" })).toBe(
+      "run_id=proj-x%26user_id%3Du1",
+    );
+  });
+});
@@ -44,6 +44,7 @@ const external = [
  "@elastic/elasticsearch",
  "chromadb",
  "weaviate-client",
+  "oracledb",
 ];

 const define = {
@@ -734,6 +734,7 @@ class MemoryClient:
        memory_depth: Optional[str] = None,
        usecase_setting: Optional[str] = None,
        multilingual: Optional[bool] = None,
+        agent_custom_instructions: Optional[str] = None,
    ) -> Dict[str, Any]:
        """Update the project settings.

@@ -744,6 +745,7 @@ class MemoryClient:
            memory_depth: Memory depth for the project.
            usecase_setting: Usecase setting for the project.
            multilingual: Whether to use the input language for memory storage and retrieval.
+            agent_custom_instructions: New extraction instructions for agent-scoped memories.

        Returns:
            Dictionary containing the API response.
@@ -768,6 +770,7 @@ class MemoryClient:
                    "memory_depth": memory_depth,
                    "usecase_setting": usecase_setting,
                    "multilingual": multilingual,
+                    "agent_custom_instructions": agent_custom_instructions,
                }.items()
                if v is not None
            },
@@ -1636,6 +1639,7 @@ class AsyncMemoryClient:
        memory_depth: Optional[str] = None,
        usecase_setting: Optional[str] = None,
        multilingual: Optional[bool] = None,
+        agent_custom_instructions: Optional[str] = None,
    ) -> Dict[str, Any]:
        """Update the project settings.

@@ -1646,6 +1650,7 @@ class AsyncMemoryClient:
            memory_depth: Memory depth for the project.
            usecase_setting: Usecase setting for the project.
            multilingual: Whether to use the input language for memory storage and retrieval.
+            agent_custom_instructions: New extraction instructions for agent-scoped memories.

        Returns:
            Dictionary containing the API response.
@@ -1670,6 +1675,7 @@ class AsyncMemoryClient:
                    "memory_depth": memory_depth,
                    "usecase_setting": usecase_setting,
                    "multilingual": multilingual,
+                    "agent_custom_instructions": agent_custom_instructions,
                }.items()
                if v is not None
            },
@@ -177,6 +177,7 @@ class BaseProject(ABC):
        self,
        custom_instructions: Optional[str] = None,
        custom_categories: Optional[List[str]] = None,
+        agent_custom_instructions: Optional[str] = None,
    ) -> Dict[str, Any]:
        """
        Update project settings.
@@ -184,6 +185,7 @@ class BaseProject(ABC):
        Args:
            custom_instructions: New instructions for the project
            custom_categories: New categories for the project
+            agent_custom_instructions: New extraction instructions for agent-scoped memories

        Returns:
            Dictionary containing the API response.
@@ -396,6 +398,7 @@ class Project(BaseProject):
        custom_categories: Optional[List[str]] = None,
        multilingual: Optional[bool] = None,
        decay: Optional[bool] = None,
+        agent_custom_instructions: Optional[str] = None,
    ) -> Dict[str, Any]:
        """
        Update project settings.
@@ -407,6 +410,7 @@ class Project(BaseProject):
            decay: Toggle Memory Decay for this project. When True, search-time
                ranking boosts recently-used memories and gently dampens stale ones; when
                False, ranking is restored to the pre-decay behaviour. Off by default.
+            agent_custom_instructions: New extraction instructions for agent-scoped memories

        Returns:
            Dictionary containing the API response.
@@ -418,10 +422,17 @@ class Project(BaseProject):
            NetworkError: If network connectivity issues occur.
            ValueError: If org_id or project_id are not set.
        """
-        if custom_instructions is None and custom_categories is None and multilingual is None and decay is None:
+        if (
+            custom_instructions is None
+            and custom_categories is None
+            and multilingual is None
+            and decay is None
+            and agent_custom_instructions is None
+        ):
            raise ValueError(
                "At least one parameter must be provided for update: "
-                "custom_instructions, custom_categories, multilingual, decay"
+                "custom_instructions, custom_categories, multilingual, decay, "
+                "agent_custom_instructions"
            )

