fix(memory): expose reasons for skipped empty-URI operations (#4168)

* 增加空uri的原因

* fix  pr bug

* 加入Memory diff

* fix(memory): preserve semantics for skipped URI reporting

Author: yeshion23333

docs/design/traj-exp-experience-learning-redesign.md

@@ -650,6 +650,7 @@ merge 输入/输出日志通过 `tracer.info(..., console=False)` 记录，避
   "trace_id": "...",
   "extracted_at": "...",
   "operations": {...},
+  "skipped_operations": [...],
   "summary": {...}
 }
 ```

docs/en/api/05-sessions.md

@@ -1633,10 +1633,19 @@ When long-term memory extraction runs successfully, the commit writes a `memory_
       }
     ]
   },
+  "skipped_operations": [
+    {
+      "memory_type": "events",
+      "page_id": 101,
+      "reason_code": "invalid_ranges",
+      "reason": "No valid event range could be resolved"
+    }
+  ],
   "summary": {
     "total_adds": 1,
     "total_updates": 1,
-    "total_deletes": 1
+    "total_deletes": 1,
+    "total_skipped": 1
   }
 }
 ```
@@ -1648,11 +1657,13 @@ When long-term memory extraction runs successfully, the commit writes a `memory_
 | `operations.adds` | array | New memories created (`uri`, `memory_type`, `after`) |
 | `operations.updates` | array | Modified memories (`uri`, `memory_type`, `before`, `after`) |
 | `operations.deletes` | array | Deleted memories (`uri`, `memory_type`, `deleted_content`) |
+| `skipped_operations` | array | Intentionally skipped operations and their stable reason codes; these do not represent file changes |
 | `summary.total_adds` | int | Number of new memories |
 | `summary.total_updates` | int | Number of modified memories |
 | `summary.total_deletes` | int | Number of deleted memories |
+| `summary.total_skipped` | int | Number of intentionally skipped operations |
 
-An empty `memory_diff.json` (all counts zero) is written when long-term memory extraction runs but produces no memory operations.
+An empty `memory_diff.json` (all counts zero) is written when long-term memory extraction runs but produces no applied or intentionally skipped operations.
 
 <a id="built-in-memory-types"></a>
 

docs/en/concepts/08-session.md

@@ -197,10 +197,19 @@ Each `session.commit()` writes a `memory_diff.json` to the archive directory, re
       }
     ]
   },
+  "skipped_operations": [
+    {
+      "memory_type": "events",
+      "page_id": 101,
+      "reason_code": "invalid_ranges",
+      "reason": "No valid event range could be resolved"
+    }
+  ],
   "summary": {
     "total_adds": 1,
     "total_updates": 1,
-    "total_deletes": 1
+    "total_deletes": 1,
+    "total_skipped": 1
   }
 }
 ```
@@ -212,9 +221,10 @@ Each `session.commit()` writes a `memory_diff.json` to the archive directory, re
 | `operations.adds` | New memories created (no `before`) |
 | `operations.updates` | Modified memories (with `before` and `after`) |
 | `operations.deletes` | Deleted memories (with `deleted_content`) |
+| `skipped_operations` | Intentionally skipped operations and their stable reason codes; these are not file changes |
 | `summary` | Counts per operation type |
 
-An empty `memory_diff.json` (all counts zero) is written even when no memory operations occurred.
+An empty `memory_diff.json` (all counts zero) is written when no applied or intentionally skipped operations occurred.
 
 ## Storage Structure
 

docs/zh/api/05-sessions.md

@@ -1603,10 +1603,19 @@ viking://user/{user_id}/sessions/{session_id}/
       }
     ]
   },
+  "skipped_operations": [
+    {
+      "memory_type": "events",
+      "page_id": 101,
+      "reason_code": "invalid_ranges",
+      "reason": "无法解析出有效的事件范围"
+    }
+  ],
   "summary": {
     "total_adds": 1,
     "total_updates": 1,
-    "total_deletes": 1
+    "total_deletes": 1,
+    "total_skipped": 1
   }
 }
 ```
@@ -1618,11 +1627,13 @@ viking://user/{user_id}/sessions/{session_id}/
 | `operations.adds` | array | 新增记忆（`uri`、`memory_type`、`after`） |
 | `operations.updates` | array | 修改记忆（`uri`、`memory_type`、`before`、`after`） |
 | `operations.deletes` | array | 删除记忆（`uri`、`memory_type`、`deleted_content`） |
+| `skipped_operations` | array | 策略性跳过的操作及稳定原因码；不代表文件变更 |
 | `summary.total_adds` | int | 新增记忆数 |
 | `summary.total_updates` | int | 修改记忆数 |
 | `summary.total_deletes` | int | 删除记忆数 |
+| `summary.total_skipped` | int | 策略性跳过的操作数 |
 
-如果长记忆抽取已运行但没有产生记忆操作，也会写入空结构的 `memory_diff.json`（所有计数为零）。
+如果长记忆抽取已运行但没有产生实际变更或策略性跳过，也会写入空结构的 `memory_diff.json`（所有计数为零）。
 
