v0.0.1

This release wires the CI pipeline to publish multi-arch GHCR images and keeps docs/deployments aligned with the new container sources and tagging behaviour.

Highlights

• Added: CI now builds and pushes multi-arch ghcr.io/mfittko/{raged-api,raged,raged-worker} images on main and v*.*.* tags, publishing deterministic main, sha-*, semantic version, and release-only latest tags so deploys can track specific releases or the current build stream.
• Changed: Helm values, deployment docs, and the contributing guide now point to the official GHCR repositories/tags and describe the publish triggers plus the latest policy for clarity around image sourcing.

Upgrade Notes

• Point deployments and tooling to the new GHCR repositories/tags described in the updated docs so you pull the published artifacts instead of legacy registries.

Links

• Compare changes

v0.0.0-rc.20260216.12

Release note generation now retries adaptively when token usage bumps into limits, reducing failures when collecting content for new releases.

Highlights

• Adaptive token retries: Release notes are now produced with an adaptive retry strategy when tokens are exhausted, making the note-generation step more resilient.

Links

• Compare v0.0.0-rc.20260216.11...v0.0.0-rc.20260216.12

v0.0.0-rc.20260216.11

This release finalizes the Postgres-only stack, adds operational safeguards, and refreshes CI/CD tooling so the API/worker pair and their deployments run more predictably.

Highlights

Added

• Docker Compose now gates qdrant/ollama startup on healthchecks so the API only begins operations once its dependencies are ready.
• A Postgres database layer boots migrations, pools connections, and enqueues enrichment work so the API owns its schema and work queue.
• Added an OpenAI-driven release creation script plus a GitHub Actions workflow to publish Docker images to GHCR, along with manual publish dispatch support for tighter release automation.

Changed

• Removed the legacy Qdrant/Neo4j/Redis scaffolding, updated CLI/docs/config/runtime messaging, and consolidated ingest/query/enrichment services and worker pipelines onto Postgres (with pgvector) and Ollama.
• The worker now communicates with the API over /internal/* endpoints, probes /healthz before becoming ready, and sends plain key-value filters so Postgres translates them to SQL.
• Startup now validates DATABASE_URL/OLLAMA_URL/QDRANT_URL before listening, warns about weak API tokens, and keeps Helm/chart defaults aligned with the renamed raged branding.
• Updated Docker tag handling, latest-tag policy, and Helm chart configuration so image publishing and version tracking behave consistently when pushing from main.

Fixed

• Hardened release-note generation and semver bumping, and clarified Git tag versus Docker tag naming so releases reflect the intended versions across tooling.

Upgrade Notes

• Ensure DATABASE_URL, OLLAMA_URL, and QDRANT_URL are set before starting the API, as it now validates them before listening.

Links

• Compare v0.0.0-rc.20260216.10...v0.0.0-rc.20260216.11