Add User Model Service and Temporal Recall for MeMind V0.1.

Introduce UMS ingest/snapshot pipeline, Context Planner with multi-source recall, runtime context injection, canonical user mapping, and session snapshot loading on auth/me. Co-authored-by: Cursor <cursoragent@cursor.com>
2026-09-03 23:25:18 +08:00
parent bf66fdf493
commit 212ff3ff80
47 changed files with 4534 additions and 6 deletions
@@ -0,0 +1,175 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://memind.local/schemas/context-plan.v1.schema.json",
+  "title": "ContextPlan",
+  "description": "Context Planner 输出 — 决定需要哪些上下文及多源检索计划",
+  "type": "object",
+  "required": ["query_type", "time", "temporal_mode", "targets", "sources", "retrievals", "context_needs"],
+  "additionalProperties": false,
+  "properties": {
+    "query_type": {
+      "type": "string",
+      "enum": [
+        "personal_temporal_recall",
+        "memory_lookup",
+        "profile_only",
+        "general_chat",
+        "task_execution"
+      ]
+    },
+    "time": {
+      "type": "object",
+      "required": ["mention_range"],
+      "additionalProperties": false,
+      "properties": {
+        "mention_range": {
+          "$ref": "#/$defs/timeRange",
+          "description": "按 observed_time / 提及时间过滤"
+        },
+        "event_range": {
+          "oneOf": [{ "$ref": "#/$defs/timeRange" }, { "type": "null" }],
+          "description": "按 event_time / 计划发生时间过滤"
+        },
+        "relative_label": {
+          "type": ["string", "null"],
+          "enum": [
+            "today",
+            "yesterday",
+            "day_before_yesterday",
+            "this_week",
+            "last_week",
+            "this_month",
+            "last_month",
+            "recent_days",
+            null
+          ]
+        }
+      }
+    },
+    "temporal_mode": {
+      "type": "string",
+      "enum": ["OCCURRED_IN", "MENTIONED_IN", "PLANNED_IN", "CREATED_IN", "DUE_IN", "AMBIGUOUS"]
+    },
+    "targets": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "required": ["type", "weight"],
+        "additionalProperties": false,
+        "properties": {
+          "type": {
+            "type": "string",
+            "enum": [
+              "calendar_event",
+              "event",
+              "schedule",
+              "todo",
+              "commitment",
+              "decision",
+              "mention",
+              "important_event"
+            ]
+          },
+          "weight": { "type": "number", "minimum": 0, "maximum": 1 }
+        }
+      }
+    },
+    "sources": {
+      "type": "object",
+      "additionalProperties": { "type": "number", "minimum": 0, "maximum": 1 },
+      "propertyNames": {
+        "enum": ["calendar", "chat", "meinput", "memory_v2", "email", "browser", "tasks", "git"]
+      }
+    },
+    "retrievals": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": ["source", "query_type", "weight"],
+        "additionalProperties": false,
+        "properties": {
+          "source": {
+            "type": "string",
+            "enum": ["calendar", "chat", "meinput", "memory_v2"]
+          },
+          "query_type": {
+            "type": "string",
+            "description": "源内检索策略，如 important_mentions | commitments_and_decisions | events"
+          },
+          "expanded_queries": {
+            "type": "array",
+            "items": { "type": "string", "maxLength": 128 }
+          },
+          "weight": { "type": "number", "minimum": 0, "maximum": 1 },
+          "skip_below": {
+            "type": "number",
+            "minimum": 0,
+            "maximum": 1,
+            "default": 0.3,
+            "description": "低于此权重的 source 跳过检索"
+          }
+        }
+      }
+    },
+    "filters": {
+      "type": "object",
+      "additionalProperties": false,
+      "properties": {
+        "importance_min": { "type": "number", "minimum": 0, "maximum": 1 },
+        "status": {
+          "type": "string",
+          "enum": ["any", "unresolved", "planned", "completed"]
+        },
+        "participants": {
+          "type": "array",
+          "items": { "type": "string" }
+        }
