Third-party version sources
This page is the single source of truth for third-party versions used by the Delphi platform — Kamailio, Asterisk, Janus, RTPEngine, coturn, PostgreSQL, Redis, the OTel collector, base images, etc. Bespoke voiceai-* images that ship from this release are tracked by ECR_TAG; see the release compatibility page for which releases are current and which services changed in each.
Each row points at the single file that pins the version. The table is refreshed during release preparation so operators can quickly check which upstream versions ship with the selected Delphi release.
Pinned versions (v0.9.14-patch1)
| Component | Version | Source-of-truth file |
|---|---|---|
| Asterisk (TelSys) | 22.9.0 | .docker/voiceai-telsys/Dockerfile — ENV ASTERISK_VER |
| Kamailio (TelPro) | 6.1.1-bookworm | .docker/voiceai-telpro/Dockerfile — FROM ghcr.io/kamailio/kamailio:... |
| RTPEngine (TelPro) — dfx.at channel | 26.0 | .docker/voiceai-rtpengine/Dockerfile — ARG RTPENGINE_VERSION |
| Janus (TelPro / WebRTC) — git tag | v1.3.3 | .docker/voiceai-webrtc/Dockerfile — ARG JANUS_VERSION |
| coturn (TelPro) | coturn/coturn:4.6.2 | .infrastructure/services/telpro/docker-compose.yml |
| PostgreSQL (Database) | postgres:17-alpine | .infrastructure/services/database/docker-compose.yaml |
| PgBouncer (Database) | edoburu/pgbouncer:v1.25.1-p0 | .infrastructure/services/database/docker-compose.yaml |
| Redis (Database) | redis:7-alpine | .infrastructure/services/database/docker-compose.yaml |
| Caddy (Web, Media) | caddy:2-alpine | .infrastructure/services/{web,media}/docker-compose.yaml |
| Squid (Squid) | ubuntu/squid:latest (unpinned, see warning below) | .infrastructure/services/squid/docker-compose.yaml |
| OpenTelemetry Collector (contrib) (every service) | otel/opentelemetry-collector-contrib:0.154.0 | .infrastructure/services/*/docker-compose.yaml |
| SigNoz (SigNoz service) | 0.129.0 (git tag v0.129.0) | .infrastructure/services/signoz/init.sh — SIGNOZ_VERSION |
| Node.js (TelAPI, TelWeb, TelPhi, Tasker, Scaler) | node:24-alpine | .docker/voiceai-{telapi,telweb,telphi,tasker,scaler}/Dockerfile — FROM node:... |
| Python (CPU AudioProc) | python:3.11-slim | .docker/voiceai-audioproc/Dockerfile — FROM python:... |
| NVIDIA CUDA (GPU AudioProc) | nvidia/cuda:12.4.1-runtime-ubuntu22.04 | .docker/voiceai-audioproc/Dockerfile — FROM nvidia/cuda:... |
| Go (media-upload build) | golang:1.24-alpine | .docker/voiceai-media-server/Dockerfile — FROM golang:... |
| Alpine (sidecar runtimes) | alpine:3.19 | .docker/voiceai-media-server/Dockerfile, telpro / voice docker-compose |
| Debian (telsys / telpro / rtpengine / webrtc runtime base) | bookworm-slim | various .docker/voiceai-*/Dockerfile |
The Squid container is the only one in the stack pinned to a floating tag. Behaviour can change unexpectedly on update.sh (which always pulls). Pin to a digest in .infrastructure/services/squid/docker-compose.yaml for reproducibility.
What's intentionally not pinned
ubuntu/squid:latest— see warning above.- The
voiceai-*image tag itself (ECR_TAG) — set per deployment via the bootstrap/opt/deployment/.env. The release notes are the source of truth for whatECR_TAGships at each release.
See also
- Release compatibility — which Delphi releases are current and which services changed in each.
- Service operations — per-service Containers tables that quote the versions on this page.
- Release notes — v0.9.14-patch1 — product changelog and operator rollout checklist for deployments.