--docker. Native remains the default when you run ragsuite init without flags.
Recommended: CLI
From a repository clone
Vector store
Vectors live in the same PostgreSQL database with the pgvector extension. The Compose Postgres image includes pgvector, and the existingpostgres_data volume is reused — there is no separate vector database container.
Upgrading from a Chroma-based release
Releases 1.0.7 and earlier stored vectors in ChromaDB. The first start after upgrading runs a one-timevector-migrate step that imports that data into pgvector and verifies every row before the API starts.
1
Back up
From your install folder (
~/ragsuite for CLI installs), back up the database and the old vector data before upgrading:2
Start with the full command
Run
ragsuite update --restart (or docker compose up -d --build from a clone). Do not use --no-deps backend for this first start — the API would start before the import.3
Check the import
In Docker mode:The log should end with
migration complete: N collection(s), M rows. Later starts report already migrated and skip the step. If verification fails, the API stays down and the old data is untouched — fix the cause and start again; the import resumes.4
Clean up
After you have verified answers in the console, delete
backend/data/chroma_db, then run docker compose up -d --remove-orphans to remove the old chromadb container.Native installs run the same import during
ragsuite start, before the API starts, and print the result in the start output. Host Postgres must have pgvector installed, and you must run CREATE EXTENSION vector; once as a superuser in the RAGSuite database. If your old vector data is in a non-default location, set CHROMA_PERSIST_PATH (native) or CHROMA_HOST_PERSIST_PATH (Docker) before the first start.Ports (Docker)
See Ports & prerequisites and Install with CLI.