Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 11 additions & 11 deletions deploy/hive-oke/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -21,23 +21,23 @@ Quorum (RFC #1) uses one shared Vexa control plane per cluster. On hive-oke it r

## Versions (all pinned by tag and digest)

Everything Vexa comes from one commit of hivecommons/vexa `main`: 3addb0a940acc42e899b2748711c93dab2467874 (chart 0.13.4).
Everything Vexa comes from one commit of hivecommons/vexa `main`: 93bc10a5d2972197642a960e885b356a5063d50a (chart 0.13.4).
It is upstream Vexa `main` plus our patches (hivecommons/vexa#2 Teams chat-write, #3 speak route, #7 Teams
per-track capture, #8 Google Meet end of turn, #9 cached and prewarmed short TTS lines).
Upstream v0.12.28 has no meeting chat-post route and its `/ws` sends no chat events; both
are on upstream `main`, and upstream publishes control-plane images only for releases.

| Component | Image |
| ---------------------------------------- | -------------------------------------------------------------------------------- |
| Chart | `deploy/helm/charts/vexa` at hivecommons/vexa 3addb0a (chart 0.13.4) |
| gateway | `ghcr.io/hivecommons/vexa-gateway:main-3addb0a-quorum` |
| admin-api | `ghcr.io/hivecommons/vexa-admin-api:main-3addb0a-quorum` |
| meeting-api | `ghcr.io/hivecommons/vexa-meeting-api:main-3addb0a-quorum` |
| runtime | `ghcr.io/hivecommons/vexa-runtime:main-3addb0a-quorum` |
| meeting bot | `ghcr.io/hivecommons/vexa-bot:main-3addb0a-quorum` |
| CPU transcription (hotwords, warm-up) | `ghcr.io/hivecommons/vexa-transcription-cpu:main-3addb0a-quorum` |
| agent-api, agent-worker (pinned, unused) | `ghcr.io/hivecommons/vexa-agent-api`, `vexa-agent-worker` `:main-3addb0a-quorum` |
| mcp (pinned, disabled) | `ghcr.io/hivecommons/vexa-mcp:main-3addb0a-quorum` |
| Chart | `deploy/helm/charts/vexa` at hivecommons/vexa 93bc10a (chart 0.13.4) |
| gateway | `ghcr.io/hivecommons/vexa-gateway:main-93bc10a-quorum` |
| admin-api | `ghcr.io/hivecommons/vexa-admin-api:main-93bc10a-quorum` |
| meeting-api | `ghcr.io/hivecommons/vexa-meeting-api:main-93bc10a-quorum` |
| runtime | `ghcr.io/hivecommons/vexa-runtime:main-93bc10a-quorum` |
| meeting bot | `ghcr.io/hivecommons/vexa-bot:main-93bc10a-quorum` |
| CPU transcription (hotwords, warm-up) | `ghcr.io/hivecommons/vexa-transcription-cpu:main-93bc10a-quorum` |
| agent-api, agent-worker (pinned, unused) | `ghcr.io/hivecommons/vexa-agent-api`, `vexa-agent-worker` `:main-93bc10a-quorum` |
| mcp (pinned, disabled) | `ghcr.io/hivecommons/vexa-mcp:main-93bc10a-quorum` |
| postgres | `docker.io/library/postgres:17.11-alpine` |
| redis | `docker.io/valkey/valkey:8.1.11-alpine` |
| S3 | `docker.io/versity/versitygw:v1.10.0` |
Expand Down Expand Up @@ -233,7 +233,7 @@ the upstream Kubernetes docs, it is minted at install time and kept in `vexa-adm

```bash
git clone https://github.com/hivecommons/vexa.git /tmp/hivecommons-vexa
git -C /tmp/hivecommons-vexa checkout 3addb0a940acc42e899b2748711c93dab2467874 # quorum.vexaCommit in values.yaml
git -C /tmp/hivecommons-vexa checkout 93bc10a5d2972197642a960e885b356a5063d50a # quorum.vexaCommit in values.yaml
KUBE_CONTEXT=hive-oke VEXA_CHART=/tmp/hivecommons-vexa/deploy/helm/charts/vexa ./install.sh
KUBE_CONTEXT=hive-oke ./smoke.sh
