// One-time backfill: mirror existing Chroma vector-memory records into the new FTS5
// keyword-search sidecar (memory-fts.ts, added 2026-08-06 hybrid-search upgrade). Only
// needed once — addVector() keeps the two in sync for everything written after this ran.
import { ChromaClient } from 'chromadb';
import { USER_FACTS_COLLECTION, DAILY_EXTRACTS_COLLECTION } from '../src/gateway/memory/memory-vector';
import { upsertFtsRecord } from '../src/gateway/memory/memory-fts';

async function backfillCollection(client: ChromaClient, name: string) {
  const col = await client.getOrCreateCollection({ name, embeddingFunction: null });
  const count = await col.count();
  console.log(`[${name}] ${count} records`);
  const batchSize = 200;
  let offset = 0;
  let written = 0;
  while (offset < count) {
    const res = await col.get({ limit: batchSize, offset });
    const ids = res.ids || [];
    const docs = res.documents || [];
    const metas = res.metadatas || [];
    for (let i = 0; i < ids.length; i++) {
      const text = docs[i] || '';
      if (!text) continue;
      const workspace = String((metas[i] as any)?.workspace || '');
      upsertFtsRecord(name, ids[i], text, workspace);
      written++;
    }
    offset += batchSize;
  }
  console.log(`[${name}] backfilled ${written} records into FTS`);
}

async function main() {
  const client = new ChromaClient({ host: 'localhost', port: 8100 });
  await backfillCollection(client, USER_FACTS_COLLECTION);
  await backfillCollection(client, DAILY_EXTRACTS_COLLECTION);
  console.log('done');
}

main().catch(e => { console.error(e); process.exit(1); });