From 6222744119e52efdbd086a07282ea8357301b571 Mon Sep 17 00:00:00 2001
From: Codex <codex@noreply.local>
Date: Wed, 1 Jul 2026 16:39:40 +0000
Subject: [PATCH] fix: surface provider decision in otel diagnostics

---
 .../apply-status-scripts.ts                   |  3 +-
 .../diagnose-code-agent-script.ts             | 83 +++++++++++++++++++
 .../platform-infra-observability/render.ts    | 27 ++++++
 3 files changed, 112 insertions(+), 1 deletion(-)

diff --git a/scripts/src/platform-infra-observability/apply-status-scripts.ts b/scripts/src/platform-infra-observability/apply-status-scripts.ts
index 62d19c00..60e26529 100644
--- a/scripts/src/platform-infra-observability/apply-status-scripts.ts
+++ b/scripts/src/platform-infra-observability/apply-status-scripts.ts
@@ -90,6 +90,7 @@ export function compactDiagnoseCodeAgentResult(value: unknown): Record<string, u
     observabilityGap: source.observabilityGap ?? null,
     businessTraceIds: source.businessTraceIds ?? null,
     businessTraceScope: source.businessTraceScope ?? null,
+    providerDecision: source.providerDecision ?? null,
     identity: compactDiagnoseIdentity(source.identity),
     agentrun: compactDiagnoseAgentRun(source.agentrun),
     hwlabReadModel: source.hwlabReadModel ?? null,
@@ -252,7 +253,7 @@ export function compactSpanList(value: unknown, limit: number): unknown[] {
     return {
       name: span.name ?? null,
       service: span.service ?? null,
-      attributes: compactRecord(attrs, ["failureKind", "terminalStatus", "status", "eventType", "idleMs", "waitingFor", "lastEventLabel", "http.route", "http.status_code", "http.response.status_code", "workbench.session_id", "workbench.trace_id", "workbench.turn_id", "workbench.read_model.route", "workbench.read_model.count", "workbench.read_model.family", "workbench.read_model.status", "workbench.read_model.reason", "hwlab.http.stage", "hwlab.http.phase", "hwlab.http.phase.outcome", "hwlab.live_builds.service_id", "hwlab.live_builds.service_kind", "hwlab.live_builds.external", "hwlab.live_builds.deploy_manifest_status", "hwlab.live_builds.artifact_catalog_status"]),
+      attributes: compactRecord(attrs, ["failureKind", "terminalStatus", "status", "eventType", "idleMs", "waitingFor", "lastEventLabel", "http.route", "http.status_code", "http.response.status_code", "workbench.session_id", "workbench.trace_id", "workbench.turn_id", "workbench.read_model.route", "workbench.read_model.count", "workbench.read_model.family", "workbench.read_model.status", "workbench.read_model.reason", "hwlab.http.stage", "hwlab.http.phase", "hwlab.http.phase.outcome", "hwlab.live_builds.service_id", "hwlab.live_builds.service_kind", "hwlab.live_builds.external", "hwlab.live_builds.deploy_manifest_status", "hwlab.live_builds.artifact_catalog_status", "providerProfile", "model", "modelId", "providerModel", "code_agent.stage", "agent.chat.session_id", "agent.chat.provider_profile", "agent.chat.provider_profile_source", "agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model", "defaultProviderProfile", "adapter", "adapterEnabled", "agentRunManagerHost", "agentRunRunnerNamespace", "agentRunSourceCommitPresent"]),
     };
   });
 }
diff --git a/scripts/src/platform-infra-observability/diagnose-code-agent-script.ts b/scripts/src/platform-infra-observability/diagnose-code-agent-script.ts
index cc67ce9c..b4cd7e6b 100644
--- a/scripts/src/platform-infra-observability/diagnose-code-agent-script.ts
+++ b/scripts/src/platform-infra-observability/diagnose-code-agent-script.ts
@@ -90,6 +90,13 @@ IMPORTANT_ATTRS = [
     "error.code", "error.category", "error.layer",
     "stage", "causeStage", "causeCode",
     "eventType", "runnerId", "attemptId", "backendProfile",
+    "providerProfile", "model", "modelId", "providerModel",
+    "code_agent.stage", "agent.chat.session_id",
+    "agent.chat.provider_profile", "agent.chat.provider_profile_source",
+    "agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model",