+      }
+    },
+    "output": {
+      "type": "object",
+      "additionalProperties": false,
+      "properties": {
+        "group_by": {
+          "type": "string",
+          "enum": ["importance", "time", "source", "type"]
+        },
+        "dedupe": { "type": "boolean", "default": true },
+        "timeline": { "type": "boolean", "default": true },
+        "wide_recall": { "type": "boolean", "default": true },
+        "max_items": { "type": "integer", "minimum": 1, "maximum": 200, "default": 50 }
+      }
+    },
+    "context_needs": {
+      "type": "object",
+      "required": ["user_snapshot", "temporal_recall", "memory_retrieval"],
+      "additionalProperties": false,
+      "properties": {
+        "user_snapshot": { "type": "string", "enum": ["REQUIRED", "OPTIONAL", "SKIP"] },
+        "temporal_recall": { "type": "string", "enum": ["REQUIRED", "OPTIONAL", "SKIP"] },
+        "memory_retrieval": { "type": "string", "enum": ["REQUIRED", "OPTIONAL", "SKIP"] }
+      }
+    },
+    "planner_meta": {
+      "type": "object",
+      "additionalProperties": false,
+      "properties": {
+        "level": { "type": "string", "enum": ["cheap_router", "semantic_planner"] },
+        "rule_hits": { "type": "array", "items": { "type": "string" } },
+        "confidence": { "type": "number", "minimum": 0, "maximum": 1 }
+      }
+    }
+  },
+  "$defs": {
+    "timeRange": {
+      "type": "object",
+      "required": ["start", "end"],
+      "additionalProperties": false,
+      "properties": {
+        "start": { "type": "string", "format": "date-time" },
+        "end": { "type": "string", "format": "date-time" }
+      }
+    }
+  }
+}
@@ -0,0 +1,146 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://memind.local/schemas/evidence-envelope.v1.schema.json",
+  "title": "EvidenceEnvelope",
+  "description": "通用 Evidence Contract v1。所有 Source（MeInput、Chat、Calendar、Browser、Git…）统一产出此结构；UMS 只消费 Envelope，不依赖各 Source 内部表结构。",
+  "type": "object",
+  "required": [
+    "evidence_id",
+    "user_id",
+    "source_type",
+    "source_ref",
+    "occurred_at",
+    "evidence_type",
+    "payload",
+    "content_hash",
+    "schema_version"
+  ],
+  "properties": {
+    "evidence_id": {
+      "type": "string",
+      "format": "uuid",
+      "description": "Evidence 全局唯一 ID（由 Source 或 Export 层生成）"
+    },
+    "user_id": {
+      "type": "string",
+      "format": "uuid",
+      "description": "MeMind h5_users.id"
+    },
+    "source_type": {
+      "type": "string",
+      "enum": ["meinput", "chat", "calendar", "browser", "git", "mail", "filesystem", "voice", "location"],
+      "description": "Evidence Source 类型，与 MeMind Source 体系对齐"
+    },
+    "source_ref": {
+      "type": "string",
+      "minLength": 1,
+      "maxLength": 512,
+      "description": "Source 内稳定引用，如 segment_id / message_id / commit_sha"
+    },
+    "occurred_at": {
+      "type": "string",
+      "format": "date-time",
+      "description": "Evidence 发生时间（用户行为时间，非入库时间）"
+    },
+    "received_at": {
+      "type": "string",
+      "format": "date-time",
+      "description": "Source 侧接收/导出时间，可选"
+    },
+    "evidence_type": {
+      "type": "string",
+      "enum": [
+        "expression_segment",
+        "expression_commit",
+        "expression_session",
+        "chat_message",
+        "calendar_event",
+        "browser_visit",
+        "git_commit",
+        "aggregate_daily",
+        "timeline_report"
+      ],
+      "description": "Evidence 语义类型，决定 payload 子 schema"
+    },
+    "payload": {
+      "type": "object",
+      "description": "类型化载荷，见 $defs"
+    },
+    "content_hash": {
+      "type": "string",
+      "pattern": "^[a-f0-9]{64}$",
+      "description": "SHA-256(canonical_json(payload)+evidence_type+source_ref)，用于幂等与去重"
+    },
+    "schema_version": {
+      "type": "integer",
+      "const": 1
+    },
+    "privacy_level": {
+      "type": "string",
+      "enum": ["normal", "restricted", "secure_skip"],
+      "default": "normal",
+      "description": "secure_skip 的 Evidence 不应离开 Source；Export 层必须过滤"
+    },
+    "locale": {
+      "type": "string",
