eavanvalkenburg · Sep 16, 2026
@@ -11,6 +11,7 @@ This feature ports the vector store abstractions, embedding generator abstractio
 | Vector search | Unified search interface with `search_type` parameter (`"vector"`, `"keyword_hybrid"`) |
 | Data model decorator | `@vectorstoremodel` decorator for defining vector store data models (supports Pydantic, dataclasses, plain classes, dicts) |
 | Agent tools | `create_vector_search_tool`, `create_upsert_tool`, `create_get_tool`, `create_delete_tool` for agent-usable vector store operations |
+| Agent providers | `VectorStoreHistoryProvider` for provider-owned scoped history and `VectorCollectionContextProvider` for caller-owned collections |
 | In-memory store | Zero-dependency vector store for testing and development |
 | 13+ connectors | Azure AI Search, Qdrant, Redis, PostgreSQL, MongoDB, Cosmos DB, Pinecone, Chroma, Weaviate, Oracle, SQL Server, FAISS |
 
@@ -130,6 +131,11 @@ Options considered:
 | `SupportsVectorUpsert` | `_vectors.py` | Protocol for collection CRUD |
 | `SupportsVectorSearch` | `_vectors.py` | Protocol for vector search |
 | `create_vector_search_tool` | `_vectors.py` | Creates AF `FunctionTool` from vector search |
+| `create_upsert_tool` | `_vectors.py` | Creates an agent tool for record upserts |
+| `create_get_tool` | `_vectors.py` | Creates an agent tool for key-based retrieval |
+| `create_delete_tool` | `_vectors.py` | Creates an agent tool for key-based deletion |
+| `VectorStoreHistoryProvider` | `_vectors.py` | Stores scoped full history with optional compacted loading and search |
+| `VectorCollectionContextProvider` | `_vectors.py` | Adds CRUD and one or more search tools for a caller-owned collection |
 
 ## Source Files Reference (SK → AF mapping)
 
@@ -437,7 +443,7 @@ Each connector follows the AF package structure:
 ---
 
 ### Phase 8: Vector Store CRUD Tools
-**Goal:** Provide a full set of agent-usable tools for CRUD operations on vector store collections.
+**Goal:** Provide agent-usable CRUD tools plus generic history and context providers for vector stores.
 **Mergeable:** Yes — adds tools without changing existing APIs.
 
 #### 8.1 — `create_upsert_tool` — tool for upserting records into a collection
@@ -447,6 +453,8 @@ Each connector follows the AF package structure:
 - Consider if this overlaps with filtered search and document when to use which
 #### 8.3 — `create_delete_tool` — tool for deleting records by key
 #### 8.4 — Tests and samples for CRUD tools
+#### 8.5 — `VectorStoreHistoryProvider` — provider-owned scoped history with optional embeddings, compaction, and search
+#### 8.6 — `VectorCollectionContextProvider` — configurable tools and instructions for a caller-owned collection
 
 ---
 
@@ -502,9 +510,18 @@ Each connector follows the AF package structure:
    - `create_upsert_tool(...)` → tool for upserting records
    - `create_get_tool(...)` → tool for retrieving records by key
    - `create_delete_tool(...)` → tool for deleting records
-   - These are separate from search and are placed in a later phase
-
-10. **Score threshold filtering**: Scoring, filter execution, score thresholds, and paging belong to the connector
+   - Upsert and delete require approval by default; get is read-only by default
+
+10. **Generic vector providers**:
+    - `VectorStoreHistoryProvider` takes a store because it owns the history model and collection. Every operation
+      is filtered by application, optional tenant/agent, provider source, and session. Compaction changes only the
+      messages loaded into model context; optional search still queries the full scoped transcript. Content arrays
+      can be stored as JSON text or msgspec-encoded MessagePack bytes to match the backing store.
+    - `VectorCollectionContextProvider` takes a caller-owned collection and adds generated instructions plus
+      configurable CRUD/search tools. Additional `create_vector_search_tool` instances can expose different filters
+      and result mappings.
+
+11. **Score threshold filtering**: Scoring, filter execution, score thresholds, and paging belong to the connector
     and backing store (ref: [SK .NET PR #13501](https://github.com/microsoft/semantic-kernel/pull/13501)). Core passes
     `score_threshold` through without requiring a known distance function or an explicit metric and does not
     post-filter returned results, including results without scores. Each connector defines its score units,

