ding-labs · zuchka · May 9, 2026 · May 9, 2026
diff --git a/docs/recipes/_template.md b/docs/recipes/_template.md
@@ -12,7 +12,7 @@ rubric.
 
 ## Prerequisites
 
-- DING binary `>= v0.3.0` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - <Platform-specific requirements: account, runtime version, etc.>
 - A notifier endpoint (Slack webhook URL, custom webhook, etc.)
 
@@ -32,7 +32,7 @@ rules:
     match:
       metric: run.exit
     condition: value > 0
-    message: "Job failed (exit {{ .exit_code }})"
+    message: "Job failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: slack
 ```

diff --git a/docs/recipes/argo-workflows.md b/docs/recipes/argo-workflows.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.7.0` — see [install](../install.md). The recipe pulls the official container image `ghcr.io/ding-labs/ding:v0.7.0` (multi-arch, scratch base) into each step's Pod via an initContainer; no need to bake DING into your workload image.
+- DING binary `>= v0.10.0` — see [install](../install.md). The recipe pulls the official container image `ghcr.io/ding-labs/ding:v0.10.0` (multi-arch, scratch base) into each step's Pod via an initContainer; no need to bake DING into your workload image.
 - Argo Workflows controller installed in the cluster, `>= v3.5` (most users on v3.5/v3.6 LTS series).
 - `kubectl` access to a namespace where you can create Workflows, ConfigMaps, and Secrets.
 - `argo` CLI installed locally (ships with the controller; one-line install per [Argo docs](https://argo-workflows.readthedocs.io/en/latest/quick-start/)).
@@ -49,7 +49,7 @@ data:
         match:
           metric: run.exit
         condition: value > 0
-        message: "Argo step {{ .pod }} (workflow {{ .workflow }}) failed with exit {{ .exit_code }}"
+        message: "Argo step {{ .pod }} (workflow {{ .workflow }}) failed with exit {{ .exit_code }} after {{ .duration_seconds }}s"
         alert:
           - notifier: slack
 ---
@@ -71,7 +71,7 @@ spec:
             name: ding-config
       initContainers:
         - name: install-ding
-          image: ghcr.io/ding-labs/ding:v0.7.0
+          image: ghcr.io/ding-labs/ding:v0.10.0
           # `ding install` self-copies the binary — works against the FROM-scratch
           # release image (no /bin/sh available). Added in DING v0.5.1.
           command: ["/ding", "install", "/shared/ding"]
@@ -173,7 +173,7 @@ spec:
         - { name: ding-config, configMap: { name: ding-config } }
       initContainers:
         - name: install-ding
-          image: ghcr.io/ding-labs/ding:v0.7.0
+          image: ghcr.io/ding-labs/ding:v0.10.0
           command: ["/ding", "install", "/shared/ding"]
           mirrorVolumeMounts: true
       container:

diff --git a/docs/recipes/buildkite.md b/docs/recipes/buildkite.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.3.0` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - A Buildkite organization with at least one agent ([free trial available](https://buildkite.com/pricing))
 - A notifier endpoint (Slack webhook URL, custom webhook, etc.)
 
@@ -33,7 +33,7 @@ rules:
     match:
       metric: run.exit
     condition: value > 0
-    message: "{{ .repo }}@{{ .branch }} failed (exit {{ .exit_code }})"
+    message: "{{ .repo }}@{{ .branch }} failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: slack
 ```
@@ -64,7 +64,7 @@ Use these in `match.labels` or `message` templates. See [Configuration](../confi
 2. Trigger a build. Confirm a successful step produces no alert.
 3. Force a failure (`exit 1` in `run-tests.sh`). Confirm the alert fires in Slack within ~5 seconds of step exit.
 
-If the alert doesn't fire, check the Buildkite build log for `ding` output. Common issues: webhook URL not exposed (env hook scope, agent vs pipeline level), or `drain_timeout` shorter than the notifier retry window — see [Configuration](../configuration.md).
+If the alert doesn't fire, check the Buildkite build log for `ding` output. Common issues: webhook URL not exposed (env hook scope, agent vs pipeline level), or `drain_timeout` shorter than the notifier retry window — see [Configuration → drain_timeout](../configuration.md#drain_timeout-and-retry-behaviour-in-ding-run).
 
 ## Native Buildkite UI surfacing
 

diff --git a/docs/recipes/gitlab-ci.md b/docs/recipes/gitlab-ci.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.3.0` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - A GitLab project with CI enabled (gitlab.com or self-hosted)
 - A notifier endpoint (Slack webhook URL, custom webhook, etc.) accessible from the runner
 
@@ -35,7 +35,7 @@ rules:
     match:
       metric: run.exit
     condition: value > 0
-    message: "Pipeline {{ .branch }} failed (exit {{ .exit_code }})"
+    message: "Pipeline {{ .branch }} failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: slack
 ```
@@ -67,7 +67,7 @@ Use these in `match.labels` for selective rules, or in `message` templates as `{
 2. Push a commit. Confirm the pipeline runs and that a successful job produces no alert.
 3. Force a failure: change `run-tests.sh` to `exit 1`. Confirm the alert fires in Slack within ~5 seconds of job exit.
 
-If the alert doesn't fire, check the GitLab CI job log for `ding` output. Common issues: webhook URL not exposed to the job (mark the variable as not "Protected" if testing on a non-protected branch), or `drain_timeout` shorter than the notifier retry window — see [Configuration → drain_timeout](../configuration.md).
+If the alert doesn't fire, check the GitLab CI job log for `ding` output. Common issues: webhook URL not exposed to the job (mark the variable as not "Protected" if testing on a non-protected branch), or `drain_timeout` shorter than the notifier retry window — see [Configuration → drain_timeout](../configuration.md#drain_timeout-and-retry-behaviour-in-ding-run).
 
 ## Native GitLab UI surfacing
 

diff --git a/docs/recipes/jenkins.md b/docs/recipes/jenkins.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.3.0` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - A Jenkins controller (any version supporting Pipeline DSL — most do)
 - A notifier endpoint (Slack webhook URL, custom webhook, etc.) reachable from the Jenkins agent
 
@@ -47,7 +47,7 @@ rules:
     match:
       metric: run.exit
     condition: value > 0
-    message: "{{ .job }} build {{ .build }} failed (exit {{ .exit_code }})"
+    message: "{{ .job }} build {{ .build }} failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: slack
 ```
@@ -77,7 +77,7 @@ Note: Jenkins doesn't expose `repo`, `branch`, or `commit` as universal env vars
 2. Trigger the job. Confirm a successful build produces no alert.
 3. Force a failure (`exit 1` in `run-tests.sh`). Confirm the alert fires in Slack within ~5 seconds of build exit.
 
-If the alert doesn't fire, check the Jenkins build console for `ding` output. Common issues: webhook credential not exposed to the job (`withCredentials` block missing or wrong `credentialsId`), or `drain_timeout` shorter than the notifier retry window — see [Configuration](../configuration.md).
+If the alert doesn't fire, check the Jenkins build console for `ding` output. Common issues: webhook credential not exposed to the job (`withCredentials` block missing or wrong `credentialsId`), or `drain_timeout` shorter than the notifier retry window — see [Configuration → drain_timeout](../configuration.md#drain_timeout-and-retry-behaviour-in-ding-run).
 
