fix(strix): qualify contextual gateway model for LiteLLM

docs/doctoring/contextual-orchestrator-vendored-sidecar.md

@@ -88,3 +88,19 @@ the full local suite passed with `1689 passed, 1 skipped, 16 subtests passed`.
 The PR-target Noema check still runs the trusted base copy until this trusted
 workflow change is merged, so its reproduction of the old error is retained as
 bootstrap evidence rather than treated as a current-head runtime result.
+
+## 2026-08-28 post-#1370 runtime correction
+
+Main push run `33141468804` confirmed that the catalog envelope correction
+reached the Strix sidecar, but LiteLLM rejected the child model
+`orchestrator/free` because it had no provider prefix. Follow-up commits
+`9f58d74` and `5aa0a20` map only the pinned gateway request to
+`openai/orchestrator/free`, fail closed when that gateway base is absent or
+not loopback, and keep the loopback sidecar receiving `orchestrator/free`.
+
+The same follow-up masks the dynamic sidecar bearer before writing `GITHUB_ENV`
+and rejects carriage returns/newlines in an override. This closes the runtime
+log exposure observed in the Noema step environment block. Focused contracts
+pass (`32 passed`) and the full local suite passes (`1689 passed, 1 skipped`).
+The main Strix rerun and an independently authorized Noema model verdict are
+still required before claiming end-to-end review completion.

docs/product-technical-gap-baseline.md

@@ -240,8 +240,9 @@ flowchart LR
 
 ## 2026-08-28 current-main routing and runtime recheck
 
-- Current protected-main candidate is `f8823a544c3c4c046977f8511f683e85f83eb496`,
-  the merge commit for #1364. #1360 is merged at
+- Current protected main is `24ee38b097dbfc1a895e1199ade48cff36431d05`,
+  the merge commit for #1370. #1364 is merged at
+  `f8823a544c3c4c046977f8511f683e85f83eb496`; #1360 is merged at
   `17052a7ca3c16db90932a4d6036b43165ddee418`.
 - The current Required OpenCode dispatch, `noema-review.yml`, `strix.yml`,
   and write-capable `pr-review-autofix.yml` all provision the pinned
@@ -260,11 +261,31 @@ flowchart LR
   in commit `861463c11a7ca8b1f9179073e2a3db9eba5aa5ab`; its current head is
   `38e0307c655823a1e474b29aae89f8cfcb1edbc0`. Focused tests and the full local
   suite pass (`1689 passed, 1 skipped, 16 subtests passed`).
-- #1370 remains open and blocked against main `f8823a5`. Its PR-target Noema
-  run `33140830199` executes the trusted base launcher and reproduces the
-  pre-fix bare-list error; its `opencode-review` check fails closed because no
-  current-head OpenCode verdict exists. These are bootstrap evidence gaps,
-  not proof that the #1370 catalog-envelope patch fails.
+- #1370 merged on `24ee38b…`; its pre-merge PR-target Noema run
+  `33140830199` executed the trusted base launcher and reproduced the
+  pre-fix bare-list error. The post-merge push run below shows that the
+  catalog-envelope fix reached the Strix sidecar successfully.
+
+## 2026-08-28 post-#1370 Strix runtime recheck
+
+- Main push run `33141468804` reached `Provision contextual-orchestrator Strix
+  sidecar` successfully, then failed in `Run Strix (quick)`. LiteLLM rejected
+  the unqualified child model `orchestrator/free` with `LLM Provider NOT
+  provided`; this is a request-shape defect, not evidence that the sidecar
+  catalog failed.
+- Follow-up commits `9f58d74` and `5aa0a20` qualify only the LiteLLM child
+  request as `openai/orchestrator/free` when the API base is the pinned
+  loopback gateway; the gateway still receives `orchestrator/free` and owns
+  discovery/failover. They also fail closed when that base is absent or not
+  the pinned loopback, register the dynamic bearer token with `::add-mask::`
+  before exporting `GITHUB_ENV`, and reject token newlines. Focused contracts
+  pass (`32 passed`); the full local suite passes (`1689 passed, 1 skipped`).
+- A real current-main Noema run on #1369 (`33141494393`) also booted the
+  sidecar and executed the Noema gate, but correctly skipped the LLM verdict
+  because the exact head had no primary OpenCode approval. A successful
+  sidecar/bootstrap step is not counted as a model-review result; post-fix
+  Strix completion and an independently authorized Noema verdict remain
+  separate evidence items.
 
