Accept memory fact source aliases

Author: SqlRush

docs/cc-100-roadmap.md

@@ -188,6 +188,8 @@ M6 补充：session-memory recall agent 和 relevant-memory selector 现在也
 
 M6 补充：model-backed memory fact extraction 现在也接受 provider-style `choices`/`outputs`/`candidates`/`generations` response wrapper 以及顶层 `message`/`content`/`text` envelope，可从 `message.content`、content-block array、`content.parts[].text` 和 fenced `json` code block 里递归恢复 JSON facts payload。
 
+M6 补充：model-backed memory fact extraction 现在接受更多 fact source aliases，包括 `sourceMessageUUID`/`source_message_uuid`、`sourceEventId`/`source_event_id`、`originId` 以及 `turn`/`event` source object，并保留 numeric source IDs 为字符串。
+
 M6 补充：compact runner 的 summary 响应现在也接受 provider-style `choices`/`outputs`/`candidates`/`generations` wrapper，可在构建 compact plan 前从 `message.content`、content-block array、`content.parts[].text` 和 fenced `json` code block 中恢复 visible summary text。
 
 M6/M7 补充：Runner 会把 `RelevantMemoryDir` 透传到 tool metadata 的 internal auto-memory path context，让 Read tool freshness prefix 和 permission internal-path policy 在同一 memory dir 配置下生效。

docs/claude-code-go-rewrite-plan.md

@@ -300,6 +300,7 @@ test/parity/                 # golden tests against TS/official behavior
 - 本轮补充：session-memory recall agent 现在接受 `summaries`、`selectedSummaries`、`relevantSummaries`、`candidateSummaries` 等 summary collection aliases，并会继续从嵌套 `summary.sessionId`/`summaryId` 恢复模型排序的 session IDs。
 - 本轮补充：session-memory recall agent 现在也接受 `sessionUri`/`sessionUrl`、`uri`/`url`/`href` 等 summary link aliases，能从 `file://.../summary.md` 或 API URL path 中恢复 session ID 并按模型顺序匹配 summary。
 - 本轮补充：model-backed memory fact extraction 现在也接受 provider-style `choices`/`outputs`/`candidates`/`generations` response wrapper 以及顶层 `message`/`content`/`text` envelope，可从 `message.content`、content-block array、`content.parts[].text` 和 fenced `json` code block 里递归恢复 JSON facts payload。
+- 本轮补充：model-backed memory fact extraction 现在接受 `sourceMessageUUID`/`source_message_uuid`、`sourceEventId`/`source_event_id`、`originId` 和 `turn`/`event` source object 等来源别名，并把 numeric source ID 保留为字符串。
 - 本轮补充：compact runner 的 summary 响应现在也接受 provider-style `choices`/`outputs`/`candidates`/`generations` wrapper，可在构建 compact plan 前从 `message.content`、content-block array、`content.parts[].text` 和 fenced `json` code block 中恢复 visible summary text。
 - 本轮补充：remote history `SDKEvent` type 现在会把 provider-style aliases 归一化为现有 canonical 事件类型，包括 `assistant_message`/`assistant_delta`、`userMessage`/`humanMessage`、`system-event`、`result_event`/`finalResult`/`response.completed`、`errorEvent`/`failureEvent` 和 `status_update`/`statusMessage`/`progress`，single-object page 与 transcript materialization 不再因事件类型拼写相邻而丢消息。
 - 本轮补充：conversation runner 会在用户消息入队后基于 compact window 计算 token warning state，并在达到 warning/error/auto-compact/blocking 阈值时发出 `token_warning` event；warning state 接入 blocking-limit env override，auto-compact 判断接入既有 `CLAUDE_AUTOCOMPACT_PCT_OVERRIDE`，避免 warning 和 compact 阈值来源分叉。

