fix(compression): truncated summaries no longer become compaction checkpoints (port of earendil-works/pi#7048)

A summarization response with finish_reason == "length" contains PARTIAL
text — the generation stopped on the output-token cap mid-summary.
Previously all compressor summarization sites accepted such responses as
complete: the cut-off text replaced the real middle turns AND was fed back
into every subsequent iterative-update prompt, compounding the loss across
compactions.

Guards added at all four summarization sites (whole bug class):
- _generate_summary: length stop raises, gets the existing one-shot
  main-model fallback (a larger output budget may finish the summary), and
  on terminal failure ABORTS compression preserving the session unchanged
  (new _last_summary_truncated_failure flag, same class as empty-content).
- _micro_summarize_one: partial rolling-summary merge is discarded; the
  exchange stays unabsorbed for a later pass.
- _build_chunk_digests: partial lean digest degrades to the
  recover-via-session_search placeholder.
- trajectory_compressor (sync + async): length stop raises into the
  existing retry/backoff loop.

_response_finish_reason() reads dict- and object-shaped responses and
returns "" when the provider omits the field, so proxies that never send
finish_reason are unaffected.

Ported from earendil-works/pi commit 97fa14e39 (pi#7048), adapted to
hermes' abort-preserving compression failure machinery.

Tests: tests/agent/test_compressor_truncated_summary_guard.py (12 tests;
sabotage-verified — disabling the guards fails 4).

trajectory_compressor.py

@@ -56,6 +56,31 @@ _project_env = Path(__file__).parent / ".env"
 load_hermes_dotenv(hermes_home=_hermes_home, project_env=_project_env)
 
 
+def _response_finish_reason(response: Any) -> str:
+    """Return ``choices[0].finish_reason`` from a dict- or object-shaped response.
+
+    Local copy of ``agent.context_compressor._response_finish_reason`` —
+    trajectory_compressor is a standalone CLI tool and deliberately avoids
+    importing the (heavy) context compressor module. Returns the lowercased
+    finish reason, or ``""`` when absent/unreadable.
+    """
+    try:
+        if isinstance(response, dict):
+            choices = response.get("choices") or [{}]
+            first = choices[0] if choices else {}
+            reason = (
+                first.get("finish_reason")
+                if isinstance(first, dict)
+                else getattr(first, "finish_reason", None)
+            )
+        else:
+            choices = getattr(response, "choices", None) or []
+            reason = getattr(choices[0], "finish_reason", None) if choices else None
+        return str(reason).strip().lower() if reason else ""
+    except Exception:
+        return ""
+
+
 def _effective_temperature_for_model(
     model: str,
     requested_temperature: float,
@@ -658,6 +683,16 @@ Write only the summary, starting with "[CONTEXT SUMMARY]:" prefix."""
                         _create_kwargs["temperature"] = summary_temperature
                     response = self.client.chat.completions.create(**_create_kwargs)
                 
+                _fr = _response_finish_reason(response)
+                if _fr == "length":
+                    # Length stop = partial summary; storing it as the turn
+                    # replacement silently truncates the trajectory's memory.
+                    # Raise so the retry/backoff loop treats it as a failure
+                    # (pi#7048 class).
+                    raise RuntimeError(
+                        "trajectory summarization hit the output token cap "
+                        "(finish_reason=length); summary is incomplete"
+                    )
                 summary = self._coerce_summary_content(response.choices[0].message.content)
                 return self._ensure_summary_prefix(summary)
                 
@@ -727,6 +762,13 @@ Write only the summary, starting with "[CONTEXT SUMMARY]:" prefix."""
                         _create_kwargs["temperature"] = summary_temperature
                     response = await self._get_async_client().chat.completions.create(**_create_kwargs)
                 
+                if _response_finish_reason(response) == "length":
+                    # Length stop = partial summary; see sync sibling above
+                    # (pi#7048 class).
+                    raise RuntimeError(
+                        "trajectory summarization hit the output token cap "
+                        "(finish_reason=length); summary is incomplete"
+                    )
                 summary = self._coerce_summary_content(response.choices[0].message.content)
                 return self._ensure_summary_prefix(summary)