 ## 5. 실행 루프와 고객의 다음 행동
 

scripts/ci/contextual_orchestrator_review_sidecar.sh

@@ -45,6 +45,14 @@ fi
 log "provider secrets present: $provider_secret_count of 5"
 
 ORCHESTRATOR_TOKEN="${ORCHESTRATOR_TOKEN:-$(python3 -c 'import secrets; print(secrets.token_urlsafe(32))')}"
+if [[ "$ORCHESTRATOR_TOKEN" == *$'\r'* || "$ORCHESTRATOR_TOKEN" == *$'\n'* ]]; then
+  fail "ORCHESTRATOR_TOKEN must not contain carriage returns or newlines"
+fi
+if [ -n "${GITHUB_ACTIONS:-}" ]; then
+  # Register the process-local bearer token before exporting it through
+  # GITHUB_ENV; later step environment blocks otherwise echo it verbatim.
+  echo "::add-mask::$ORCHESTRATOR_TOKEN"
+fi
 
 mkdir -p "$ORCHESTRATOR_WORK"
 rm -rf "$ORCHESTRATOR_SOURCE"

scripts/ci/strix_quick_gate.sh

@@ -2499,6 +2499,10 @@ resolved_llm_api_base_for_model() {
 	fi
 
 	if [ -z "$api_base_file" ]; then
+		if is_contextual_orchestrator_model "$model"; then
+			echo "ERROR: contextual-orchestrator Strix scans require LLM_API_BASE_FILE to select the pinned loopback gateway." >&2
+			return 2
+		fi
 		if is_github_models_model "$model"; then
 			echo "ERROR: GitHub Models Strix scans require LLM_API_BASE_FILE to select the GitHub Models inference endpoint." >&2
 			return 2
@@ -2516,8 +2520,17 @@ resolved_llm_api_base_for_model() {
 	llm_api_base_value="${llm_api_base_value%%:generateContent*}"
 	llm_api_base_value="$(trim_whitespace "$llm_api_base_value")"
 	if [ -z "$llm_api_base_value" ]; then
+		if is_contextual_orchestrator_model "$model"; then
+			echo "ERROR: contextual-orchestrator Strix scans require a non-empty pinned loopback API base." >&2
+			return 2
+		fi
 		return 0
 	fi
+	if is_contextual_orchestrator_model "$model" &&
+		! is_contextual_orchestrator_api_base "$llm_api_base_value"; then
+		echo "ERROR: contextual-orchestrator Strix scans require the pinned loopback API base." >&2
+		return 2
+	fi
 	if [[ "$llm_api_base_value" =~ [[:space:][:cntrl:]] ]]; then
 		echo "ERROR: LLM_API_BASE must not contain whitespace or control characters." >&2
 		return 2
@@ -2547,6 +2560,16 @@ child_model_for_api_base() {
 	local model="$1"
 	local llm_api_base_value="$2"
 
+	# LiteLLM requires an explicit provider prefix even when the gateway is an
+	# OpenAI-compatible local endpoint. Keep the public gateway model name, but
+	# qualify only the child process model so the request still carries
+	# orchestrator/free to contextual-orchestrator.
+	if is_contextual_orchestrator_model "$model" &&
+		is_contextual_orchestrator_api_base "$llm_api_base_value"; then
+		printf '%s\n' 'openai/orchestrator/free'
+		return 0
+	fi
+
 	if [ -n "$llm_api_base_value" ] && is_github_models_api_base "$llm_api_base_value"; then
 		case "$model" in
 		github_models/openai/*)