docs/first-second-parity-audit.md

@@ -155,6 +155,7 @@ M6 progress now includes:
 - `internal/memory`: model-backed memory fact extraction now also unwraps provider-style `choices`, `outputs`, `candidates`, and `generations` responses plus top-level `message`/`content`/`text`/`output_text` envelopes, recovering JSON facts payloads from nested `message.content`, content-block arrays, `content.parts[].text`, and fenced `json` code blocks.
 - `internal/memory`: model-backed memory fact extraction now also traverses API-shaped fact collections such as `observations`, `notes`, `findings`, `records`, `resources`, `data.resource.attributes`, and `edge.node`, while accepting text aliases such as `note`, `description`, `body`, `message`, `observation`, and `finding`.
 - `internal/memory`: model-backed memory fact extraction now accepts additional kind aliases such as `user_pref`, `requirement`, `action_item`, `outcome`, `conclusion`, `tool_usage`, and `command_run`, normalizing them into preference/request/decision/tool facts.
+- `internal/memory`: model-backed memory fact extraction now accepts additional source aliases such as `sourceMessageUUID`, `source_message_uuid`, `sourceEventId`, `source_event_id`, `originId`, and nested `turn`/`event` source objects, preserving numeric source IDs as strings.
 - `internal/session`: remote-history response parsing now also unwraps provider-style `choices`, `outputs`, `candidates`, and `generations` responses plus top-level `message`/`content`/`text`/`output_text` envelopes, recovering event page JSON plus pagination from nested `message.content`, content-block arrays, `content.parts[].text`, and fenced `json` code blocks.
 - `internal/session`: remote-history pagination now accepts `starting_after`/`startingAfter`/`after*` cursor aliases from page fields and link URL query parameters when deriving the next `before_id`.
 - `internal/session`: project session listing and pagination, prompt-history lock/buffered flush/field aliases, lightweight transcript index/title/text-preview inference, line-offset transcript indexing/window/byte-budget-window/parent-chain/resume/tail/byte-budget-tail loading, AI-title/last-prompt/task/agent/PR/worktree transcript metadata loading and type/field aliases, transcript message/session UUID field aliases, tombstone metadata loading with target/session/reason aliases and delete/relink replay, agent-scoped content replacement metadata/record field-alias loading, session-scoped metadata reappend including AI-title/last-prompt/task-summary, streaming transcript search snippets for resume/search UI, official `subagents/agent-*.jsonl` transcript layout with legacy sidechain listing, agent metadata sidecar read/write/field aliases, sidechain runtime start/append/finish/cancel/fail summary bridging plus parent-chain append/finish, sidechain manager orchestration for spawn/append/finish/cancel/fail/list/resume, sidechain state/list/resume support with content-field aliases, sidechain resume context construction, sidechain conversation and agent-scoped content-replacement reconstruction, transcript tail/window loading and byte-budget tail loading for bounded-memory resume/UI paths, lightweight transcript metadata loading, remote-history token refresh retry, page-field/event-list/records/entries/last-id/cursor/event-id/has-next aliases plus wrapped-data/links/paging/bare-array/keyed-event-map responses, before_id resume, fallback field fill, and duplicate-aware parent linking during transcript materialization.

internal/memory/agent.go

@@ -8,6 +8,7 @@ import (
 	"net/url"
 	"path/filepath"
 	"sort"
+	"strconv"
 	"strings"
 
