Merge remote-tracking branch 'origin/main' into HEAD

Author: masterkain

.drone.jsonnet

@@ -2,7 +2,7 @@ local releaseBranch = 'release-please--branches--main--components--codex-pooler'
 local registry = 'registry.icorete.ch';
 local image = 'registry.icorete.ch/icoretech/codex-pooler';
 local helmVersion = 'v4.2.2';
-local nodeImage = 'node:26.3.1-slim';
+local nodeImage = 'node:26.4.0-slim';
 
 [
   {

.github/workflows/docs.yml

@@ -24,7 +24,7 @@ jobs:
       - name: Setup Node.js
         uses: actions/setup-node@v6
         with:
-          node-version: '26.3.1'
+          node-version: '26.4.0'
           cache: npm
           cache-dependency-path: docs-site/package-lock.json
 
@@ -58,7 +58,7 @@ jobs:
       - name: Setup Node.js
         uses: actions/setup-node@v6
         with:
-          node-version: '26.3.1'
+          node-version: '26.4.0'
           cache: npm
           cache-dependency-path: docs-site/package-lock.json
 

.release-please-manifest.json

@@ -1,3 +1,3 @@
 {
-  ".": "0.4.3"
+  ".": "0.4.4"
 }

CHANGELOG.md

@@ -1,5 +1,23 @@
 # Changelog
 
+## [0.4.4](https://github.com/icoretech/codex-pooler/compare/codex-pooler-v0.4.3...codex-pooler-v0.4.4) (2026-06-27)
+
+
+### Features
+
+* **gateway:** coordinate websocket rollout drains ([5c990ad](https://github.com/icoretech/codex-pooler/commit/5c990ad10867467f2929e5f2ee8369a319e08433))
+
+
+### Bug Fixes
+
+* **accounting:** exclude unknown usage from consumption totals ([4d32260](https://github.com/icoretech/codex-pooler/commit/4d322607bf4e49572b74a665395635822d16c647))
+* **accounting:** exclude unknown usage from reporting totals ([7e0a0a2](https://github.com/icoretech/codex-pooler/commit/7e0a0a2af5003a65f90332c3be3b6555151c672d))
+* **deps:** update dependency daisyui to ^5.6.0 ([#85](https://github.com/icoretech/codex-pooler/issues/85)) ([724a4e6](https://github.com/icoretech/codex-pooler/commit/724a4e6ebbecb8df6f08cae8ab74b42775915c28))
+* **gateway:** classify public sse transport interruptions ([e03967d](https://github.com/icoretech/codex-pooler/commit/e03967d3ed2894701a4e833cca58bf31d6d6ca72))
+* **gateway:** reject websocket starts during rollout drain ([bf5b3a8](https://github.com/icoretech/codex-pooler/commit/bf5b3a82aa50dc519bd1b9f3d4f6d483d9f76332))
+* **operations:** mark readiness unavailable during rollout drain ([59e0f34](https://github.com/icoretech/codex-pooler/commit/59e0f34e0f1f347a8cd04c1195668e2aa67aff23))
+* **websocket:** finalize rollout drains as owner drained ([166879e](https://github.com/icoretech/codex-pooler/commit/166879efa6c47cc7786830d301c240f8cfb6f8ee))
+
 ## [0.4.3](https://github.com/icoretech/codex-pooler/compare/codex-pooler-v0.4.2...codex-pooler-v0.4.3) (2026-06-25)
 
 

Dockerfile

@@ -1,7 +1,7 @@
 ARG DEBIAN_MIRROR=
 ARG DEBIAN_SECURITY_MIRROR=
 
-FROM node:26.3.1-slim AS assets_deps
+FROM node:26.4.0-slim AS assets_deps
 
 ENV NPM_CONFIG_UPDATE_NOTIFIER=false
 

README.md

@@ -473,7 +473,7 @@ model:
   provider: openai-api
   base_url: http://localhost:4000/v1
   api_mode: codex_responses
-  context_length: 272000
+  context_length: 258400
   supports_vision: true
 
