Fix TiDB auto-embedding fallback and migration

[YoungYo]

Summary

This PR fixes TiDB auto-embedding schema drift and makes search degrade gracefully instead of surfacing hard 500 errors.

Problem

In the TiDB auto-embedding flow, a tenant can end up with a plain VECTOR column instead of a proper EMBED_TEXT() generated column.

When that happens:

• memory writes may still succeed
• semantic search can fail at query time with errors like:
  • VEC_EMBED_COSINE_DISTANCE() first argument must be a vector embedding column generated by EMBED_TEXT()

So the broken state is subtle: data appears to be written correctly, but recall/search fails later.

What changed

• detect when memories.embedding / sessions.embedding are not EMBED_TEXT() generated columns
• rebuild the affected tables into the correct TiDB auto-embedding shape when in-place migration is not supported
• recreate vector indexes against the repaired schema
• degrade search gracefully when auto vector search is unavailable instead of returning a hard failure

Why rebuild is needed

TiDB does not always allow converting a plain vector column into a generated EMBED_TEXT() column in place.

For that reason, the repair path falls back to:

1. creating a correctly-shaped replacement table
2. copying the data
3. swapping the tables in place

This keeps the fix practical for already-drifted tenants.

Scope

This PR does not change the overall tenant model or TiDB provisioning flow.
It only repairs schema drift for TiDB auto-embedding tables and prevents search from failing hard while the schema is not yet usable for vector search.

Validation

Validated against TiDB Cloud with auto embedding enabled:

• mnemo-server starts successfully
• memory creation succeeds
• auto-embedding schema is repaired into STORED GENERATED
• hybrid search succeeds again after migration
• OpenClaw -> mem9 -> TiDB flow was verified end to end

[netlify]

👷 Deploy request for mem9 pending review.

A Netlify team Owner will need to approve the deploy before you can run your build.

Are you a team Owner? Visit the deploys page to approve it →

Need more help? Learn more in the Netlify docs →

Name	Link
🔨 Latest commit	b849a07

[you06]

rest LGTM

[you06]

Missing FTS index.

mem9/server/internal/tenant/zero.go

Lines 203 to 209 in 594a24a

	if p.ftsEnabled {
	_, err := db.ExecContext(ctx,
	`ALTER TABLE memories ADD FULLTEXT INDEX idx_fts_content (content) WITH PARSER MULTILINGUAL ADD_COLUMNAR_REPLICA_ON_DEMAND`)
	if err != nil && !IsIndexExistsError(err) {
	return fmt.Errorf("init schema: fulltext index: %w", err)
	}
	}