Merge pull request 'Switch to shttp transport' (#16) from streamable-http into main

Reviewed-on: #16

config.yaml

@@ -3,7 +3,7 @@ google_cloud_location: us-central1
 
 firestore_db: bnt-orquestador-cognitivo-firestore-bdo-dev
 
-mcp_remote_url: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app/sse"
+mcp_remote_url: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app/mcp"
 # audience sin la ruta, para emitir el ID Token:
 mcp_audience: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app"
 

src/va_agent/agent.py

@@ -4,7 +4,7 @@ from google import genai
 from google.adk.agents.llm_agent import Agent
 from google.adk.runners import Runner
 from google.adk.tools.mcp_tool import McpToolset
-from google.adk.tools.mcp_tool.mcp_session_manager import SseConnectionParams
+from google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPConnectionParams
 from google.cloud.firestore_v1.async_client import AsyncClient
 
 from va_agent.auth import auth_headers_provider
@@ -12,7 +12,7 @@ from va_agent.config import settings
 from va_agent.session import FirestoreSessionService
 
 toolset = McpToolset(
-    connection_params=SseConnectionParams(url=settings.mcp_remote_url),
+    connection_params=StreamableHTTPConnectionParams(url=settings.mcp_remote_url),
     header_provider=auth_headers_provider,
 )
 

src/va_agent/auth.py

@@ -1,6 +1,6 @@
 """ID-token auth for Cloud Run → Cloud Run calls."""
 
-import threading
+import logging
 import time
 
 from google.adk.agents.readonly_context import ReadonlyContext
@@ -10,18 +10,33 @@ from google.oauth2 import id_token
 
 from va_agent.config import settings
 
-_REFRESH_MARGIN = 300  # refresh 5 min before expiry
+logger = logging.getLogger(__name__)
+
+_REFRESH_MARGIN = 900  # refresh 15 min before expiry
 
-_lock = threading.Lock()
 _token: str | None = None
 _token_exp: float = 0.0
 
 
+def _fetch_token() -> tuple[str, float]:
+    """Fetch a fresh ID token (blocking I/O)."""
+    tok = id_token.fetch_id_token(GAuthRequest(), settings.mcp_audience)
+    exp = jwt.decode(tok, verify=False)["exp"]
+    return tok, exp
+
+
 def auth_headers_provider(_ctx: ReadonlyContext | None = None) -> dict[str, str]:
-    """Return Authorization headers with a cached ID token."""
-    global _token, _token_exp  # noqa: PLW0603
-    with _lock:
-        if _token is None or time.time() >= _token_exp - _REFRESH_MARGIN:
-            _token = id_token.fetch_id_token(GAuthRequest(), settings.mcp_audience)
-            _token_exp = jwt.decode(_token, verify=False)["exp"]
+    """Return Authorization headers, refreshing the cached token when needed.
+
+    With Streamable HTTP transport every tool call is a fresh HTTP
+    request, so returning a valid token here is sufficient — no
+    background refresh loop required.
+    """
+    global _token, _token_exp
+
+    if _token is not None and time.time() < _token_exp - _REFRESH_MARGIN:
         return {"Authorization": f"Bearer {_token}"}
+
+    tok, exp = _fetch_token()
+    _token, _token_exp = tok, exp
+    return {"Authorization": f"Bearer {tok}"}

tests/test_auth.py

@@ -75,8 +75,7 @@ class TestAuthHeadersProvider:
     ) -> None:
         mock_settings.mcp_audience = "https://my-service"
 
-        # First token expires within the refresh margin (5 min).
-        first_exp = time.time() + 100  # < 300s margin
+        first_exp = time.time() + 100  # < 900s margin
         second_exp = time.time() + 3600
         mock_fetch.side_effect = [
             _make_fake_token(first_exp),