fix(copilot): update rate limit comment with realistic token estimates, remove synthetic IDs from billing metadata

- config.py: Update per-turn token estimate from 10-15K to 25-35K avg
  (accounting for context growth), adjust daily turn estimate to ~70-100
- helpers.py: Remove synthetic graph/node IDs from UsageTransactionMetadata
  since they're fake data; keep only real identifiers (node_exec_id, block_id)

autogpt_platform/backend/backend/copilot/config.py

@@ -71,8 +71,10 @@ class ChatConfig(BaseSettings):
     )
 
     # Rate limiting — token-based limits per day and per week.
-    # Each CoPilot turn consumes ~10-15K tokens (system prompt + tool schemas + response),
-    # so 2.5M daily allows ~170-250 turns/day which is reasonable for normal use.
+    # Per-turn token cost varies with context size: ~10-15K for early turns,
+    # ~30-50K mid-session, up to ~100K pre-compaction. Average across a
+    # session with compaction cycles is ~25-35K tokens/turn, so 2.5M daily
+    # allows ~70-100 turns/day.
     # Checked both at the HTTP layer (routes.py pre-turn) and in the SDK service
     # just before compaction, since context can grow fast with heavy workloads.
     #

autogpt_platform/backend/backend/copilot/tools/helpers.py

@@ -147,9 +147,6 @@ async def execute_block(
                     user_id=user_id,
                     cost=cost,
                     metadata=UsageTransactionMetadata(
-                        graph_exec_id=synthetic_graph_id,
-                        graph_id=synthetic_graph_id,
-                        node_id=synthetic_node_id,
                         node_exec_id=node_exec_id,
                         block_id=block_id,
                         block=block.name,