        payload = self._prepare_params(
@@ -430,6 +441,7 @@ class Project(BaseProject):
                "custom_categories": custom_categories,
                "multilingual": multilingual,
                "decay": decay,
+                "agent_custom_instructions": agent_custom_instructions,
            }
        )
        response = self._client.patch(
@@ -445,6 +457,7 @@ class Project(BaseProject):
                "custom_categories": custom_categories,
                "multilingual": multilingual,
                "decay": decay,
+                "agent_custom_instructions": agent_custom_instructions,
                "sync_type": "sync",
            },
        )
@@ -709,6 +722,7 @@ class AsyncProject(BaseProject):
        custom_categories: Optional[List[str]] = None,
        multilingual: Optional[bool] = None,
        decay: Optional[bool] = None,
+        agent_custom_instructions: Optional[str] = None,
    ) -> Dict[str, Any]:
        """
        Update project settings.
@@ -720,6 +734,7 @@ class AsyncProject(BaseProject):
            decay: Toggle Memory Decay for this project. When True, search-time
                ranking boosts recently-used memories and gently dampens stale ones; when
                False, ranking is restored to the pre-decay behaviour. Off by default.
+            agent_custom_instructions: New extraction instructions for agent-scoped memories

        Returns:
            Dictionary containing the API response.
@@ -731,10 +746,17 @@ class AsyncProject(BaseProject):
            NetworkError: If network connectivity issues occur.
            ValueError: If org_id or project_id are not set.
        """
-        if custom_instructions is None and custom_categories is None and multilingual is None and decay is None:
+        if (
+            custom_instructions is None
+            and custom_categories is None
+            and multilingual is None
+            and decay is None
+            and agent_custom_instructions is None
+        ):
            raise ValueError(
                "At least one parameter must be provided for update: "
-                "custom_instructions, custom_categories, multilingual, decay"
+                "custom_instructions, custom_categories, multilingual, decay, "
+                "agent_custom_instructions"
            )

        payload = self._prepare_params(
@@ -743,6 +765,7 @@ class AsyncProject(BaseProject):
                "custom_categories": custom_categories,
                "multilingual": multilingual,
                "decay": decay,
+                "agent_custom_instructions": agent_custom_instructions,
            }
        )
        response = await self._client.patch(
@@ -758,6 +781,7 @@ class AsyncProject(BaseProject):
                "custom_categories": custom_categories,
                "multilingual": multilingual,
                "decay": decay,
+                "agent_custom_instructions": agent_custom_instructions,
                "sync_type": "async",
            },
        )
@@ -28,6 +28,9 @@ class AddMemoryOptions(BaseModel):
        default=None, description="Custom categories for memory classification"
    )
    custom_instructions: Optional[str] = Field(default=None, description="Custom instructions for fact extraction")
+    agent_custom_instructions: Optional[str] = Field(
+        default=None, description="Custom instructions for fact extraction from agent-scoped memories"
+    )
    timestamp: Optional[int] = Field(default=None, description="Unix timestamp for the memory")
    expiration_date: Optional[str] = Field(default=None, description="Expiration date in YYYY-MM-DD format")
    structured_data_schema: Optional[Dict[str, Any]] = Field(
@@ -107,6 +110,9 @@ class ProjectUpdateOptions(BaseModel):
    """Options for project update operations."""

    custom_instructions: Optional[str] = Field(default=None, description="Custom instructions for fact extraction")
+    agent_custom_instructions: Optional[str] = Field(
+        default=None, description="Custom instructions for fact extraction from agent-scoped memories"
+    )
    custom_categories: Optional[List[Dict[str, Any]]] = Field(
        default=None, description="Custom categories for classification"
    )
@@ -1,3 +1,4 @@
+# Copyright (c) 2026, Oracle and/or its affiliates.
 """Pydantic configuration for the Oracle AI Vector Search integration."""

 import re
@@ -91,8 +92,9 @@ class OracleAIVectorSearchConfig(BaseModel):
                exclude_none=True,
            )

-        if self.index_accuracy and not (0 < self.index_accuracy <= 100):
-            raise ValueError("`index_accuracy` must be between 1 and 100")
+        if self.index_accuracy is not None:
+            if not (0 < self.index_accuracy <= 100):
+                raise ValueError("`index_accuracy` must be between 1 and 100")

        if not (0 < self.embedding_model_dims):
            raise ValueError("`embedding_model_dims` must be bigger than 0")
@@ -404,13 +404,18 @@ def _build_filters_and_metadata(
    return base_metadata_template, effective_query_filters