+    "defaultProviderProfile", "adapter", "adapterEnabled",
+    "agentRunManagerHost", "agentRunRunnerNamespace",
+    "agentRunSourceCommitPresent",
     "sourceCommit", "jobName", "podName", "logPath",
     "toolName", "type", "itemType", "itemId", "status", "exitCode",
     "durationMs", "cwd", "processId", "command", "commandFingerprint",
@@ -407,6 +414,13 @@ IMPORTANT_ATTRS = [
     "error.code", "error.category", "error.layer",
     "stage", "causeStage", "causeCode",
     "eventType", "runnerId", "attemptId", "backendProfile",
+    "providerProfile", "model", "modelId", "providerModel",
+    "code_agent.stage", "agent.chat.session_id",
+    "agent.chat.provider_profile", "agent.chat.provider_profile_source",
+    "agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model",
+    "defaultProviderProfile", "adapter", "adapterEnabled",
+    "agentRunManagerHost", "agentRunRunnerNamespace",
+    "agentRunSourceCommitPresent",
     "sourceCommit", "jobName", "podName", "logPath",
     "toolName", "type", "itemType", "itemId", "status", "exitCode",
     "durationMs", "cwd", "processId", "command", "commandFingerprint",
@@ -972,6 +986,13 @@ IMPORTANT_ATTRS = [
     "error.code", "error.category", "error.layer",
     "stage", "causeStage", "causeCode",
     "eventType", "runnerId", "attemptId", "backendProfile",
+    "providerProfile", "model", "modelId", "providerModel",
+    "code_agent.stage", "agent.chat.session_id",
+    "agent.chat.provider_profile", "agent.chat.provider_profile_source",
+    "agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model",
+    "defaultProviderProfile", "adapter", "adapterEnabled",
+    "agentRunManagerHost", "agentRunRunnerNamespace",
+    "agentRunSourceCommitPresent",
     "sourceCommit", "jobName", "podName", "logPath",
     "toolName", "type", "itemType", "itemId", "status", "exitCode",
     "durationMs", "cwd", "processId", "command", "commandFingerprint",
@@ -1412,6 +1433,13 @@ def tiny_span(item):
         "hwlab.live_builds.service_id", "hwlab.live_builds.service_kind",
         "hwlab.live_builds.external", "hwlab.live_builds.deploy_manifest_status",
         "hwlab.live_builds.artifact_catalog_status",
+        "providerProfile", "model", "modelId", "providerModel",
+        "code_agent.stage", "agent.chat.session_id",
+        "agent.chat.provider_profile", "agent.chat.provider_profile_source",
+        "agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model",
+        "defaultProviderProfile", "adapter", "adapterEnabled",
+        "agentRunManagerHost", "agentRunRunnerNamespace",
+        "agentRunSourceCommitPresent",
     ):
         if key in attrs:
             keep_attrs[key] = attrs[key]
@@ -1450,6 +1478,55 @@ def identity_from_spans(spans):
             identity[key] = preferred if preferred is not None else fallback
     return identity
 
+def attr_first(attrs, keys):
+    if not isinstance(attrs, dict):
+        return None
+    for key in keys:
+        value = attrs.get(key)
+        if value not in (None, ""):
+            return value
+    return None
+
+def provider_decision_summary(spans):
+    rows = []
+    for item in spans:
+        attrs = item.get("attributes", {}) if isinstance(item.get("attributes"), dict) else {}
+        name = str(item.get("name") or "")
+        if name != "provider_decision" and attr_first(attrs, ("agent.chat.provider_profile", "providerProfile", "defaultProviderProfile")) in (None, ""):
+            continue
+        row = {
+            "span": name or None,
+            "service": item.get("service"),
+            "traceId": attr_first(attrs, ("traceId", "workbench.trace_id", "workbench.turn_id")),
+            "sessionId": attr_first(attrs, ("agent.chat.session_id", "sessionId", "workbench.session_id")),
+            "providerProfile": attr_first(attrs, ("agent.chat.provider_profile", "providerProfile", "backendProfile", "defaultProviderProfile")),
