pgtapes

pgtapes builds the PostgreSQL container image that tapes runs on: a CloudNativePG base image with pg_duckdb and pgvector built in.

The published image:

public.ecr.aws/g4e5l3z3/papercomputeco/postgres:17.7-pgduckdb-1.1.1

Built for linux/amd64 and linux/arm64. The registry also carries older tags you should not use — see Image tags.

Why this image exists

Tapes is a session-capture server. It stores captured agent sessions in PostgreSQL and uses pgvector for semantic search over span embeddings, so it needs a PostgreSQL image with that extension present and loadable rather than a stock one. pg_duckdb is preloaded alongside it to give an analytical query path over the same data.

Because the base is a CloudNativePG image, one build serves both deployment shapes: docker run (or Compose) for a local stack, and the CloudNativePG operator in Kubernetes, without being rebuilt. The two paths configure themselves differently — that split is the main thing worth understanding about this image, and Run the image walks both.

This repository contains no server code and no database schema. It is a Dockerfile, an init script, and a build module; the tapes server owns its own migrations. To run tapes itself, see https://tapes.dev/docs/installation/ — tapes local up provisions this published image for a local stack.

Where to go

• Run the image — docker run and CloudNativePG, and what each path does and does not configure for you.
• What is in the image — base, extension versions, and the mechanisms the build adds.
• Image tags — the tag scheme, and which tags on the registry to avoid.
• Build and publish — building the image from this repository.