 <a id="内置记忆类型"></a>
 

docs/zh/concepts/08-session.md

@@ -197,10 +197,19 @@ LLM 去重决策 → candidate(skip/create/none) + item(merge/delete)
       }
     ]
   },
+  "skipped_operations": [
+    {
+      "memory_type": "events",
+      "page_id": 101,
+      "reason_code": "invalid_ranges",
+      "reason": "无法解析出有效的事件范围"
+    }
+  ],
   "summary": {
     "total_adds": 1,
     "total_updates": 1,
-    "total_deletes": 1
+    "total_deletes": 1,
+    "total_skipped": 1
   }
 }
 ```
@@ -212,9 +221,10 @@ LLM 去重决策 → candidate(skip/create/none) + item(merge/delete)
 | `operations.adds` | 新增的记忆（无 `before`） |
 | `operations.updates` | 修改的记忆（含 `before` 和 `after`） |
 | `operations.deletes` | 删除的记忆（含 `deleted_content`） |
+| `skipped_operations` | 策略性跳过的操作及稳定原因码；不代表文件变更 |
 | `summary` | 各操作类型的计数 |
 
-即使没有记忆操作，也会写入空结构的 `memory_diff.json`（所有计数为零）。
+如果没有实际变更或策略性跳过，也会写入空结构的 `memory_diff.json`（所有计数为零）。
 
 ## 存储结构
 

openviking/session/compressor_v3.py

@@ -346,6 +346,9 @@ async def _build_memory_diff(
             adds=adds,
             updates=updates,
             deletes=deletes,
+            skipped_operations=_serialize_skipped_operations(
+                getattr(result, "skipped_operations", [])
+            ),
         )
 
     @tracer(ignore_result=True)
@@ -363,6 +366,7 @@ async def extract_long_term_memories(
         allow_self_memory: bool = True,
         allowed_peer_ids: Optional[set[str]] = None,
         event_search_tags: Optional[List[str]] = None,
+        peer_memory_enabled: bool = True,
     ):
         if not agent_evolution_enabled:
             effective_types = (
@@ -400,6 +404,7 @@ async def extract_long_term_memories(
                 archive_uri=archive_uri,
                 allowed_memory_types=allowed_memory_types,
                 allow_self_memory=allow_self_memory,
+                peer_memory_enabled=peer_memory_enabled,
                 allowed_peer_ids=allowed_peer_ids,
                 event_search_tags=event_search_tags,
             )
@@ -601,6 +606,7 @@ async def _extract_user_memories(
         archive_uri: Optional[str] = None,
         allowed_memory_types: Optional[set[str]] = None,
         allow_self_memory: bool = True,
+        peer_memory_enabled: bool = True,
         allowed_peer_ids: Optional[set[str]] = None,
         event_search_tags: Optional[List[str]] = None,
     ) -> "_V3ExtractionResult":
@@ -642,6 +648,7 @@ async def _extract_user_memories(
             allowed_memory_types=allowed_memory_types,
             allow_self=allow_self_memory,
             allowed_peer_ids=allowed_peer_ids,
+            peer_memory_enabled=peer_memory_enabled,
         )
         isolation_handler.prepare_messages()
         context_provider._isolation_handler = isolation_handler
@@ -682,6 +689,7 @@ async def _extract_user_memories(
                     "allowed_memory_types": allowed_memory_types,
                     "allow_self": allow_self_memory,
                     "allowed_peer_ids": allowed_peer_ids,
+                    "peer_memory_enabled": peer_memory_enabled,
                 },
                 metadata={
                     "source_extraction_id": extraction_id,
@@ -719,6 +727,9 @@ async def _extract_user_memories(
             cases=canonical_cases,
             memory_diff=memory_diff,
             case_uri_by_name=_case_uri_by_name(canonical_cases, patch_operations, result),
+            skipped_operations=_serialize_skipped_operations(
+                getattr(result, "skipped_operations", [])
+            ),
         )
 
     def _session_skill_extraction_enabled(self) -> bool:
@@ -1223,6 +1234,7 @@ class _V3ExtractionResult:
     cases: list[Case] = field(default_factory=list)
     memory_diff: dict[str, Any] | None = None
     case_uri_by_name: dict[str, str] = field(default_factory=dict)
+    skipped_operations: list[dict[str, Any]] = field(default_factory=list)
 
 
 @dataclass(slots=True)
@@ -2041,6 +2053,22 @@ def _same_memory_file(before: Optional[MemoryFile], after: Optional[MemoryFile])
     )
 
 
+def _serialize_skipped_operations(items: Any) -> list[dict[str, Any]]:
+    serialized: list[dict[str, Any]] = []
+    for item in list(items or []):
+        if isinstance(item, dict):
+            payload = dict(item)
+        else:
+            model_dump = getattr(item, "model_dump", None)
+            if not callable(model_dump):
+                continue
+            payload = model_dump(mode="json", exclude_none=True)
+        if isinstance(payload, dict):
+            payload.pop("source", None)
+            serialized.append(payload)
+    return serialized
+
+
 def _v3_extraction_response(
     *,
     contexts: list[Context],
@@ -2051,10 +2079,9 @@ def _v3_extraction_response(
 
     Historically ``extract_long_term_memories`` returned ``list[Context]`` and
     a number of direct callers still index/compare the return value as a list.