scripts/ci/test_strix_quick_gate.sh

@@ -3321,6 +3321,18 @@ success|runtime-env-forwarding|custom-openai-compatible-preserves-effort|vertex-
 		echo "scan ok"
 		exit 0
 		;;
+	contextual-orchestrator-gateway-model-qualification)
+		if [ "${STRIX_LLM:-}" != "openai/orchestrator/free" ]; then
+			echo "gateway model was not provider-qualified for LiteLLM" >&2
+			exit 10
+		fi
+		if [ "${LLM_API_BASE:-}" != "http://127.0.0.1:18080/v1" ]; then
+			echo "gateway API base was not preserved" >&2
+			exit 11
+		fi
+		echo "scan ok through contextual-orchestrator gateway"
+		exit 0
+		;;
 	scan-working-directory-isolated)
 		if [ "$PWD" = "$target_path" ] || [[ "$PWD" == "$target_path"/* ]]; then
 			echo "Error: Strix process inherited the untrusted scan target as cwd" >&2
@@ -6068,6 +6080,30 @@ run_filtered_gate_case_if_requested() {
 			"vertex_ai/ready-primary" \
 			"<unset>"
 		;;
+	contextual-orchestrator-missing-api-base-fails-closed)
+		run_gate_case "contextual-orchestrator-missing-api-base-fails-closed" \
+			"orchestrator/free" \
+			"" \
+			"2" \
+			"require LLM_API_BASE_FILE to select the pinned loopback gateway" \
+			"0" \
+			"" \
+			"" \
+			"contextual_orchestrator" \
+			""
+		;;
+	contextual-orchestrator-gateway-model-qualification)
+		run_gate_case "contextual-orchestrator-gateway-model-qualification" \
+			"orchestrator/free" \
+			"" \
+			"0" \
+			"scan ok through contextual-orchestrator gateway" \
+			"1" \
+			"openai/orchestrator/free" \
+			"http://127.0.0.1:18080/v1" \
+			"contextual_orchestrator" \
+			"http://127.0.0.1:18080/v1"
+		;;
 	pr-rust-workspace-context)
 		run_gate_case "pr-rust-workspace-context" \
 			"openai/gpt-4o-mini" \
@@ -9747,6 +9783,28 @@ run_gate_case "success" \
 	"vertex_ai/ready-primary" \
 	"<unset>"
 
+run_gate_case "contextual-orchestrator-missing-api-base-fails-closed" \
+	"orchestrator/free" \
+	"" \
+	"2" \
+	"require LLM_API_BASE_FILE to select the pinned loopback gateway" \
+	"0" \
+	"" \
+	"" \
+	"contextual_orchestrator" \
+	""
+
+run_gate_case "contextual-orchestrator-gateway-model-qualification" \
+	"orchestrator/free" \
+	"" \
+	"0" \
+	"scan ok through contextual-orchestrator gateway" \
+	"1" \
+	"openai/orchestrator/free" \
+	"http://127.0.0.1:18080/v1" \
+	"contextual_orchestrator" \
+	"http://127.0.0.1:18080/v1"
+
 run_gate_case "success-with-critical-report" \
 	"vertex_ai/ready-primary" \
 	"" \

tests/test_contextual_orchestrator_review_sidecar_contract.py

@@ -82,6 +82,11 @@ def test_sidecar_feeds_discovery_and_policy_artifacts_to_the_launcher() -> None:
 def test_sidecar_exports_gateway_env_for_review_steps() -> None:
     """The gateway address and bearer token land in GITHUB_ENV for later steps."""
     text = _read(SIDECAR)
+    assert 'echo "::add-mask::$ORCHESTRATOR_TOKEN"' in text
+    assert "ORCHESTRATOR_TOKEN must not contain carriage returns or newlines" in text
+    assert text.index('echo "::add-mask::$ORCHESTRATOR_TOKEN"') < text.index(
+        'if [ -n "$ORCHESTRATOR_GITHUB_ENV" ]; then'
+    )
     assert "CONTEXTUAL_ORCHESTRATOR_BASE_URL=http://%s:%s\\n' \"$ORCHESTRATOR_HOST\" \"$ORCHESTRATOR_PORT\"" in text
     assert "CONTEXTUAL_ORCHESTRATOR_TOKEN=%s\\n' \"$ORCHESTRATOR_TOKEN\"" in text
     assert "CONTEXTUAL_ORCHESTRATOR_EVIDENCE=%s\\n' \"$policy_report\"" in text