+def _escape_scope_value(val: Any) -> str:
+    """Escape the structural delimiters of the session scope key."""
+    return str(val).replace("%", "%25").replace("&", "%26").replace("=", "%3D")
+
+
 def _build_session_scope(filters):
    """Build deterministic session scope string from entity IDs."""
    parts = []
    for key in sorted(["user_id", "agent_id", "run_id"]):
        val = filters.get(key)
        if val:
-            parts.append(f"{key}={val}")
+            parts.append(f"{key}={_escape_scope_value(val)}")
    return "&".join(parts)


@@ -791,6 +796,11 @@ class Memory(MemoryBase):
                are treated as general conversational/factual memories.
            prompt (str, optional): Prompt to use for the memory creation. Defaults to None.

+        Note:
+            `search()` and `get_all()` scope queries via `filters={"user_id": "...", "agent_id": "...", "run_id": "..."}` —
+            they reject top-level `user_id`/`agent_id`/`run_id` arguments. `add()` accepts them top-level, but passing
+            the same arguments to `search()`/`get_all()` raises a `ValueError`; use the `filters` form there instead.
+

        Returns:
            dict: A dictionary containing the result of the memory addition operation, typically
@@ -2447,6 +2457,12 @@ class AsyncMemory(MemoryBase):
                                         Pass "procedural_memory" to create procedural memories.
            prompt (str, optional): Prompt to use for the memory creation. Defaults to None.
            llm (BaseChatModel, optional): LLM class to use for generating procedural memories. Defaults to None. Useful when user is using LangChain ChatModel.
+
+        Note:
+            `search()` and `get_all()` scope queries via `filters={"user_id": "...", "agent_id": "...", "run_id": "..."}` —
+            they reject top-level `user_id`/`agent_id`/`run_id` arguments. `add()` accepts them top-level, but passing
+            the same arguments to `search()`/`get_all()` raises a `ValueError`; use the `filters` form there instead.
+
        Returns:
            dict: A dictionary containing the result of the memory addition operation.
        """
@@ -1,3 +1,4 @@
+# Copyright (c) 2026, Oracle and/or its affiliates.
 """Oracle AI Vector Search vector store integration for mem0."""

 import array
@@ -245,25 +246,34 @@ class OracleAIVectorSearch(VectorStoreBase):
            self.client = oracledb.connect(**self.config.connection_params)
            self._owns_client = True

-        if not (hasattr(self.client, "thin") and self.client.thin):
-            if oracledb.clientversion()[:2] < (23, 4):
-                raise RuntimeError(
-                    f"Oracle DB client driver version {'.'.join(map(str, oracledb.clientversion()))} "
-                    "not supported, must be >=23.4 for vector support"
+        try:
+            if not (hasattr(self.client, "thin") and self.client.thin):
+                if oracledb.clientversion()[:2] < (23, 4):
+                    raise RuntimeError(
+                        f"Oracle DB client driver version {'.'.join(map(str, oracledb.clientversion()))} "
+                        "not supported, must be >=23.4 for vector support"
+                    )
+
+            if isinstance(self.client, oracledb.Connection):
+                db_version = tuple([int(v) for v in self.client.version.split(".")])
+            else:
+                with self.client.acquire() as conn:
+                    db_version = tuple([int(v) for v in conn.version.split(".")])
+
+            if db_version < (23, 4):
+                raise ValueError(
+                    f"Oracle DB version {'.'.join(map(str, db_version))} not supported, "
+                    "must be >=23.4 for vector support"
                )

-        if isinstance(self.client, oracledb.Connection):
-            db_version = tuple([int(v) for v in self.client.version.split(".")])
-        else:
-            with self.client.acquire() as conn:
-                db_version = tuple([int(v) for v in conn.version.split(".")])
-
-        if db_version < (23, 4):
-            raise ValueError(
-                f"Oracle DB version {'.'.join(map(str, db_version))} not supported, must be >=23.4 for vector support"
-            )
-
-        self.create_col()
+            self.create_col()
+        except Exception:
+            if self._owns_client:
+                try:
+                    self.client.close()
+                except Exception:
+                    pass
+            raise