 agent:
@@ -495,10 +495,11 @@ mcp_servers:
 ```
 
 Current Codex Pooler releases also expose an SDK-readable `context_length` value
-on `/v1/models`, derived from the Codex `context_window` metadata, so Hermes'
-automatic probes can resolve the 272k Pooler window. Keep `context_length: 272000`
-in Hermes config as a safe explicit override for older Pooler deployments or stale
-model metadata.
+on `/v1/models`, derived from the effective Codex `context_window` metadata, so
+Hermes' automatic probes can resolve the Pooler window. For `gpt-5.5`, the raw
+Codex window remains 272000 and the effective advertised value is currently
+258400. Keep `context_length: 258400` in Hermes config as a safe explicit
+override for older Pooler deployments or stale model metadata.
 
 Remote HTTP MCP servers require Hermes' `mcp` extra. If
 `hermes mcp test codex_pooler` reports `mcp.client.streamable_http is not
@@ -531,7 +532,7 @@ model:
   default: gpt-5.5
   provider: openai-codex
   base_url: http://localhost:4000/v1
-  context_length: 272000
+  context_length: 258400
   supports_vision: true
 
 agent:
@@ -1503,8 +1504,8 @@ Choose the deployment path that matches how you want to operate Codex Pooler:
 The Kubernetes path uses the
 [`icoretech/codex-pooler` chart](https://github.com/icoretech/helm/tree/main/charts/codex-pooler)
 from the iCoreTech Helm repository. The chart runs one release image as separate
-web, worker, scheduler, and migration roles. For a real install, pin `image.tag`
-to an immutable release instead of using `latest`.
+web, worker, scheduler, and migration roles. For a real install, pin the chart
+`--version`; the chart defaults `image.tag` to the matching `appVersion`.
 
 ## Need more Codex?
 

assets/package-lock.json

@@ -2,7 +2,7 @@
   "dependencies": {
     "apexcharts": "^5.15.2",
     "cally": "^0.9.2",
-    "daisyui": "^5.5.23",
+    "daisyui": "^5.6.3",
     "topbar": "^3.0.1",
     "uqr": "^0.1.3"
   }

assets/package.json

@@ -4,8 +4,16 @@ if System.get_env("PHX_SERVER") in ~w(true 1) do
   config :codex_pooler, CodexPoolerWeb.Endpoint, server: true
 end
 
+websocket_drain_timeout_ms =
+  CodexPooler.Gateway.Transports.Websocket.RolloutDrain.configured_timeout_ms()
+
+endpoint_shutdown_timeout_ms = websocket_drain_timeout_ms + 5_000
+
 config :codex_pooler, CodexPoolerWeb.Endpoint,
-  http: [port: String.to_integer(System.get_env("PORT", "4000"))]
+  http: [
+    port: String.to_integer(System.get_env("PORT", "4000")),
+    thousand_island_options: [shutdown_timeout: endpoint_shutdown_timeout_ms]
+  ]
 
 config :codex_pooler,
        :websocket_owner_forwarding_enabled,
@@ -93,7 +101,8 @@ if config_env() == :prod do
     url: [host: host, port: 443, scheme: "https"],
     http: [
       port: String.to_integer(System.get_env("PORT", "4000")),
-      ip: {0, 0, 0, 0}
+      ip: {0, 0, 0, 0},
+      thousand_island_options: [shutdown_timeout: endpoint_shutdown_timeout_ms]
     ],
     secret_key_base: secret_key_base
 

config/runtime.exs

@@ -40,7 +40,7 @@ model:
   provider: openai-api
   base_url: https://codex-pooler.example.com/v1
   api_mode: codex_responses
-  context_length: 272000
+  context_length: 258400
   supports_vision: true
 
 agent:
@@ -63,7 +63,7 @@ mcp_servers:
 
 Use `http://localhost:4000/v1` and `http://localhost:4000/mcp` for a local Codex Pooler instance.
 
-Current Codex Pooler releases also expose an SDK-readable `context_length` value on `/v1/models`, derived from the Codex `context_window` metadata, so Hermes' automatic probes can resolve the 272k Pooler window. Keep `context_length: 272000` in Hermes config as a safe explicit override for older Pooler deployments or stale model metadata.
+Current Codex Pooler releases also expose an SDK-readable `context_length` value on `/v1/models`, derived from the effective Codex `context_window` metadata, so Hermes' automatic probes can resolve the Pooler window. For `gpt-5.5`, the raw Codex window remains 272000 and the effective advertised value is currently 258400. Keep `context_length: 258400` in Hermes config as a safe explicit override for older Pooler deployments or stale model metadata.
 
 Remote HTTP MCP servers require Hermes' `mcp` extra. If `hermes mcp test codex_pooler` reports `mcp.client.streamable_http is not available`, install MCP support into the Hermes environment, following the [Hermes MCP Integration docs](https://hermes-agent.nousresearch.com/docs/user-guide/features/mcp), and rerun the test.
 
@@ -93,7 +93,7 @@ model:
   default: gpt-5.5
   provider: openai-codex
   base_url: https://codex-pooler.example.com/v1
-  context_length: 272000
+  context_length: 258400
   supports_vision: true
 
 agent: