dominus-sdk-python 2.1.2__tar.gz → 2.1.3__tar.gz

{dominus_sdk_python-2.1.2 → dominus_sdk_python-2.1.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dominus-sdk-python
-Version: 2.1.2
+Version: 2.1.3
 Summary: Python SDK for the Dominus Orchestrator Platform
 Author-email: CareBridge Systems <dev@carebridge.io>
 License: Proprietary

{dominus_sdk_python-2.1.2 → dominus_sdk_python-2.1.3}/dominus/helpers/core.py

@@ -64,6 +64,7 @@ async def _get_service_jwt(psk_token: str, base_url: str) -> str:
     Get service JWT by calling /api/warden/mint with PSK.
 
     Uses circuit breaker to prevent retry storms during service outages.
+    Retries on 401/5xx with exponential backoff (orchestrator cold start handling).
 
     Args:
         psk_token: PSK token (DOMINUS_TOKEN)
@@ -73,7 +74,7 @@ async def _get_service_jwt(psk_token: str, base_url: str) -> str:
         JWT token string
 
     Raises:
-        RuntimeError: If circuit is open or auth fails
+        RuntimeError: If circuit is open or auth fails after retries
     """
     # Circuit breaker check
     if not sovereign_circuit_breaker.can_execute():
@@ -94,32 +95,85 @@ async def _get_service_jwt(psk_token: str, base_url: str) -> str:
     body_json = {"method": "auth.self", "params": {}}
     body_b64 = _b64_encode(body_json)
 
-    try:
-        async with httpx.AsyncClient(base_url=base_url, headers=headers, timeout=30.0) as client:
-            response = await client.post("/api/warden/mint", content=body_b64)
-            response.raise_for_status()
+    # Retry loop for JWT minting (handles orchestrator cold start)
+    JWT_MINT_RETRIES = 3
+    last_error = None
 
-            # Decode base64 response
-            result = _b64_decode(response.text)
+    for attempt in range(JWT_MINT_RETRIES):
+        try:
+            async with httpx.AsyncClient(base_url=base_url, headers=headers, timeout=30.0) as client:
+                response = await client.post("/api/warden/mint", content=body_b64)
+
+                # Check for retryable status codes before raise_for_status
+                if response.status_code == 401 or response.status_code >= 500:
+                    if attempt < JWT_MINT_RETRIES - 1:
+                        delay = exponential_backoff_with_jitter(attempt, base_delay=2.0, max_delay=10.0)
+                        print(
+                            f"[Dominus] JWT mint returned {response.status_code}, "
+                            f"retrying in {delay:.1f}s (attempt {attempt + 1}/{JWT_MINT_RETRIES})"
+                        )
+                        await asyncio.sleep(delay)
+                        continue
+
+                response.raise_for_status()
+
+                # Decode base64 response
+                result = _b64_decode(response.text)
 
-            if not result.get("success"):
-                error_msg = result.get("error", "Unknown auth error")
-                sovereign_circuit_breaker.record_failure()
-                raise RuntimeError(f"Auth error: {error_msg}")
+                if not result.get("success"):
+                    error_msg = result.get("error", "Unknown auth error")
+                    sovereign_circuit_breaker.record_failure()
+                    raise RuntimeError(f"Auth error: {error_msg}")
+
+                data = result.get("data", {})
+                jwt = data.get("access_token") or data.get("token")
+                if not jwt:
+                    sovereign_circuit_breaker.record_failure()
+                    raise RuntimeError("No JWT token in auth response")
+
+                # Success - record it
+                sovereign_circuit_breaker.record_success()
+                return jwt
+
+        except httpx.TimeoutException as e:
+            last_error = e
+            if attempt < JWT_MINT_RETRIES - 1:
+                delay = exponential_backoff_with_jitter(attempt, base_delay=2.0, max_delay=10.0)
+                print(
+                    f"[Dominus] JWT mint timed out, "
+                    f"retrying in {delay:.1f}s (attempt {attempt + 1}/{JWT_MINT_RETRIES})"
+                )
+                await asyncio.sleep(delay)
+                continue
+            sovereign_circuit_breaker.record_failure()
+            raise RuntimeError(f"Failed to get JWT: {e}") from e
+
+        except httpx.NetworkError as e:
+            last_error = e
+            if attempt < JWT_MINT_RETRIES - 1:
+                delay = exponential_backoff_with_jitter(attempt, base_delay=2.0, max_delay=10.0)
+                print(
+                    f"[Dominus] JWT mint network error ({e}), "
+                    f"retrying in {delay:.1f}s (attempt {attempt + 1}/{JWT_MINT_RETRIES})"
+                )
+                await asyncio.sleep(delay)
+                continue
+            sovereign_circuit_breaker.record_failure()
+            raise RuntimeError(f"Failed to get JWT: {e}") from e
 
-            data = result.get("data", {})
-            jwt = data.get("access_token") or data.get("token")
-            if not jwt:
+        except httpx.HTTPStatusError as e:
+            last_error = e
+            # 4xx errors (except 401 which is retried above) should not be retried
+            if 400 <= e.response.status_code < 500 and e.response.status_code != 401:
                 sovereign_circuit_breaker.record_failure()
-                raise RuntimeError("No JWT token in auth response")
-
-            # Success - record it
-            sovereign_circuit_breaker.record_success()
-            return jwt
-
-    except (httpx.TimeoutException, httpx.NetworkError, httpx.HTTPStatusError) as e:
-        sovereign_circuit_breaker.record_failure()
-        raise RuntimeError(f"Failed to get JWT: {e}") from e
+                raise RuntimeError(f"Failed to get JWT: {e}") from e
+            # For other errors, we've already handled retries in the status check above
+            sovereign_circuit_breaker.record_failure()
+            raise RuntimeError(f"Failed to get JWT: {e}") from e
+
+    # Should not reach here, but just in case
+    sovereign_circuit_breaker.record_failure()
+    raise RuntimeError(f"Failed to get JWT after {JWT_MINT_RETRIES} retries: {last_error}")
 
 
 def _get_architect_url(psk_token: str = None, sovereign_url: str = None, environment: str = None) -> str:

{dominus_sdk_python-2.1.2 → dominus_sdk_python-2.1.3}/dominus_sdk_python.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dominus-sdk-python
-Version: 2.1.2
+Version: 2.1.3
 Summary: Python SDK for the Dominus Orchestrator Platform
 Author-email: CareBridge Systems <dev@carebridge.io>
 License: Proprietary

{dominus_sdk_python-2.1.2 → dominus_sdk_python-2.1.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "dominus-sdk-python"
-version = "2.1.2"
+version = "2.1.3"
 description = "Python SDK for the Dominus Orchestrator Platform"
 readme = "README.md"
 license = {text = "Proprietary"}