    @contextmanager
    def _get_cursor(self, commit: bool = False):
@@ -89,6 +89,10 @@ def _build_filter_conditions(filters):
                    raise ValueError(f"Unsupported filter operator: {op}")
                template, is_numeric = OPERATOR_SQL_MAP[op]
                if op in ("in", "nin"):
+                    if not isinstance(op_value, list):
+                        raise ValueError(
+                            f"Filter operator {op!r} for key {key!r} requires a list value, got {type(op_value).__name__}"
+                        )
                    str_list = [str(v) for v in op_value]
                    conditions.append(template)
                    params.extend([key, str_list])
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"

 [project]
 name = "mem0ai"
-version = "2.0.16"
+version = "2.0.18"
 description = "Long-term memory for AI Agents"
 authors = [
    { name = "Mem0", email = "support@mem0.ai" }
@@ -0,0 +1,131 @@
+import itertools
+
+import pytest
+
+from mem0.memory.main import Memory, _build_session_scope, _escape_scope_value
+
+DELIMITER_VALUES = [
+    "u1",
+    "r1",
+    "a1",
+    "%",
+    "&",
+    "=",
+    "%25",
+    "%26",
+    "%3D",
+    "a==",
+    "a1&run_id=r1",
+    "a1&user_id=u1",
+    "r1&user_id=u1",
+    "a1&run_id=r1&user_id=u1",
+]
+
+
+class TestBuildSessionScope:
+    """Tests that _build_session_scope produces a unique key per id combination."""
+
+    def test_ordinary_ids_produce_unchanged_scope_keys(self):
+        """Ids without delimiter characters keep producing the pre-fix key format."""
+        cases = [
+            ({"user_id": "550e8400-e29b-41d4-a716-446655440000"}, "user_id=550e8400-e29b-41d4-a716-446655440000"),
+            ({"agent_id": "agent.assistant:v2"}, "agent_id=agent.assistant:v2"),
+            ({"run_id": "12345"}, "run_id=12345"),
+            (
+                {"user_id": "user@example.com", "agent_id": "support-bot"},
+                "agent_id=support-bot&user_id=user@example.com",
+            ),
+            (
+                {"user_id": "u1", "agent_id": "a1", "run_id": "r1"},
+                "agent_id=a1&run_id=r1&user_id=u1",
+            ),
+        ]
+        for filters, expected in cases:
+            assert _build_session_scope(filters) == expected
+
+    def test_ids_containing_delimiters_do_not_collide(self):
+        """A value that embeds the join syntax no longer maps to the same key as the equivalent split filters."""
+        collapsed_run = {"run_id": "proj-x&user_id=u1"}
+        split_run = {"user_id": "u1", "run_id": "proj-x"}
+        assert _build_session_scope(collapsed_run) != _build_session_scope(split_run)
+        assert _build_session_scope(collapsed_run) == "run_id=proj-x%26user_id%3Du1"
+
+        collapsed_agent = {"run_id": "proj-y&agent_id=a1"}
+        split_agent = {"agent_id": "a1", "run_id": "proj-y"}
+        assert _build_session_scope(collapsed_agent) != _build_session_scope(split_agent)
+
+    def test_ids_containing_delimiters_get_a_new_key_format(self):
+        """Ids holding a delimiter character map to a new key, so their buffer starts empty once after upgrade."""
+        assert _build_session_scope({"user_id": "dXNlcl9pZDE="}) == "user_id=dXNlcl9pZDE%3D"
+        assert _build_session_scope({"agent_id": "x&y"}) == "agent_id=x%26y"
+        assert _build_session_scope({"run_id": "50% off"}) == "run_id=50%25 off"
+
+    def test_scope_keys_are_unique_per_filter_combination(self):
+        """Every distinct filter dict built from delimiter-heavy id values maps to a distinct scope key."""
+        keys = ["user_id", "agent_id", "run_id"]
+        seen = {}
+        for size in range(1, len(keys) + 1):
+            for key_subset in itertools.combinations(keys, size):
+                for combo in itertools.product(DELIMITER_VALUES, repeat=size):
+                    filters = dict(zip(key_subset, combo))
+                    scope = _build_session_scope(filters)
+                    if scope in seen:
+                        assert seen[scope] == filters, f"{seen[scope]} and {filters} both map to {scope!r}"
+                    else:
+                        seen[scope] = filters
+
+
+class TestEscapeScopeValue:
+    """Tests for the low-level per-value escaping helper."""
+
+    def test_non_string_input_is_stringified(self):
+        assert _escape_scope_value(42) == "42"
+
+    def test_percent_is_escaped_before_other_delimiters(self):
+        assert _escape_scope_value("%26") == "%2526"
+        assert _escape_scope_value("%26") != "%26"
+
+    def test_each_delimiter_is_escaped(self):
+        assert _escape_scope_value("%") == "%25"
+        assert _escape_scope_value("&") == "%26"
+        assert _escape_scope_value("=") == "%3D"
+
+
+class TestSessionScopeWiring:
+    """Tests that the add pipeline keys the conversation buffer through the builder."""
+
+    @pytest.fixture
+    def memory(self, mocker):
+        mocker.patch("mem0.memory.main.capture_event")
+        mock_embedder = mocker.MagicMock()
+        mock_embedder.return_value.embed.return_value = [0.1, 0.2, 0.3]
+        mocker.patch("mem0.utils.factory.EmbedderFactory.create", mock_embedder)
+        mock_vector_store = mocker.MagicMock()
+        mock_vector_store.return_value.search.return_value = []
+        mocker.patch(
+            "mem0.utils.factory.VectorStoreFactory.create",
+            side_effect=[mock_vector_store.return_value, mocker.MagicMock()],
+        )
+        mocker.patch("mem0.utils.factory.LlmFactory.create", mocker.MagicMock())
+        mocker.patch("mem0.memory.storage.SQLiteManager", mocker.MagicMock())
+
+        memory = Memory()
+        memory.config = mocker.MagicMock()
+        memory.config.custom_instructions = None
+        memory.custom_instructions = None
+        memory.api_version = "v1.1"
+        memory.db.get_last_messages = mocker.MagicMock(return_value=[])
+        memory.db.save_messages = mocker.MagicMock()
+        memory.llm.generate_response.return_value = '{"memory": []}'
+        return memory
+
+    def test_add_pipeline_uses_the_escaped_key(self, memory):
+        """The pipeline must route through the builder, not assemble the key inline."""
+        memory._add_to_vector_store(
+            messages=[{"role": "user", "content": "hello"}],
+            metadata={},
+            filters={"run_id": "proj-x&user_id=u1"},
+            infer=True,
+        )
+        assert memory.db.get_last_messages.call_args[0][0] == "run_id=proj-x%26user_id%3Du1"
+        assert memory.db.save_messages.call_args[0][1] == "run_id=proj-x%26user_id%3Du1"
@@ -440,3 +440,51 @@ class TestValidateApiKeyHttpError:

        assert not isinstance(exc_info.value, requests.exceptions.JSONDecodeError)
        assert "Error:" in str(exc_info.value)
+
+
+class TestAddAgentCustomInstructions:
+    """Per-request override of the project-level setting, forwarded to the add payload."""
+
+    def _mock_add(self, client):
+        response = MagicMock()
+        response.json.return_value = {"results": []}
+        response.raise_for_status.return_value = None
+        client.client.post.return_value = response
+        return response
+
+    def test_kwarg_reaches_the_add_payload(self, mock_memory_client):
+        self._mock_add(mock_memory_client)
+
+        mock_memory_client.add(
+            "hello",
+            filters={"agent_id": "a1"},
+            agent_custom_instructions="remember tool failures",
+        )
+
+        _, kwargs = mock_memory_client.client.post.call_args
+        assert kwargs["json"]["agent_custom_instructions"] == "remember tool failures"
+
+    def test_typed_option_reaches_the_add_payload(self, mock_memory_client):
+        from mem0.client.types import AddMemoryOptions
+
+        self._mock_add(mock_memory_client)
+
+        mock_memory_client.add(
+            "hello",
+            AddMemoryOptions(
+                filters={"agent_id": "a1"},
+                agent_custom_instructions="remember tool failures",
+            ),
+        )
+
+        _, kwargs = mock_memory_client.client.post.call_args
+        assert kwargs["json"]["agent_custom_instructions"] == "remember tool failures"
+
+    def test_absent_when_not_passed(self, mock_memory_client):
+        """Callers that don't use the feature send an unchanged payload."""
+        self._mock_add(mock_memory_client)
+
+        mock_memory_client.add("hello", filters={"user_id": "u1"})
+
+        _, kwargs = mock_memory_client.client.post.call_args
+        assert "agent_custom_instructions" not in kwargs["json"]
@@ -3,8 +3,8 @@ parameter-passthrough surface.