-    Commit orchestration now also understands the execution-memory style
-    ``{"contexts": ..., "session_skills": ...}`` shape so it can count
-    session skills.  Preserve the old list shape unless there are actual
-    session skills to report.
+    Commit orchestration also understands a structured response for session
+    skills. Preserve the old list shape unless there are actual session skills
+    to report.
     """
     skill_dicts: list[dict[str, Any]] = []
     seen: set[str] = set()
@@ -2075,7 +2102,9 @@ def _make_memory_diff(
     adds: list[dict[str, Any]],
     updates: list[dict[str, Any]],
     deletes: list[dict[str, Any]],
+    skipped_operations: Optional[list[dict[str, Any]]] = None,
 ) -> dict[str, Any]:
+    skipped = list(skipped_operations or [])
     return {
         "archive_uri": archive_uri,
         "trace_id": tracer.get_trace_id() or None,
@@ -2085,10 +2114,12 @@ def _make_memory_diff(
             "updates": list(updates),
             "deletes": list(deletes),
         },
+        "skipped_operations": skipped,
         "summary": {
             "total_adds": len(adds),
             "total_updates": len(updates),
             "total_deletes": len(deletes),
+            "total_skipped": len(skipped),
         },
     }
 
@@ -2101,12 +2132,16 @@ def _merge_memory_diffs(
     adds: list[dict[str, Any]] = []
     updates: list[dict[str, Any]] = []
     deletes: list[dict[str, Any]] = []
+    skipped_operations: list[dict[str, Any]] = []
     trace_id = tracer.get_trace_id() or None
     for diff in diffs:
         if not isinstance(diff, dict):
             continue
         if trace_id is None and diff.get("trace_id"):
             trace_id = str(diff.get("trace_id"))
+        skipped_operations.extend(
+            item for item in diff.get("skipped_operations", []) if isinstance(item, dict)
+        )
         operations = diff.get("operations")
         if not isinstance(operations, dict):
             continue
@@ -2118,6 +2153,7 @@ def _merge_memory_diffs(
         adds=adds,
         updates=updates,
         deletes=deletes,
+        skipped_operations=skipped_operations,
     )
     merged["trace_id"] = trace_id
     return merged
@@ -2130,7 +2166,8 @@ def _memory_diff_has_changes(diff: Any) -> bool:
     if not isinstance(summary, dict):
         return False
     return any(
-        int(summary.get(key) or 0) > 0 for key in ("total_adds", "total_updates", "total_deletes")
+        int(summary.get(key) or 0) > 0
+        for key in ("total_adds", "total_updates", "total_deletes", "total_skipped")
     )
 
 

openviking/session/memory/dataclass.py

@@ -169,6 +169,39 @@ class MemoryOperationSource(BaseModel):
     extracted_at: Optional[str] = None
 
 
+class MemoryOperationSkipCode(str, Enum):
+    """Stable reason codes for intentionally skipped memory operations."""
+
+    MEMORY_TYPE_FILTERED = "memory_type_filtered"
+    SELF_MEMORY_DISABLED = "self_memory_disabled"
+    PEER_MEMORY_DISABLED = "peer_memory_disabled"
+    INVALID_PEER_ID = "invalid_peer_id"
+    PEER_NOT_ALLOWED = "peer_not_allowed"
+    INVALID_RANGES = "invalid_ranges"
+    AMBIGUOUS_TARGET = "ambiguous_target"
+    NO_WRITABLE_TARGET = "no_writable_target"
+
+
+class MemoryOperationSkip(BaseModel):
+    """Internal policy/validation decision explaining why no URI was produced."""
+
+    reason_code: MemoryOperationSkipCode
+    reason: str
+
+
+class SkippedMemoryOperation(BaseModel):
+    """Structured, task-visible record for one intentionally skipped operation."""
+
+    memory_type: str
+    page_id: Optional[int] = None
+    uri: Optional[str] = None
+    reason_code: MemoryOperationSkipCode
+    reason: str
+    # Source is used only to scope shared streaming-batch results back to the
+    # submitting commit. It must never be serialized into the public task result.
+    source: Optional[MemoryOperationSource] = Field(default=None, exclude=True)
+
+
 # ============================================================================
 # Memory Field and Schema Definitions
 # ============================================================================
@@ -298,6 +331,9 @@ class ResolvedOperation(BaseModel):
     uris: List[str]
     page_id: Optional[int] = None  # Temporary page_id for link resolution (not persisted)
     source: Optional[MemoryOperationSource] = None
+    # Runtime-only resolution decision. It is deliberately excluded from model
+    # serialization so it cannot enter later LLM merge prompts or memory files.
+    resolution_skip: Optional[MemoryOperationSkip] = Field(default=None, exclude=True)
     # Custom scalar tags (already normalized as "key=value") to attach to this
     # operation's memories in the vector index. None means "no tags"; used by
     # event-memory auto-tagging. Not persisted in the memory file content.