 ## Tradeoffs / known limitations
 

diff --git a/docs/recipes/kubernetes-jobs.md b/docs/recipes/kubernetes-jobs.md
@@ -14,7 +14,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.7.0` — see [install](../install.md). The recipe pulls the official container image `ghcr.io/ding-labs/ding:v0.7.0` (multi-arch, scratch base) into your Pod via an initContainer; no need to bake DING into your workload image.
+- DING binary `>= v0.10.0` — see [install](../install.md). The recipe pulls the official container image `ghcr.io/ding-labs/ding:v0.10.0` (multi-arch, scratch base) into your Pod via an initContainer; no need to bake DING into your workload image.
 - A Kubernetes cluster `>= 1.21` for the primary wrapper pattern below. The sidecar alternative documented in [Configuration](#sidecar-alternative-k8s-129) requires `>= 1.29` for native sidecar lifecycle.
 - `kubectl` access to a namespace where you can create Jobs, ConfigMaps, and Secrets.
 - A notifier endpoint (Slack webhook URL or custom webhook) you can store in a Kubernetes Secret.
@@ -58,7 +58,7 @@ data:
         match:
           metric: run.exit
         condition: value > 0
-        message: "{{ .pod }} (Job {{ .job_name }}) failed with exit {{ .exit_code }}"
+        message: "{{ .pod }} (Job {{ .job_name }}) failed with exit {{ .exit_code }} after {{ .duration_seconds }}s"
         alert:
           - notifier: slack
 ---
@@ -82,7 +82,7 @@ spec:
             name: ding-config
       initContainers:
         - name: install-ding
-          image: ghcr.io/ding-labs/ding:v0.7.0
+          image: ghcr.io/ding-labs/ding:v0.10.0
           # `ding install` self-copies the binary — works against the FROM-scratch
           # release image (no /bin/sh available). Added in DING v0.5.1.
           command: ["/ding", "install", "/shared/ding"]
@@ -184,7 +184,7 @@ spec:
     spec:
       initContainers:
         - name: ding
-          image: ghcr.io/ding-labs/ding:v0.7.0
+          image: ghcr.io/ding-labs/ding:v0.10.0
           restartPolicy: Always       # native sidecar — K8s 1.29+
           command: ["/ding", "serve", "--config", "/etc/ding/ding.yaml"]
           # ...volumeMounts for config + downward-API env block
@@ -225,7 +225,7 @@ rules:
   - name: job_failed
     match: { metric: run.exit }
     condition: value > 0
-    message: "Job failed (exit {{ .exit_code }})"
+    message: "Job failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: k8s
 ```

diff --git a/docs/recipes/mlflow.md b/docs/recipes/mlflow.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.6.0` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - `mlflow >= 2.0` (`pip install mlflow`)
 - An MLflow tracking URI: local SQLite for dev; remote tracking server like Databricks or self-hosted (`mlflow server`) for production deep-links to work
 - A notifier endpoint (Slack webhook URL is the canonical example)
@@ -49,7 +49,7 @@ rules:
     match: { metric: run.exit }
     condition: value > 0
     message: |
-      MLflow run failed (exit {{ .exit_code }})
+      MLflow run failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)
       <{{ .tracking_uri }}/#/experiments/{{ .experiment_id }}/runs/{{ .run_id }}|View run in MLflow UI>
     alert:
       - notifier: slack
@@ -87,7 +87,7 @@ A Slack message during training when `val_loss` exceeds threshold:
 …and on training-process exit:
 
 > 🔔 `training_failed`
-> MLflow run failed (exit 1)
+> MLflow run failed (exit 1 after 42s)
 > [View run in MLflow UI](#)
 
 The deep-link in the second message takes you straight to the MLflow run page. All alerts are auto-tagged with `run_id`, `runner=mlflow`, `experiment_id`, `tracking_uri`.
@@ -121,7 +121,7 @@ mlflow run . --env-manager=local
 #   3. labels include run_id, experiment_id, tracking_uri
 ```
 
-If the alert doesn't fire, check the `mlflow run` log for `ding` output. Common issues: `SLACK_WEBHOOK_URL` not exported in the shell that ran `mlflow run`, or `drain_timeout` shorter than the notifier retry window — see [Configuration](../configuration.md).
+If the alert doesn't fire, check the `mlflow run` log for `ding` output. Common issues: `SLACK_WEBHOOK_URL` not exported in the shell that ran `mlflow run`, or `drain_timeout` shorter than the notifier retry window — see [Configuration → drain_timeout](../configuration.md#drain_timeout-and-retry-behaviour-in-ding-run).
 