+      "description": "如 zh-CN，可选"
+    }
+  },
+  "additionalProperties": false,
+  "$defs": {
+    "ExpressionSegmentPayload": {
+      "type": "object",
+      "required": ["text", "start_at", "end_at", "event_ids"],
+      "properties": {
+        "text": { "type": "string", "minLength": 1 },
+        "start_at": { "type": "string", "format": "date-time" },
+        "end_at": { "type": "string", "format": "date-time" },
+        "event_ids": { "type": "array", "items": { "type": "string", "format": "uuid" }, "minItems": 1 },
+        "event_count": { "type": "integer", "minimum": 1 },
+        "context": {
+          "type": "object",
+          "properties": {
+            "app": { "type": ["string", "null"] },
+            "app_bundle_id": { "type": ["string", "null"] },
+            "device_id": { "type": "string" },
+            "scene": { "type": "string" },
+            "input_method": { "type": ["string", "null"] },
+            "language": { "type": ["string", "null"] }
+          }
+        }
+      },
+      "additionalProperties": false
+    },
+    "AggregateDailyPayload": {
+      "type": "object",
+      "required": ["day", "stats"],
+      "properties": {
+        "day": { "type": "string", "format": "date" },
+        "stats": {
+          "type": "object",
+          "properties": {
+            "events": { "type": "integer", "minimum": 0 },
+            "segments": { "type": "integer", "minimum": 0 },
+            "chars": { "type": "integer", "minimum": 0 },
+            "peak_hour": { "type": "integer", "minimum": 0, "maximum": 23 },
+            "top_apps": { "type": "array", "items": { "type": "string" } },
+            "top_terms": { "type": "array", "items": { "type": "string" } }
+          }
+        }
+      },
+      "additionalProperties": false
+    },
+    "TimelineReportPayload": {
+      "type": "object",
+      "required": ["report_type", "period_key", "edition", "status", "content"],
+      "properties": {
+        "report_type": { "type": "string", "enum": ["daily_close", "weekly_review"] },
+        "period_key": { "type": "string" },
+        "edition": { "type": "integer", "minimum": 1 },
+        "status": { "type": "string", "enum": ["published", "revoked"] },
+        "content": { "type": "object" }
+      },
+      "additionalProperties": false
+    }
+  }
+}
@@ -0,0 +1,187 @@
+-- memind_user_model v0 — User Model Service (UMS)
+-- 库：memind_user_model（与 meinput、goose 并列，独立 RDS 库或 schema）
+-- UMS 只 ingest Evidence Envelope v1，不读 meinput.mi_* 表。
+
+-- ── L0: 原始 Evidence（immutable append）────────────────────────────────────
+
+CREATE TABLE IF NOT EXISTS um_evidence (
+  evidence_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  source_type VARCHAR(32) NOT NULL,
+  source_ref VARCHAR(512) NOT NULL,
+  evidence_type VARCHAR(64) NOT NULL,
+  occurred_at DATETIME(3) NOT NULL,
+  received_at DATETIME(3) NOT NULL DEFAULT CURRENT_TIMESTAMP(3),
+  content_hash CHAR(64) NOT NULL,
+  schema_version INT NOT NULL DEFAULT 1,
+  privacy_level VARCHAR(32) NOT NULL DEFAULT 'normal',
+  payload_json JSON NOT NULL,
+  UNIQUE KEY uq_um_evidence_content (user_id, content_hash),
+  KEY idx_um_evidence_user_time (user_id, occurred_at DESC),
+  KEY idx_um_evidence_source (user_id, source_type, occurred_at DESC)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── L1: Signal（观察到了什么 — 确定性，可复算）────────────────────────────
+
+CREATE TABLE IF NOT EXISTS um_signals (
+  signal_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  signal_type VARCHAR(64) NOT NULL,
+  -- 例：term_frequency | app_usage | entity_mention | rhythm_hour | segment_count
+  dimension_key VARCHAR(128) NOT NULL,
+  -- 例：term:MeInput | app:com.tencent.xin | hour:15
+  window_start DATETIME(3) NOT NULL,
+  window_end DATETIME(3) NOT NULL,
+  value_json JSON NOT NULL,
+  -- 例：{"count":126,"chars":8400,"distinct_days":7}
+  evidence_ids JSON NOT NULL,
+  -- 指向 um_evidence.evidence_id 数组
+  computed_at BIGINT NOT NULL,
+  content_hash CHAR(64) NOT NULL,
+  UNIQUE KEY uq_um_signal (user_id, signal_type, dimension_key, window_start, window_end, content_hash),
