fix(backend/copilot): route copilot cost logging through DatabaseManagerAsyncClient

The copilot executor's token_tracking.py was using schedule_cost_log()
which calls execute_raw_with_schema() directly on the Prisma singleton.
In the copilot_executor process, Prisma is not reliably connected due to
event-loop binding issues, causing ClientNotConnectedError on every turn.

Fix: route cost logging through platform_cost_db() -> DatabaseManagerAsyncClient
RPC (same approach already used by the block executor). Also fix
_copilot_block_name() to extract only the service tag from the log prefix
(e.g. "[SDK][session-id][T1]" -> "copilot:SDK") instead of the full suffix.

Update cost_tracking.py drain to drain token_tracking._pending_log_tasks,
and update token_tracking_test.py mocks to match new call site.

autogpt_platform/backend/backend/copilot/token_tracking.py

@@ -9,19 +9,41 @@ Both the baseline (OpenRouter) and SDK (Anthropic) service layers need to:
 This module extracts that common logic so both paths stay in sync.
 """
 
+import asyncio
 import logging
+import re
 
-from backend.data.platform_cost import (
-    PlatformCostEntry,
-    schedule_cost_log,
-    usd_to_microdollars,
-)
+from backend.data.db_accessors import platform_cost_db
+from backend.data.platform_cost import PlatformCostEntry, usd_to_microdollars
 
 from .model import ChatSession, Usage
 from .rate_limit import record_token_usage
 
 logger = logging.getLogger(__name__)
 
+# Hold strong references to in-flight cost log tasks to prevent GC.
+_pending_log_tasks: set["asyncio.Task[None]"] = set()
+
+
+def _schedule_cost_log(entry: PlatformCostEntry) -> None:
+    """Schedule a fire-and-forget cost log via DatabaseManagerAsyncClient RPC."""
+
+    async def _safe_log() -> None:
+        try:
+            await platform_cost_db().log_platform_cost(entry)
+        except Exception:
+            logger.exception(
+                "Failed to log platform cost for user=%s provider=%s block=%s",
+                entry.user_id,
+                entry.provider,
+                entry.block_name,
+            )
+
+    task = asyncio.create_task(_safe_log())
+    _pending_log_tasks.add(task)
+    task.add_done_callback(_pending_log_tasks.discard)
+
+
 # Identifiers used by PlatformCostLog for copilot turns (not tied to a real
 # block/credential in the block_cost_config or credentials_store tables).
 COPILOT_BLOCK_ID = "copilot"
@@ -29,8 +51,10 @@ COPILOT_CREDENTIAL_ID = "copilot_system"
 
 
 def _copilot_block_name(log_prefix: str) -> str:
-    """Turn a log prefix like ``"[SDK]"`` into a stable block_name
-    ``"copilot:SDK"``. Empty prefix becomes just ``"copilot"``."""
+    """Extract stable block_name from ``"[SDK][session][T1]"`` -> ``"copilot:SDK"``."""
+    match = re.search(r"\[([A-Za-z][A-Za-z0-9_]*)\]", log_prefix)
+    if match:
+        return f"{COPILOT_BLOCK_ID}:{match.group(1)}"
     tag = log_prefix.strip(" []")
     return f"{COPILOT_BLOCK_ID}:{tag}" if tag else COPILOT_BLOCK_ID
 
@@ -148,7 +172,7 @@ async def persist_and_record_usage(
             tracking_type = "tokens"
             tracking_amount = total_tokens
 
-        schedule_cost_log(
+        _schedule_cost_log(
             PlatformCostEntry(
                 user_id=user_id,
                 graph_exec_id=session_id,

autogpt_platform/backend/backend/copilot/token_tracking_test.py

@@ -298,8 +298,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(
@@ -337,8 +339,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(
@@ -368,8 +372,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(
@@ -391,8 +397,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(
@@ -418,8 +426,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(
@@ -445,8 +455,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(
@@ -470,8 +482,10 @@ class TestPlatformCostLogging:
                 new_callable=AsyncMock,
             ),
             patch(
-                "backend.data.platform_cost.log_platform_cost_safe",
-                new=mock_log,
+                "backend.copilot.token_tracking.platform_cost_db",
+                return_value=type(
+                    "FakePlatformCostDb", (), {"log_platform_cost": mock_log}
+                )(),
             ),
         ):
             await persist_and_record_usage(

autogpt_platform/backend/backend/data/db_accessors.py

@@ -142,3 +142,9 @@ def credit_db():
         credit_db = get_database_manager_async_client()
 
     return credit_db
+
+
+def platform_cost_db():
+    from backend.util.clients import get_database_manager_async_client
+
+    return get_database_manager_async_client()

autogpt_platform/backend/backend/executor/cost_tracking.py

@@ -58,12 +58,21 @@ async def drain_pending_cost_logs(timeout: float = 5.0) -> None:
                 len(still_pending),
                 timeout,
             )
-    # Also drain copilot cost log tasks (platform_cost._pending_log_tasks)
-    from backend.data.platform_cost import (  # noqa: PLC0415
-        drain_pending_cost_logs as _drain_copilot,
+    # Also drain copilot cost log tasks (token_tracking._pending_log_tasks)
+    from backend.copilot.token_tracking import (  # noqa: PLC0415
+        _pending_log_tasks as _copilot_tasks,
     )
 
-    await _drain_copilot()
+    copilot_pending = list(_copilot_tasks)
+    if copilot_pending:
+        logger.info("Draining %d copilot cost log task(s)", len(copilot_pending))
+        _, still_pending = await asyncio.wait(copilot_pending, timeout=timeout)
+        if still_pending:
+            logger.warning(
+                "%d copilot cost log task(s) did not complete within %.1fs",
+                len(still_pending),
+                timeout,
+            )
 
 
 def _schedule_log(