```
Expand Down
8 changes: 4 additions & 4 deletions deploy/hive-oke/transcription.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -40,8 +40,8 @@ spec:
automountServiceAccountToken: false
containers:
- name: transcription
# main-3addb0a-quorum (linux/amd64, linux/arm64), hivecommons/vexa 3addb0a (hotwords, warm-up)
image: ghcr.io/hivecommons/vexa-transcription-cpu:main-3addb0a-quorum@sha256:5dd732b156a00813e7b02e4d30f2e851647e689ef0f9eb29375d4c511079b097
# main-93bc10a-quorum (linux/amd64, linux/arm64), hivecommons/vexa 93bc10a (hotwords, warm-up)
image: ghcr.io/hivecommons/vexa-transcription-cpu:main-93bc10a-quorum@sha256:37158bdcdffe378060994b07135759b782bb6aff6f38e0e36ab53c1321309d50
env:
- name: WORKER_ID
value: "1"
Expand Down Expand Up @@ -118,7 +118,7 @@ spec:
# Fetch faster-whisper `base.en` at a pinned commit of Systran/faster-whisper-base.en into the
# model cache, and point the cache's refs/main at it so the offline server resolves it.
- name: fetch-model
image: ghcr.io/hivecommons/vexa-transcription-cpu:main-3addb0a-quorum@sha256:5dd732b156a00813e7b02e4d30f2e851647e689ef0f9eb29375d4c511079b097
image: ghcr.io/hivecommons/vexa-transcription-cpu:main-93bc10a-quorum@sha256:37158bdcdffe378060994b07135759b782bb6aff6f38e0e36ab53c1321309d50
env:
- name: MODEL_REVISION
value: 3d3d5dee26484f91867d81cb899cfcf72b96be6c
Expand Down Expand Up @@ -148,7 +148,7 @@ spec:
containers:
- name: transcription
# Same image and digest as the batch tier above.
image: ghcr.io/hivecommons/vexa-transcription-cpu:main-3addb0a-quorum@sha256:5dd732b156a00813e7b02e4d30f2e851647e689ef0f9eb29375d4c511079b097
image: ghcr.io/hivecommons/vexa-transcription-cpu:main-93bc10a-quorum@sha256:37158bdcdffe378060994b07135759b782bb6aff6f38e0e36ab53c1321309d50
env:
- name: WORKER_ID
value: "rt"
Expand Down
20 changes: 10 additions & 10 deletions deploy/hive-oke/values.yaml
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# Vexa Helm values for the shared Quorum meeting service on hive-oke (namespace `meetings`).
#
# Chart: deploy/helm/charts/vexa from hivecommons/vexa main at commit 3addb0a940acc42e899b2748711c93dab2467874 (chart 0.13.4,
# Chart: deploy/helm/charts/vexa from hivecommons/vexa main at commit 93bc10a5d2972197642a960e885b356a5063d50a (chart 0.13.4,
# appVersion 0.13.4). Install with ./install.sh, which also creates the Secrets and applies the
# namespace policy rendered from the `quorum` block below.
#
Expand All @@ -18,7 +18,7 @@
# ---- Quorum deployment knobs (read by install.sh, ignored by the chart) ----
quorum:
# The hivecommons/vexa commit the chart is taken from and every Vexa image is built from.
vexaCommit: 3addb0a940acc42e899b2748711c93dab2467874
vexaCommit: 93bc10a5d2972197642a960e885b356a5063d50a
# Most meeting-bot Pods that may run at once in `meetings`. Enforced twice:
# 1. ResourceQuota `meetings-quota`: pods = controlPlanePods + maxConcurrentBots
# (bot Pods are bare Pods the runtime creates, so a quota cannot select only them);
Expand Down Expand Up @@ -72,27 +72,27 @@ mcp:
enabled: false
image:
repository: ghcr.io/hivecommons/vexa-mcp
tag: main-3addb0a-quorum@sha256:d4f802bce02ff8e22022b4f708651d6ba2459772f231b54d09fd73f037d3a84e
tag: main-93bc10a-quorum@sha256:dcd8ecc676c11e556abfbd4782e49dc5418f871127d2a05c5d8e8d553bb64611

gateway:
replicaCount: 2
image:
repository: ghcr.io/hivecommons/vexa-gateway
tag: main-3addb0a-quorum@sha256:ac40a5d374a8fef37dd4df7a31dc97b83e56555e5b636bf42e58a6d3f4a6691e
tag: main-93bc10a-quorum@sha256:76803e7958da25f9b989b554773d65bc8390c3d291ee80c10a0756545e2e632d
service:
type: ClusterIP

adminApi:
replicaCount: 2
image:
repository: ghcr.io/hivecommons/vexa-admin-api
tag: main-3addb0a-quorum@sha256:5ecf337d6d63c415bc2a0ce6a78e02df8c71e3dc8ecc28cc9486bef3c22a0e63
tag: main-93bc10a-quorum@sha256:0db02dc2551139e6b74561839f2e621eb73570e49cb702e94c97d4dea1ed323b

meetingApi:
replicaCount: 2
image:
repository: ghcr.io/hivecommons/vexa-meeting-api
tag: main-3addb0a-quorum@sha256:6f7e881be4f643bdd3b860c50635270e2c3d597f7ad80b476532bee832b3e569
tag: main-93bc10a-quorum@sha256:bd6a7aeaa034c1dd760aa84c79588c783845bd8c73302385384b74795688edf8
# Participants see this name; it says the bot takes notes (RFC: behaviour in the meeting).
defaultBotName: "Hive notes (AI notetaker)"
transcriptionServiceUrl: http://vexa-transcription:8000
Expand All @@ -117,16 +117,16 @@ runtime:
replicaCount: 1
image:
repository: ghcr.io/hivecommons/vexa-runtime
tag: main-3addb0a-quorum@sha256:d5a51457840a0e349415e266aa21ad706a75daed9c640c645454b2a620ee3529
tag: main-93bc10a-quorum@sha256:1c815b101a91dfd6b269030bc1a6223d708043efe9eadb533eb4f116cd51b020
# The meeting bot, from the same commit as everything else (hivecommons/vexa#15: lines are
# synthesized and played sentence by sentence from a per-sentence cache, the microphone is
# opened once at join and carries only the speech clip, the Meet final follows the pause by
# one transcription round trip, Teams utterances cross whole and Teams chat text is typed in
# and checked before send). install.sh also pre-pulls this image on every bot node.
browserImage: ghcr.io/hivecommons/vexa-bot:main-3addb0a-quorum@sha256:5ae0a60e4123d7e236b9896f882005201273dc9eb474582b955e9d685d28bda7
browserImage: ghcr.io/hivecommons/vexa-bot:main-93bc10a-quorum@sha256:2170dbde3c706cea6214a925c2b89c2341b0b9fd1eebc89ba95f829fb44dd511
# Agent workers are not used by Quorum (agentApi is off); pinned anyway so nothing is mutable.
agentImage: ghcr.io/hivecommons/vexa-agent-api:main-3addb0a-quorum@sha256:f3e38905b1ed0c2d53e5cc8c205b1d27349b30e5c74df842d01cf6ea2ec18de3
agentWorkerImage: ghcr.io/hivecommons/vexa-agent-worker:main-3addb0a-quorum@sha256:4bb8d561bd1e386c2ef3d1539437e133e74eb3342056da3366fe03570e74ddf0
agentImage: ghcr.io/hivecommons/vexa-agent-api:main-93bc10a-quorum@sha256:86db5f578b637f248f200ca9a7f88c794179b0c63a9389fbe944c5a689c6aac9
agentWorkerImage: ghcr.io/hivecommons/vexa-agent-worker:main-93bc10a-quorum@sha256:c40bd70cfb7721036398e0f14e7e529de5b2a16fd52594aff9248903b7a2470f
# Each spawned bot Pod: Guaranteed QoS, request == limit. Sizing per upstream production
# guidance (deploy docs: ~1.0-1.3Gi steady, Teams spikes past 2Gi; do not go below 2560Mi).
workloadResources:
Expand Down
Loading