 Verifies the kwarg → JSON payload mapping for every supported field
 (``custom_instructions``, ``custom_categories``, ``multilingual``,
-``decay``), the ValueError when no field is provided, and the
-URL/method shape. The HTTP layer is mocked.
+``decay``, ``agent_custom_instructions``), the ValueError when no field
+is provided, and the URL/method shape. The HTTP layer is mocked.
 """

 from unittest.mock import MagicMock, patch
@@ -84,6 +84,50 @@ class TestProjectUpdateDecay:
        assert args[0] == "/api/v1/orgs/organizations/org1/projects/proj1/"


+class TestProjectUpdateAgentCustomInstructions:
+    def test_agent_custom_instructions_sent_in_payload(self, project):
+        proj, http = project
+        proj.update(agent_custom_instructions="remember tool failures")
+        assert _patch_payload(http) == {"agent_custom_instructions": "remember tool failures"}
+
+    def test_agent_custom_instructions_alone_satisfies_the_guard(self, project):
+        """It is a standalone field, so setting only it must not raise."""
+        proj, http = project
+        proj.update(agent_custom_instructions="remember tool failures")
+        assert http.patch.called
+
+    def test_empty_string_round_trips_to_clear_the_field(self, project):
+        """An empty string must survive the ``is not None`` filter, not be dropped as falsy."""
+        proj, http = project
+        proj.update(agent_custom_instructions="")
+        assert _patch_payload(http) == {"agent_custom_instructions": ""}
+
+    def test_combined_with_custom_instructions(self, project):
+        """Both sets in one PATCH: the split-instruction configuration."""
+        proj, http = project
+        proj.update(
+            custom_instructions="remember user preferences",
+            agent_custom_instructions="remember tool failures",
+        )
+        assert _patch_payload(http) == {
+            "custom_instructions": "remember user preferences",
+            "agent_custom_instructions": "remember tool failures",
+        }
+
+    def test_omitted_when_none(self, project):
+        """Callers that don't pass it must send an unchanged payload."""
+        proj, http = project
+        proj.update(custom_instructions="be concise")
+        payload = _patch_payload(http)
+        assert payload == {"custom_instructions": "be concise"}
+        assert "agent_custom_instructions" not in payload
+
+    def test_no_args_raises_with_agent_instructions_in_message(self, project):
+        proj, _ = project
+        with pytest.raises(ValueError, match=r"agent_custom_instructions"):
+            proj.update()
+
+
 class TestProjectUpdateBackwardsCompat:
    def test_multilingual_only_still_works(self, project):
        """Pre-decay callers (multilingual only) keep working unchanged."""
@@ -1,3 +1,4 @@
+# Copyright (c) 2026, Oracle and/or its affiliates.
 import os
 import uuid
 from contextlib import nullcontext
@@ -267,6 +268,16 @@ def test_index_parameters_reject_non_string_keys():
        )


+def test_index_accuracy_rejects_zero():
+    with pytest.raises(ValueError, match="index_accuracy.*between 1 and 100"):
+        OracleAIVectorSearchConfig(
+            collection_name=_unique_collection_name(),
+            embedding_model_dims=DIM,
+            client=object(),
+            index_accuracy=0,
+        )
+
+
 def test_index_parameters_canonicalize_int_subclasses():
    class FormattedInt(int):
        def __format__(self, format_spec):
@@ -331,6 +342,27 @@ def test_config_rejects_none_for_non_optional_fields(field, value):
        OracleAIVectorSearchConfig(client=object(), **{field: value})