+            "providerProfileSource": attr_first(attrs, ("agent.chat.provider_profile_source", "providerProfileSource")),
+            "defaultProviderProfile": attr_first(attrs, ("defaultProviderProfile",)),
+            "model": attr_first(attrs, ("agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model", "model", "modelId", "providerModel")),
+            "adapter": attr_first(attrs, ("adapter",)),
+            "adapterEnabled": attr_first(attrs, ("adapterEnabled",)),
+            "managerHost": attr_first(attrs, ("agentRunManagerHost",)),
+            "runnerNamespace": attr_first(attrs, ("agentRunRunnerNamespace",)),
+            "sourceCommitPresent": attr_first(attrs, ("agentRunSourceCommitPresent",)),
+        }
+        compact = {key: value for key, value in row.items() if value not in (None, "")}
+        if compact:
+            rows.append(compact)
+    primary = rows[0] if rows else {}
+    return {
+        "providerProfile": primary.get("providerProfile"),
+        "providerProfileSource": primary.get("providerProfileSource"),
+        "defaultProviderProfile": primary.get("defaultProviderProfile"),
+        "model": primary.get("model"),
+        "adapter": primary.get("adapter"),
+        "adapterEnabled": primary.get("adapterEnabled"),
+        "managerHost": primary.get("managerHost"),
+        "runnerNamespace": primary.get("runnerNamespace"),
+        "sourceCommitPresent": primary.get("sourceCommitPresent"),
+        "spanCount": len(rows),
+        "decisions": rows[:5],
+    }
+
 def span_business_trace_id(item):
     attrs = item.get("attributes", {}) if isinstance(item.get("attributes"), dict) else {}
     for key in ("traceId", "workbench.trace_id", "workbench.turn_id"):
@@ -2036,6 +2113,7 @@ def candidate_score(trace_id, meta, trace_body, trace_rc, trace_err):
         "terminalStatus": agentrun.get("terminalStatus"),
         "errorSpanCount": len(error_spans),
         "businessTraceScope": business_scope,
+        "providerDecision": provider_decision_summary(spans),
         "candidateQuality": candidate_quality,
         "lowConfidence": candidate_quality != "normal",
         "rootTraceName": meta_root_name or None,
@@ -2187,6 +2265,7 @@ idle_warning_spans = [item for item in spans if str(item.get("name") or "") == "
 http_summary = http_status_summary(spans)
 read_model = hwlab_read_model_summary(spans)
 identity = identity_from_spans(spans)
+provider_decision = provider_decision_summary(spans)
 agentrun_authority = agentrun_authority_summary(identity)
 agentrun = agentrun_summary(spans, agentrun_authority)
 lag = projection_lag_summary(agentrun, read_model)
@@ -2254,6 +2333,9 @@ summary = {
         "failureKind": agentrun.get("failureKind"),
         "observabilityGap": observability_gap.get("status"),
         "businessTraceScope": business_scope.get("mode"),
+        "providerProfile": provider_decision.get("providerProfile"),
+        "providerModel": provider_decision.get("model"),
+        "providerAdapter": provider_decision.get("adapter"),
     },
 }
 evidence = {
@@ -2284,6 +2366,7 @@ payload = {
     "observabilityGap": observability_gap,
     "businessTraceIds": business_trace_ids[:20],
     "businessTraceScope": business_scope,
+    "providerDecision": provider_decision,
     "identity": identity,
     "agentrun": {
         "terminalStatus": agentrun.get("terminalStatus"),
diff --git a/scripts/src/platform-infra-observability/render.ts b/scripts/src/platform-infra-observability/render.ts
index 395498b8..2a2a31c9 100644
--- a/scripts/src/platform-infra-observability/render.ts
+++ b/scripts/src/platform-infra-observability/render.ts
@@ -382,6 +382,7 @@ export function renderTraceTable(input: {
   const turnRows = traceTurnStatusRows(allSpans);
   const projectionRows = traceProjectionRows(allSpans);
   const projectionSyncRows = traceProjectionSyncRows(allSpans);
+  const providerRows = traceProviderDecisionRows(allSpans);
   const countRows = asArray(input.result.spanNameCounts).slice(0, 8).map((item) => {
