fix: context pressure warning fires at 85% of context_length, not threshold (#538)

_emit_context_pressure computed compaction_progress as
real_tokens / threshold_tokens. With threshold_percent=0.50,
this fires at 85% of 50% = 42.5% of total context.

Fix: compute progress relative to context_length so the
warning fires at 85% of actual context capacity.
Same fix applied to the post-compression reset check.

Closes #538

run_agent.py

@@ -6006,7 +6006,7 @@ class AIAgent:
         # alone exceeds the warning level), keep the flag set to prevent
         # spamming the user with repeated warnings every loop iteration.
         if self.context_compressor.threshold_tokens > 0:
-            _post_progress = _compressed_est / self.context_compressor.threshold_tokens
+            _post_progress = _compressed_est / self.context_compressor.context_length if self.context_compressor.context_length else 0
             if _post_progress < 0.85:
                 self._context_pressure_warned = False
 
@@ -8996,7 +8996,7 @@ class AIAgent:
                     # Does not inject into messages — just prints to CLI output
                     # and fires status_callback for gateway platforms.
                     if _compressor.threshold_tokens > 0:
-                        _compaction_progress = _real_tokens / _compressor.threshold_tokens
+                        _compaction_progress = _real_tokens / _compressor.context_length if _compressor.context_length else 0
                         if _compaction_progress >= 0.85 and not self._context_pressure_warned:
                             self._context_pressure_warned = True
                             self._emit_context_pressure(_compaction_progress, _compressor)