@@ -138,7 +138,9 @@ only to approved first-party endpoints.
 | 17 | `core.session_store` | Agent session store | `agent_framework.SessionStore` / `FileSessionStore` |
 | 18 | `core.agent_hooks` | Agent Hooks middleware | `agent_framework.create_agent_hooks_middleware` |
 | 19 | `core.vector_stores` | Vector store abstractions | `BaseVectorCollection` / `BaseVectorSearch` operations |
-| 20–31 | _reserved_ | core growth | — |
+| 20 | `core.vector_store_history_provider` | Vector-store history provider | `agent_framework.VectorStoreHistoryProvider` |
+| 21 | `core.vector_collection_context_provider` | Vector-collection context provider | `agent_framework.VectorCollectionContextProvider` |
+| 22–31 | _reserved_ | core growth | — |
 | 32 | `orchestration.sequential` | Sequential orchestration | `agent_framework_orchestrations.SequentialBuilder` |
 | 33 | `orchestration.concurrent` | Concurrent orchestration | `agent_framework_orchestrations.ConcurrentBuilder` |
 | 34 | `orchestration.group_chat` | Group-chat orchestration | `agent_framework_orchestrations.GroupChatBuilder` |

@@ -112,6 +112,12 @@ The vector store API is experimental under the shared `VECTOR_STORES` feature ID
   filter execution, score thresholds (including provider-defined/default metrics), and paging. Use native backend
   execution where available, otherwise an explicit connector-local fallback or reject unsupported options
 - **`create_vector_search_tool`** - Creates an agent tool from any `SupportsVectorSearch` implementation
+- **`create_upsert_tool` / `create_get_tool` / `create_delete_tool`** - Create agent tools for collection CRUD;
+  upsert and delete require approval by default, while get does not
+- **`VectorStoreHistoryProvider`** - Stores full scoped conversation history in a provider-owned collection;
+  optional embeddings enable session-scoped history search, while optional compaction affects only loaded context
+- **`VectorCollectionContextProvider`** - Adds instructions and configurable CRUD/search tools for a caller-owned
+  collection, including independently configured additional search tools
 - **`InMemoryCollection` / `InMemoryStore`** - Dependency-free, process-local development and test implementation;
   cosine scoring scales finite inputs, all metrics reject non-finite scores, and unsupported distance functions
   fail before record scanning. Hamming scores/thresholds use the fraction of unequal dimensions, not a count.

@@ -310,8 +310,13 @@
         "SearchType",
         "SupportsVectorSearch",
         "SupportsVectorUpsert",
+        "VectorCollectionContextProvider",
         "VectorStoreCollectionDefinition",
         "VectorStoreField",
+        "VectorStoreHistoryProvider",
+        "create_delete_tool",
+        "create_get_tool",
+        "create_upsert_tool",
         "create_vector_search_tool",
         "register_vectorstoremodel",
         "vectorstoremodel",
@@ -633,8 +638,10 @@
     "UsageDetails",
     "UserInputRequiredException",
     "ValidationTypeEnum",
+    "VectorCollectionContextProvider",
     "VectorStoreCollectionDefinition",
     "VectorStoreField",
+    "VectorStoreHistoryProvider",
     "Workflow",
     "WorkflowAgent",
     "WorkflowBuilder",
@@ -667,8 +674,11 @@
     "create_agent_hooks_middleware_from_emitter",
     "create_always_approve_tool_response",
     "create_always_approve_tool_with_arguments_response",
+    "create_delete_tool",
     "create_edge_runner",
+    "create_get_tool",
     "create_harness_agent",
+    "create_upsert_tool",
     "create_vector_search_tool",
     "detect_media_type_from_base64",
     "enqueue_messages",

@@ -266,8 +266,13 @@ from ._vectors import (
     SearchType,
     SupportsVectorSearch,
     SupportsVectorUpsert,
+    VectorCollectionContextProvider,
     VectorStoreCollectionDefinition,
     VectorStoreField,
+    VectorStoreHistoryProvider,
+    create_delete_tool,
+    create_get_tool,
+    create_upsert_tool,
     create_vector_search_tool,
     register_vectorstoremodel,
     vectorstoremodel,
@@ -588,8 +593,10 @@ __all__ = [
     "UsageDetails",
     "UserInputRequiredException",
     "ValidationTypeEnum",
+    "VectorCollectionContextProvider",
     "VectorStoreCollectionDefinition",
     "VectorStoreField",
+    "VectorStoreHistoryProvider",
     "Workflow",
     "WorkflowAgent",
     "WorkflowBuilder",
@@ -622,8 +629,11 @@ __all__ = [
     "create_agent_hooks_middleware_from_emitter",
     "create_always_approve_tool_response",
     "create_always_approve_tool_with_arguments_response",
+    "create_delete_tool",
     "create_edge_runner",
+    "create_get_tool",
     "create_harness_agent",
+    "create_upsert_tool",
     "create_vector_search_tool",
     "detect_media_type_from_base64",
     "enqueue_messages",

@@ -56,6 +56,8 @@ class FeatureIndex(IntEnum):
     CORE_SESSION_STORE = 17
     CORE_AGENT_HOOKS = 18
     CORE_VECTOR_STORES = 19
+    CORE_VECTOR_STORE_HISTORY_PROVIDER = 20
+    CORE_VECTOR_COLLECTION_CONTEXT_PROVIDER = 21
 
 
 # This environment variable is reserved by the Foundry hosting environment to