+  KEY idx_um_signals_user_window (user_id, window_end DESC)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── L2: Candidate（系统推断了什么 — hypothesis）────────────────────────────
+
+CREATE TABLE IF NOT EXISTS um_candidates (
+  candidate_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  candidate_type VARCHAR(64) NOT NULL,
+  -- project | interest | preference | relation | routine | focus
+  hypothesis_json JSON NOT NULL,
+  -- 例：{"type":"project","name":"MeInput","role":"active"}
+  status ENUM(
+    'observed',
+    'open',
+    'accepted',
+    'rejected',
+    'conflicted',
+    'revoked',
+    'memory_eligible',
+    'memory_promoted'
+  ) NOT NULL DEFAULT 'observed',
+  promotion_score DECIMAL(6,4) NOT NULL DEFAULT 0,
+  confidence DECIMAL(6,4) NOT NULL DEFAULT 0,
+  signal_ids JSON NOT NULL,
+  evidence_ids JSON NOT NULL,
+  first_seen_at DATETIME(3) NOT NULL,
+  last_seen_at DATETIME(3) NOT NULL,
+  version INT NOT NULL DEFAULT 1,
+  parent_candidate_id CHAR(36) NULL,
+  conflict_with_id CHAR(36) NULL,
+  created_at BIGINT NOT NULL,
+  updated_at BIGINT NOT NULL,
+  KEY idx_um_candidates_user_status (user_id, status, last_seen_at DESC),
+  KEY idx_um_candidates_user_type (user_id, candidate_type, promotion_score DESC)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── L3: User Model Graph ───────────────────────────────────────────────────
+
+CREATE TABLE IF NOT EXISTS um_entities (
+  entity_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  entity_type VARCHAR(32) NOT NULL,
+  -- person | project | topic | tool | preference_dim | role
+  canonical_name VARCHAR(256) NOT NULL,
+  aliases_json JSON NULL,
+  status ENUM('active', 'archived', 'revoked') NOT NULL DEFAULT 'active',
+  created_at BIGINT NOT NULL,
+  updated_at BIGINT NOT NULL,
+  UNIQUE KEY uq_um_entity (user_id, entity_type, canonical_name(128))
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+CREATE TABLE IF NOT EXISTS um_attributes (
+  attribute_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  entity_id CHAR(36) NULL,
+  attr_key VARCHAR(128) NOT NULL,
+  value_json JSON NOT NULL,
+  confidence DECIMAL(6,4) NOT NULL,
+  decay_halflife_days INT NOT NULL DEFAULT 90,
+  effective_weight DECIMAL(6,4) NOT NULL DEFAULT 1,
+  evidence_ids JSON NOT NULL,
+  candidate_id CHAR(36) NULL,
+  first_seen_at DATETIME(3) NOT NULL,
+  last_seen_at DATETIME(3) NOT NULL,
+  status ENUM('active', 'superseded', 'revoked') NOT NULL DEFAULT 'active',
+  version INT NOT NULL DEFAULT 1,
+  content_hash CHAR(64) NOT NULL,
+  KEY idx_um_attr_user_key (user_id, attr_key, status, effective_weight DESC),
+  CONSTRAINT fk_um_attr_entity FOREIGN KEY (entity_id) REFERENCES um_entities(entity_id) ON DELETE SET NULL
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+CREATE TABLE IF NOT EXISTS um_relations (
+  relation_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  from_entity_id CHAR(36) NOT NULL,
+  rel_type VARCHAR(64) NOT NULL,
+  to_entity_id CHAR(36) NOT NULL,
+  confidence DECIMAL(6,4) NOT NULL,
+  evidence_ids JSON NOT NULL,
+  candidate_id CHAR(36) NULL,
+  first_seen_at DATETIME(3) NOT NULL,
+  last_seen_at DATETIME(3) NOT NULL,
+  status ENUM('active', 'superseded', 'revoked') NOT NULL DEFAULT 'active',
+  KEY idx_um_rel_user (user_id, from_entity_id, rel_type),
+  CONSTRAINT fk_um_rel_from FOREIGN KEY (from_entity_id) REFERENCES um_entities(entity_id) ON DELETE CASCADE,
+  CONSTRAINT fk_um_rel_to FOREIGN KEY (to_entity_id) REFERENCES um_entities(entity_id) ON DELETE CASCADE
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── L4: Profile Projection（Graph 的结构化投影，非 narrative）──────────────
+
+CREATE TABLE IF NOT EXISTS um_profile_versions (
+  profile_version INT NOT NULL,
+  user_id CHAR(36) NOT NULL,
+  structured_json JSON NOT NULL,
+  content_hash CHAR(64) NOT NULL,
+  parent_version INT NULL,
+  materialize_reason VARCHAR(64) NOT NULL,
+  -- daily_consolidation | candidate_accepted | threshold | manual
+  created_at BIGINT NOT NULL,
+  status ENUM('active', 'superseded', 'revoked') NOT NULL DEFAULT 'active',
+  PRIMARY KEY (user_id, profile_version)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── L5: Agent Snapshot（Profile Projection 的渲染视图，Session 加载）────────
+
+CREATE TABLE IF NOT EXISTS um_profile_snapshots (
+  snapshot_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  projection VARCHAR(32) NOT NULL DEFAULT 'default',
+  -- default | coding | schedule
+  profile_version INT NOT NULL,
+  fast_revision INT NOT NULL DEFAULT 0,
+  snapshot_json JSON NOT NULL,
+  byte_size INT NOT NULL,
+  content_hash CHAR(64) NOT NULL,
+  created_at BIGINT NOT NULL,
+  expires_at BIGINT NULL,
+  status ENUM('active', 'superseded', 'revoked') NOT NULL DEFAULT 'active',
+  KEY idx_um_snapshot_user_proj (user_id, projection, status, created_at DESC)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── 辅助：ingest 游标（按 user + source_type）────────────────────────────────
+
+CREATE TABLE IF NOT EXISTS um_ingest_cursors (
+  user_id CHAR(36) NOT NULL,
+  source_type VARCHAR(32) NOT NULL,
+  cursor_value VARCHAR(256) NOT NULL,
+  updated_at BIGINT NOT NULL,
+  PRIMARY KEY (user_id, source_type)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
+
+-- ── v0.1 预留：Memory lineage（暂不写入，schema 先冻结）────────────────────
+
+CREATE TABLE IF NOT EXISTS um_memory_promotions (
+  promotion_id CHAR(36) PRIMARY KEY,
+  user_id CHAR(36) NOT NULL,
+  candidate_id CHAR(36) NOT NULL,
+  attribute_id CHAR(36) NULL,
+  memind_candidate_id VARCHAR(64) NULL,
+  content_hash CHAR(64) NOT NULL,
+  promoted_at BIGINT NOT NULL,
+  revoked_at BIGINT NULL,
+  status ENUM('promoted', 'revoked') NOT NULL DEFAULT 'promoted',
+  KEY idx_um_promo_user (user_id, status)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
@@ -0,0 +1,105 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://memind.local/schemas/timeline-item.v1.schema.json",
+  "title": "TimelineItem",
+  "description": "Temporal Recall 统一时间线条目 — 跨 Source 归一化模型",
+  "type": "object",
+  "required": [
+    "timeline_item_id",
+    "user_id",
+    "source",
+    "type",
+    "observed_time",
+    "title",
+    "content",
+    "importance",
+    "confidence",
+    "source_ref",
+    "status"
+  ],
+  "additionalProperties": false,
+  "properties": {
+    "timeline_item_id": {
+      "type": "string",
+      "format": "uuid",
+      "description": "归一化后的 timeline 条目 ID"
+    },
+    "user_id": {
+      "type": "string",
+      "format": "uuid"
+    },
+    "source": {
+      "type": "string",
+      "enum": ["calendar", "chat", "meinput", "memory_v2", "email", "browser", "tasks", "git", "location", "notes"]
+    },
+    "type": {
+      "type": "string",
+      "enum": ["event", "todo", "decision", "mention", "commitment"]
+    },
+    "event_time": {
+      "type": ["string", "null"],
+      "format": "date-time",
+      "description": "事件计划/实际发生时间；未知时为 null"
+    },
+    "observed_time": {
+      "type": "string",
+      "format": "date-time",
+      "description": "首次被记录/输入/提到的时间"
+    },
+    "title": {
+      "type": "string",
+      "maxLength": 512
+    },
+    "content": {
+      "type": "string",
+      "maxLength": 8192
+    },
+    "importance": {
+      "type": "number",
+      "minimum": 0,
+      "maximum": 1
+    },
+    "confidence": {
+      "type": "number",
+      "minimum": 0,
+      "maximum": 1,
+      "description": "抽取/归一化置信度"
+    },
+    "recall_score": {
+      "type": "number",
+      "minimum": 0,
+      "maximum": 1,
+      "description": "最终召回排序分"
+    },
+    "source_ref": {
+      "type": "string",
+      "maxLength": 512,
+      "description": "源系统引用，如 meinput:segment:<id>"
+    },
+    "participants": {
+      "type": "array",
+      "items": { "type": "string", "maxLength": 128 },
+      "default": []
+    },
+    "status": {
+      "type": "string",
+      "enum": ["planned", "completed", "mentioned", "unknown"]
+    },
+    "merged_from": {
+      "type": "array",
+      "items": { "type": "string", "maxLength": 512 },
+      "description": "Dedupe 合并前的 source_ref 列表"
+    },
+    "app_bundle_id": {
+      "type": ["string", "null"],
+      "description": "MeInput 来源 App"
+    },
+    "app_name": {
+      "type": ["string", "null"]
+    },
+    "metadata": {
+      "type": "object",
+      "additionalProperties": true
+    }
+  }
+}