RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.
Go agent-harnessagent-native-databaseagentic-aiagentic-retrievalagentic-searchaiai-agentscontext-enginecontext-engineeringcontext-managementharness-engineeringknowledge-compilationragretrieval-augmented-generationretrieval-harnesssearch-harness
Fix: handle tenant_llm without id (#18541)
Lynn618c4599b10e
diff --git a/tools/scripts/mysql_migration.py b/tools/scripts/mysql_migration.pyindex bf7973f2f6..ea400f25dc 100644--- a/tools/scripts/mysql_migration.py+++ b/tools/scripts/mysql_migration.py@@ -772,9 +772,11 @@ class TenantModelStage(MigrationStage):# We cannot JOIN tenant_model_instance on api_key directly because the instance# stage deduped api_keys (stripping is_tools), so a plain SQL equality won't# match records whose api_key was merged. Count at the provider level instead.+ # Use a constant instead of tl.id: very old tenant_llm tables may lack the+ # id column, and the count only cares about the number of rows.cursor = self.db.execute_sql(f"SELECT COUNT(*) FROM ("- f" SELECT tl.id "+ f" SELECT 1 "f" FROM tenant_llm tl "f" INNER JOIN tenant_model_provider tmp ON tmp.tenant_id = tl.tenant_id AND tmp.provider_name = tl.llm_factory "f" WHERE {status_condition} "@@ -818,6 +820,14 @@ class TenantModelStage(MigrationStage):logger.info("[CREATE TABLE ONLY] Target table created/verified, skipping data migration")return 0, self.target_tables+ # Very old tenant_llm tables may lack the id column, while the migration+ # SELECT below relies on it (source_id for audit logging). Add an+ # auto-increment id column in place before running the data comparison+ # and inserts. Fail hard when it cannot be ensured: silently returning+ # would let the runner mark this stage as completed without migrating.+ if not self._ensure_tenant_llm_id_column():+ raise RuntimeError("tenant_llm has no 'id' column and it could not be added; aborting tenant_model stage")+status_condition = self._build_status_condition()# Load all tenant_model_instance records into memory for Python-level matching.@@ -896,6 +906,35 @@ class TenantModelStage(MigrationStage):return rows_inserted, self.target_tables+ def _ensure_tenant_llm_id_column(self) -> bool:+ """Ensure tenant_llm has an id column, adding an auto-increment one if missing.++ Very old tenant_llm tables may predate the id column, while the migration+ SELECT relies on it (tl.id is carried as source_id for audit logging).+ The id is only used as a unique audit key, so it is added as an+ AUTO_INCREMENT UNIQUE column rather than PRIMARY KEY: old tables may+ already have another primary key. MySQL back-fills existing rows with+ sequential values when adding an AUTO_INCREMENT column.++ Returns:+ True if the column exists (or was added successfully), False otherwise.+ """+ if self.db.column_exists("tenant_llm", "id"):+ return True++ logger.warning("Source table 'tenant_llm' has no 'id' column, adding auto-increment id column")+ if self.dry_run:+ logger.info("[DRY RUN] Would ALTER TABLE tenant_llm to add auto-increment id column")+ return False++ try:+ self.db.execute_sql("ALTER TABLE tenant_llm ADD COLUMN id BIGINT AUTO_INCREMENT UNIQUE")+ except Exception as e:+ logger.error(f"Failed to add auto-increment id column to tenant_llm: {e}")+ return False+ logger.info("Added auto-increment id column to tenant_llm")+ return True+def _build_instance_lookup(self) -> dict:"""Load all tenant_model_instance records, indexed by (provider_id, canonical_api_key).