 	"ccgo/internal/api/anthropic"
@@ -2400,18 +2401,18 @@ func rawMemoryFactFromMap(value map[string]any) (rawMemoryFact, bool) {
 		MessageText:          stringMapField(value, "message"),
 		Observation:          stringMapField(value, "observation"),
 		Finding:              stringMapField(value, "finding"),
-		SourceUUID:           stringMapField(value, "source_uuid"),
-		SourceUUIDCamel:      stringMapField(value, "sourceUuid"),
-		SourceID:             stringMapField(value, "source_id"),
-		SourceIDCamel:        stringMapField(value, "sourceId"),
-		Source:               nestedIDFromValue(value["source"]),
-		MessageUUID:          stringMapField(value, "message_uuid"),
-		MessageUUIDCamel:     stringMapField(value, "messageUuid"),
-		MessageID:            stringMapField(value, "message_id"),
-		MessageIDCamel:       stringMapField(value, "messageId"),
-		SourceMessageID:      stringMapField(value, "source_message_id"),
-		SourceMessageIDCamel: stringMapField(value, "sourceMessageId"),
-		UUID:                 stringMapField(value, "uuid"),
+		SourceUUID:           idMapField(value, "source_uuid", "sourceUUID"),
+		SourceUUIDCamel:      idMapField(value, "sourceUuid"),
+		SourceID:             idMapField(value, "source_id", "sourceID", "source_event_id", "source_event_uuid", "origin_id", "origin_uuid"),
+		SourceIDCamel:        idMapField(value, "sourceId", "sourceEventId", "sourceEventID", "sourceEventUuid", "sourceEventUUID", "originId", "originUuid", "originUUID"),
+		Source:               sourceIDFromFactMap(value),
+		MessageUUID:          idMapField(value, "message_uuid", "messageUUID"),
+		MessageUUIDCamel:     idMapField(value, "messageUuid"),
+		MessageID:            idMapField(value, "message_id", "messageID"),
+		MessageIDCamel:       idMapField(value, "messageId"),
+		SourceMessageID:      idMapField(value, "source_message_id", "source_message_uuid", "sourceMessageID", "sourceMessageUUID"),
+		SourceMessageIDCamel: idMapField(value, "sourceMessageId", "sourceMessageUuid"),
+		UUID:                 idMapField(value, "uuid"),
 	}
 	if fact.MessageUUID == "" {
 		if _, ok := value["message"].(map[string]any); ok {
@@ -2433,6 +2434,19 @@ func rawMemoryFactFromMap(value map[string]any) (rawMemoryFact, bool) {
 	return fact, true
 }
 
+func idMapField(value map[string]any, keys ...string) string {
+	return idMapFieldDepth(value, 0, keys...)
+}
+
+func idMapFieldDepth(value map[string]any, depth int, keys ...string) string {
+	for _, key := range keys {
+		if id := nestedIDFromValueDepth(value[key], depth+1); id != "" {
+			return id
+		}
+	}
+	return ""
+}
+
 func stringMapField(value map[string]any, keys ...string) string {
 	for _, key := range keys {
 		if text := textFromValue(value[key]); text != "" {
@@ -2442,15 +2456,75 @@ func stringMapField(value map[string]any, keys ...string) string {
 	return ""
 }
 
+func sourceIDFromFactMap(value map[string]any) string {
+	for _, key := range []string{
+		"source",
+		"origin",
+		"source_message",
+		"sourceMessage",
+		"source_event",
+		"sourceEvent",
+		"source_turn",
+		"sourceTurn",
+		"event",
+		"turn",
+	} {
+		if id := nestedIDFromValue(value[key]); id != "" {
+			return id
+		}
+	}
+	return ""
+}
+
 func nestedIDFromValue(value any) string {
-	if text := directStringValue(value); text != "" {
-		return text
+	return nestedIDFromValueDepth(value, 0)
+}
+
+func nestedIDFromValueDepth(value any, depth int) string {
+	if depth > 6 {
+		return ""
+	}
+	if id := directIDValue(value); id != "" {
+		return id
 	}
 	object, ok := value.(map[string]any)
 	if !ok {
 		return ""
 	}
-	return stringMapField(object, "source_uuid", "sourceUuid", "source_id", "sourceId", "message_uuid", "messageUuid", "message_id", "messageId", "uuid", "id")
+	if id := idMapFieldDepth(object, depth+1,
+		"source_uuid", "sourceUuid", "sourceUUID",
+		"source_id", "sourceId", "sourceID",
+		"source_message_uuid", "sourceMessageUuid", "sourceMessageUUID",
+		"source_message_id", "sourceMessageId", "sourceMessageID",
+		"message_uuid", "messageUuid", "messageUUID",
+		"message_id", "messageId", "messageID",
+		"event_uuid", "eventUuid", "eventUUID",
+		"event_id", "eventId", "eventID",
+		"turn_uuid", "turnUuid", "turnUUID",
+		"turn_id", "turnId", "turnID",
+		"origin_uuid", "originUuid", "originUUID",
+		"origin_id", "originId", "originID",
+		"uuid", "id",
+	); id != "" {
+		return id
+	}
+	for _, key := range []string{
+		"source",
+		"origin",
+		"source_message",
+		"sourceMessage",
+		"source_event",
+		"sourceEvent",
+		"source_turn",
+		"sourceTurn",
+		"event",
+		"turn",
+	} {
+		if id := nestedIDFromValueDepth(object[key], depth+1); id != "" {
+			return id
+		}
+	}
+	return ""
 }
 