 ## Tradeoffs / known limitations
 

diff --git a/docs/recipes/modal.md b/docs/recipes/modal.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.8.0` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - Modal CLI (`pip install modal`) authenticated via `modal token new`
 - A Modal account (free tier with $30/mo credit covers this recipe end-to-end)
 - A notifier endpoint (Slack webhook URL is the canonical example)
@@ -67,7 +67,7 @@ rules:
   - name: training_failed
     match: { metric: run.exit }
     condition: value > 0
-    message: "Modal function {{ .function_name }} (task {{ .modal_task_id }}) failed (exit {{ .exit_code }})"
+    message: "Modal function {{ .function_name }} (task {{ .modal_task_id }}) failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: slack
 ```
@@ -91,7 +91,7 @@ A Slack message during training when `val_loss` exceeds threshold:
 …and on function exit:
 
 > 🔔 `training_failed`
-> Modal function trainer (task ta-abc123def) failed (exit 1)
+> Modal function trainer (task ta-abc123def) failed (exit 1 after 287s)
 
 The `modal_task_id` matches the task ID visible in the Modal dashboard, so the Slack alert is one click away from the function's logs and metrics.
 
@@ -105,7 +105,7 @@ DING does **not** auto-detect Modal — Modal's runtime owns the container entry
 | `MODAL_FUNCTION_NAME` | The function's Python name |
 | `MODAL_APP_NAME` | The Modal `App` name (for multi-function apps) |
 
-Emit any subset as flat top-level JSON keys (DING's ingester at `internal/ingester/json.go` extracts top-level strings as labels and numbers as floats; nested objects are skipped). Use them in `match.labels` or `message` template variables. See [Configuration](../configuration.md) for the full reference.
+Emit any subset as flat top-level JSON keys — DING extracts top-level strings as labels and numbers as floats; nested objects are skipped. Use them in `match.labels` or `message` template variables. See [Configuration](../configuration.md) for the full notifier reference.
 
 ## Verification
 

diff --git a/docs/recipes/ray.md b/docs/recipes/ray.md
@@ -4,7 +4,7 @@
 
 ## Prerequisites
 
-- DING binary `>= v0.7.1` — see [install](../install.md)
+- DING binary `>= v0.10.0` — see [install](../install.md)
 - `ray >= 2.0` (`pip install "ray[default]"`; add `train`/`tune` extras as needed for your workload)
 - A running Ray cluster: local single-node (`ray start --head`) for dev; KubeRay/Anyscale/EKS for production
 - A notifier endpoint (Slack webhook URL is the canonical example)
@@ -39,7 +39,7 @@ rules:
   - name: training_failed
     match: { metric: run.exit }
     condition: value > 0
-    message: "Ray job {{ .run_id }} failed (exit {{ .exit_code }})"
+    message: "Ray job {{ .run_id }} failed (exit {{ .exit_code }} after {{ .duration_seconds }}s)"
     alert:
       - notifier: slack
 ```
@@ -112,7 +112,7 @@ A Slack message during training when `val_loss` exceeds threshold:
 …and on training-process exit:
 
 > 🔔 `training_failed`
-> Ray job raysubmit_abcdef1234567890 failed (exit 1)
+> Ray job raysubmit_abcdef1234567890 failed (exit 1 after 1843s)
 
 All Path A alerts are auto-tagged with `run_id` + `runner=ray`. The `run_id` matches the UUID printed by `ray job list`.
 
@@ -125,7 +125,7 @@ All Path A alerts are auto-tagged with `run_id` + `runner=ray`. The `run_id` mat
 | `run_id` | `RAY_JOB_ID` | Ray job UUID matching `ray job list` output |
 | `runner` | `"ray"` (set by runctx) | |
 
-Use these in `match.labels` or `message` template variables. See [Configuration](../configuration.md) for the full reference.
+Use these in `match.labels` or `message` template variables. See [Configuration](../configuration.md) for the full notifier reference.
 
 ## Verification
 
@@ -146,7 +146,7 @@ ray job list
 ray stop
 ```
 
-If the alert doesn't fire, check the Ray driver logs (`ray job logs <id>`) for `ding` output. Common issues: `SLACK_WEBHOOK_URL` not forwarded via `--runtime-env-json`, or `drain_timeout` shorter than the notifier retry window — see [Configuration](../configuration.md).
+If the alert doesn't fire, check the Ray driver logs (`ray job logs <id>`) for `ding` output. Common issues: `SLACK_WEBHOOK_URL` not forwarded via `--runtime-env-json`, or `drain_timeout` shorter than the notifier retry window — see [Configuration → drain_timeout](../configuration.md#drain_timeout-and-retry-behaviour-in-ding-run).
 
 ## Tradeoffs / known limitations