     const row = asPlainRecord(item) ?? {};
     return [shortenEnd(textValue(row.name), 64), textValue(row.count)];
@@ -406,6 +407,9 @@ export function renderTraceTable(input: {
     "Identity:",
     formatTable(["FIELD", "VALUE"], identityRows.length > 0 ? identityRows : [["-", "-"]]),
     "",
+    "Provider decision:",
+    formatTable(["SERVICE", "PROFILE", "SOURCE", "DEFAULT", "MODEL", "ADAPTER", "RUNNER_NS", "SESSION"], providerRows.length > 0 ? providerRows : [["-", "-", "-", "-", "-", "-", "-", "-"]]),
+    "",
     "Key spans:",
     formatTable(["NAME", "SERVICE", "ROUTE", "STATUS", "DUR_MS", "RES_MS", "FETCH_REQ", "REQ_WAIT", "RESP_XFER", "PROTO", "DETAIL"], spanRows.length > 0 ? spanRows : [["-", "-", "-", "-", "-", "-", "-", "-", "-", "-", "-"]]),
     "",
@@ -483,6 +487,27 @@ function traceRunnerRows(spans: Record<string, unknown>[]): string[][] {
   return dedupeRows(rows).slice(0, 12);
 }
 
+function traceProviderDecisionRows(spans: Record<string, unknown>[]): string[][] {
+  const rows = spans.filter((span) => {
+    const attrs = asPlainRecord(span.attributes);
+    return textValue(span.name) === "provider_decision"
+      || spanColumnAttr(attrs, ["agent.chat.provider_profile", "providerProfile", "defaultProviderProfile"]) !== "-";
+  }).map((span) => {
+    const attrs = asPlainRecord(span.attributes);
+    return [
+      shortenEnd(textValue(span.service), 20),
+      shortenEnd(spanColumnAttr(attrs, ["agent.chat.provider_profile", "providerProfile", "backendProfile", "defaultProviderProfile"]), 22),
+      shortenEnd(spanColumnAttr(attrs, ["agent.chat.provider_profile_source", "providerProfileSource"]), 12),
+      shortenEnd(spanColumnAttr(attrs, ["defaultProviderProfile"]), 22),
+      shortenEnd(spanColumnAttr(attrs, ["agent.chat.model", "agent.chat.model_id", "agent.chat.provider_model", "model", "modelId", "providerModel"]), 28),
+      shortenEnd(spanColumnAttr(attrs, ["adapter"]), 16),
+      shortenEnd(spanColumnAttr(attrs, ["agentRunRunnerNamespace"]), 16),
+      shortenMiddle(spanColumnAttr(attrs, ["agent.chat.session_id", "sessionId", "workbench.session_id"]), 24),
+    ];
+  });
+  return dedupeRows(rows).slice(0, 8);
+}
+
 function traceReadWindowRows(spans: Record<string, unknown>[]): string[][] {
   const rows = spans.filter((span) => textValue(span.name) === "trace_events_read").map((span) => {
     const attrs = asPlainRecord(span.attributes);
@@ -678,6 +703,7 @@ export function renderDiagnoseCodeAgentTable(input: {
   const evidence = asPlainRecord(input.result.evidence);
   const observabilityGap = asPlainRecord(input.result.observabilityGap);
   const servicePath = asPlainRecord(input.result.servicePath);
+  const providerDecision = asPlainRecord(input.result.providerDecision);
   const rootCauses = asArray(input.result.rootCauseCandidates).map((item) => asPlainRecord(item) ?? {});
   const http = asPlainRecord(input.result.http);
   const services = joinValues(input.result.services, 54);
@@ -727,6 +753,7 @@ export function renderDiagnoseCodeAgentTable(input: {
     `  requested runId=${requestedRunId} commandId=${requestedCommandId} sessionId=${requestedSessionId} runnerJobId=${requestedRunnerJobId}`,
     `  observed runId=${observedRunId} commandId=${observedCommandId} sessionId=${observedSessionId} runnerJobId=${observedRunnerJobId} runnerId=${textValue(identity?.runnerId)}`,
     `  backendProfile=${textValue(identity?.backendProfile)} sourceCommit=${shortenMiddle(textValue(identity?.sourceCommit), 20)}`,
+    `  providerProfile=${textValue(providerDecision?.providerProfile)} model=${textValue(providerDecision?.model)} source=${textValue(providerDecision?.providerProfileSource)} adapter=${textValue(providerDecision?.adapter)} runnerNamespace=${textValue(providerDecision?.runnerNamespace)}`,
     "",
     "Root causes:",
     formatTable(["CODE", "CONF", "SUMMARY"], rootRows.length > 0 ? rootRows : [["-", "-", "-"]]),