Files
Enterprise-Knowledge-Base/internal/application/repository/wiki_log_entry.go
T
wizardchen d0144f3586 perf(wiki): move ingest log to event table and index to on-demand API
Fixes O(n²) write amplification during wiki ingest on large KBs. Previously
every ingest/retract op re-wrote the single `slug='log'` row end-to-end and
every batch re-wrote the entire `slug='index'` directory markdown. On a 40k-
doc KB the log row grew to tens of MB and the index row to several MB, so
each batch triggered giant TOAST updates that dominated ingest wall time.

Log: new `wiki_log_entries` event table (`id DESC` indexed per KB) replaces
the single TEXT row. Batch ingest now collects entries and flushes them
once per batch via `AppendBatch`. Each entry stores `pages_affected` as
JSONB `[{slug,title}]` so the UI can render real titles; custom Scan falls
back to legacy `[]string` so older rows still deserialize.

Index: `wiki_pages[slug=index].content` keeps only the LLM-generated intro
(a few KB). The directory is now served by a structured paginated API
(`GetIndexView`) that reads `slug/title/summary` per type with cursor
pagination, so the agent and the frontend only pull the slice they need.
`RebuildIndexPage` degrades to a no-op; agent `wiki_read_page('index')`
synthesizes a small top-K overview and points callers at `wiki_search`.

Ingest resilience: LLM calls wrap with 3-attempt exponential backoff on
transient errors (5xx/408/429, transport resets/timeouts). Summary/extract
failures now bubble up so the batch's failed-op requeue path runs instead
of silently dropping the doc.

Frontend: sidebar Index/Log entries switch to dedicated views. Index view
streams intro → Summary → Entity → Concept → Synthesis → Comparison via
IntersectionObserver (with a nextTick re-check so small KBs still load
every section). Log view uses cursor pagination; pages_affected renders
titles with slug tooltip.

- new migration: migrations/versioned/000040_wiki_log_entries.{up,down}.sql
- tests: log repo pagination + legacy Scan, ListByTypeLight windowing,
  renderIndexOverviewForAgent output, isTransientLLMError classifier

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-08 21:48:52 +08:00

93 lines
2.7 KiB
Go

package repository
import (
"context"
"errors"
"fmt"
"strconv"
"github.com/Tencent/WeKnora/internal/types"
"github.com/Tencent/WeKnora/internal/types/interfaces"
"gorm.io/gorm"
)
// wikiLogEntryRepository implements interfaces.WikiLogEntryRepository.
type wikiLogEntryRepository struct {
db *gorm.DB
}
// NewWikiLogEntryRepository constructs a GORM-backed WikiLogEntryRepository.
func NewWikiLogEntryRepository(db *gorm.DB) interfaces.WikiLogEntryRepository {
return &wikiLogEntryRepository{db: db}
}
// AppendBatch inserts every entry in one statement. Empty batches are a
// no-op so callers can invoke unconditionally at the end of a wiki ingest
// batch without guarding against the "no events this round" case.
func (r *wikiLogEntryRepository) AppendBatch(ctx context.Context, entries []*types.WikiLogEntry) error {
if len(entries) == 0 {
return nil
}
return r.db.WithContext(ctx).Create(&entries).Error
}
// List returns up to `limit` entries strictly older than `cursor`, newest
// first. `cursor` is the stringified ID of the oldest entry from the
// previous page; an empty string starts from the newest. Callers get back
// a nextCursor string to pass on the next request — empty means no more.
//
// `limit` is clamped to [1, 200]. Values outside that range are coerced
// silently to keep the handler simple.
func (r *wikiLogEntryRepository) List(
ctx context.Context,
kbID string,
cursor string,
limit int,
) ([]*types.WikiLogEntry, string, error) {
if limit <= 0 {
limit = 50
}
if limit > 200 {
limit = 200
}
q := r.db.WithContext(ctx).
Where("knowledge_base_id = ?", kbID).
Order("id DESC").
Limit(limit)
if cursor != "" {
cursorID, err := strconv.ParseUint(cursor, 10, 64)
if err != nil {
return nil, "", fmt.Errorf("invalid cursor %q: %w", cursor, err)
}
q = q.Where("id < ?", cursorID)
}
var entries []*types.WikiLogEntry
if err := q.Find(&entries).Error; err != nil {
return nil, "", err
}
nextCursor := ""
// Only emit a cursor when we actually filled the page — a short page
// is guaranteed to be the tail, so returning a cursor would just cause
// the frontend to fire a final empty request.
if len(entries) == limit {
nextCursor = strconv.FormatUint(entries[len(entries)-1].ID, 10)
}
return entries, nextCursor, nil
}
// DeleteByKB drops every log entry for a KB. No "affected rows" signal is
// surfaced — missing rows are a legitimate state (e.g., a KB that was
// created before wiki_log_entries existed) and not a failure condition.
func (r *wikiLogEntryRepository) DeleteByKB(ctx context.Context, kbID string) error {
if kbID == "" {
return errors.New("wiki log entries: empty kb id")
}
return r.db.WithContext(ctx).
Where("knowledge_base_id = ?", kbID).
Delete(&types.WikiLogEntry{}).Error
}