Increase max_steps 7 -> 12 to reduce premature cutoffs

25% run showed questions hitting max_steps while still making progress, forcing garbled non-tool-call final answers. TokenPacer sleeps stayed short (6-8s), so there's TPM headroom to let runs go longer.

app.py

@@ -137,7 +137,12 @@ class BasicAgent:
                     "collections", "statistics", "datetime", "io", "openpyxl", "PIL",
                 ]),
             ],
-            max_steps=7,  # keep runs short: growing history otherwise blows the TPM budget by step ~6
+            # 12 (was 7): the 25% run showed several questions still making
+            # real progress (finding the right source) when cut off at max
+            # steps, forcing a garbled non-tool-call final answer instead of
+            # a real one. TokenPacer sleeps stayed short (6-8s) at 7 steps,
+            # so there's TPM budget headroom to let runs go longer.
+            max_steps=12,
             step_callbacks=[
                 MemoryTrimmer(),
                 TokenPacer(tokens_per_minute_budget=int(os.getenv("RATE_LIMIT_TOKENS_PER_MINUTE", "25000"))),