+def test_init_closes_owned_client_when_post_connect_setup_fails(monkeypatch):
+    fake_connection = MagicMock(spec=oracledb.Connection)
+    fake_connection.thin = True
+    fake_connection.version = "23.4.0.0"
+
+    monkeypatch.setattr(oracledb, "connect", MagicMock(return_value=fake_connection))
+    monkeypatch.setattr(OracleAIVectorSearch, "create_col", MagicMock(side_effect=RuntimeError("boom")))
+    monkeypatch.setattr(OracleAIVectorSearch, "__del__", lambda self: None)
+
+    with pytest.raises(RuntimeError, match="boom"):
+        OracleAIVectorSearch(
+            collection_name=_unique_collection_name(),
+            embedding_model_dims=DIM,
+            connection_params={"user": "u", "password": "p", "dsn": "d"},
+            use_connection_pool=False,
+            do_create_index=False,
+        )
+
+    fake_connection.close.assert_called_once()
+
+
@pytest.mark.parametrize(
    ("metric", "distance", "expected_score"),
    [
@@ -2508,3 +2508,20 @@ class TestBuildFilterConditions(unittest.TestCase):
    def test_numeric_scalar_becomes_string(self):
        conditions, params = _build_filter_conditions({"priority": 42})
        self.assertEqual(params, ["priority", "42"])
+
+    def test_in_rejects_string_value(self):
+        """Passing a string to 'in' would iterate characters and produce a misleading ANY() clause."""
+        with self.assertRaises(ValueError, msg="Expected ValueError for non-list 'in' value"):
+            _build_filter_conditions({"user_id": {"in": "alice"}})
+
+    def test_in_rejects_dict_value(self):
+        with self.assertRaises(ValueError):
+            _build_filter_conditions({"user_id": {"in": {"$gt": 0}}})
+
+    def test_nin_rejects_string_value(self):
+        with self.assertRaises(ValueError):
+            _build_filter_conditions({"user_id": {"nin": "alice"}})
+
+    def test_in_accepts_list_value(self):
+        conditions, params = _build_filter_conditions({"user_id": {"in": ["alice", "bob"]}})
+        self.assertEqual(params, ["user_id", ["alice", "bob"]])
Author	SHA1	Message	Date
Kartik	c427a453a8	docs(changelog): backfill Python v2.0.18 and TypeScript v3.1.6 SDK entries (#6918 )	2026-08-12 00:25:18 +05:30
Kartik	f5b4300449	docs: redirect deprecated OSS graph memory page to platform graph memory (#6914 )	2026-08-12 00:25:02 +05:30
Kartik	71f2ebefa3	fix(memory): escape delimiters when building the session scope key (#6892 )	2026-08-11 23:23:55 +05:30
Hrushikesh Yadav	35a125585e	fix(pgvector): raise ValueError when 'in'/'nin' filter value is not a list (#6879 )	2026-08-11 20:15:07 +05:30
Harsh Vardhan Gupta	4debc58a83	fix(security): patch 8 HIGH + 18 MEDIUM Vanta vulnerabilities across 4 pnpm workspaces (#6847 )	2026-08-07 19:04:33 +05:30
mintlify[bot]	b42cfdd5c8	Fix broken links: unblock link check in oracledb.mdx (#6849 ) Co-authored-by: mintlify[bot] <109931778+mintlify[bot]@users.noreply.github.com>	2026-08-07 18:16:06 +05:30
Elif Sema Balcioglu	6fe6140dba	fix(vector-stores/oracledb): Fix accuracy bug (#6848 )	2026-08-07 18:07:12 +05:30
Diwakar Ray Yadav	b05dc2740f	docs(memory): note filters-based scoping for search/get_all on add() (#6758 )	2026-08-07 14:03:09 +05:30
Kartik	4a0a9a92a6	fix(ts-oss, py): release Oracle client on init failure, validate insert batches (#6839 )	2026-08-06 18:16:25 +05:30
Kartik	beea626f0a	perf(ts-oss): cut Oracle vector store round trips per review feedback (#6835 )	2026-08-06 15:12:14 +05:30
Himanshu	3f39fba28f	fix(n8n): MIT license + themed icons for verified-node vetting (#6804 ) Co-authored-by: kartik-mem0 <kartik.labhshetwar@mem0.ai>	2026-08-06 00:00:13 +05:30
Saket Aryan	12c47f5249	feat(sdk, docs): expose agent_custom_instructions for agent-scoped extraction (#6809 )	2026-08-05 22:12:10 +05:30
Kartik	3f717e5459	docs: frame graph memory as a Platform feature, removed from OSS (#6808 )	2026-08-05 18:17:07 +05:30
Kartik	18021dd106	feat(ts-oss): add Oracle AI Vector Search vector store (#6690 )	2026-08-05 10:55:33 +05:30
pratik	fad0e0e415	fix(ts-sdk): await identity before building project-scoped URLs (#6802 )	2026-08-04 16:31:18 -07:00