 func textFromValue(value any) string {
@@ -2481,6 +2555,41 @@ func directStringValue(value any) string {
 	return strings.TrimSpace(text)
 }
 
+func directIDValue(value any) string {
+	switch typed := value.(type) {
+	case string:
+		return strings.TrimSpace(typed)
+	case json.Number:
+		return strings.TrimSpace(typed.String())
+	case float64:
+		return strings.TrimSpace(strconv.FormatFloat(typed, 'f', -1, 64))
+	case float32:
+		return strings.TrimSpace(strconv.FormatFloat(float64(typed), 'f', -1, 32))
+	case int:
+		return strconv.Itoa(typed)
+	case int8:
+		return strconv.FormatInt(int64(typed), 10)
+	case int16:
+		return strconv.FormatInt(int64(typed), 10)
+	case int32:
+		return strconv.FormatInt(int64(typed), 10)
+	case int64:
+		return strconv.FormatInt(typed, 10)
+	case uint:
+		return strconv.FormatUint(uint64(typed), 10)
+	case uint8:
+		return strconv.FormatUint(uint64(typed), 10)
+	case uint16:
+		return strconv.FormatUint(uint64(typed), 10)
+	case uint32:
+		return strconv.FormatUint(uint64(typed), 10)
+	case uint64:
+		return strconv.FormatUint(typed, 10)
+	default:
+		return ""
+	}
+}
+
 type rawMemoryFact struct {
 	Kind                 string `json:"kind"`
 	Type                 string `json:"type"`

internal/memory/memory_test.go

@@ -1359,6 +1359,36 @@ func TestMemoryAgentExtractsNestedFactSourceObjects(t *testing.T) {
 	}
 }
 
+func TestMemoryAgentExtractsAdditionalFactSourceAliases(t *testing.T) {
+	client := &fakeMemoryClient{response: &anthropic.Response{
+		ID:    "msg_memory_source_aliases",
+		Type:  "message",
+		Role:  "assistant",
+		Model: "sonnet",
+		Content: []contracts.ContentBlock{contracts.NewTextBlock(`{"facts":[
+			{"kind":"preference","text":"prefer source message uuid aliases","sourceMessageUUID":"user_1"},
+			{"type":"decision","content":"keep source event aliases","source_event_id":1234},
+			{"category":"request","summary":"support turn source objects","turn":{"id":"turn_1"}}
+		]}`)},
+	}}
+	result, err := (Agent{Client: client}).Extract(context.Background(), []contracts.Message{msgs.UserText("Remember more source aliases")}, ExtractOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.Fallback || len(result.Facts) != 3 {
+		t.Fatalf("result = %#v", result)
+	}
+	if !hasMemoryFact(result.Facts, FactPreference, "prefer source message uuid aliases", "user_1") {
+		t.Fatalf("sourceMessageUUID fact missing = %#v", result.Facts)
+	}
+	if !hasMemoryFact(result.Facts, FactDecision, "keep source event aliases", "1234") {
+		t.Fatalf("source event alias fact missing = %#v", result.Facts)
+	}
+	if !hasMemoryFact(result.Facts, FactRequest, "support turn source objects", "turn_1") {
+		t.Fatalf("turn source object fact missing = %#v", result.Facts)
+	}
+}
+
 func TestMemoryAgentExtractsStructuredFactText(t *testing.T) {
 	client := &fakeMemoryClient{response: &anthropic.Response{
 		ID:    "msg_memory_structured_text",