refactor(agent/context_compressor): use the failure-kind dataclass directly; compact threshold/arg helpers

This commit is contained in:
Teknium
2026-09-02 20:23:47 -07:00
parent 932f7532ca
commit 71bfa94ff9
+17 -39
View File
@@ -1048,7 +1048,6 @@ def _extract_tool_call_name_and_args(tool_call: Any) -> tuple[str, str]:
if isinstance(tool_call, dict):
fn = tool_call.get("function") or {}
return str(fn.get("name") or "unknown"), str(fn.get("arguments") or "")
fn = getattr(tool_call, "function", None)
if fn is None:
return "unknown", ""
@@ -1634,17 +1633,13 @@ def _summarize_tool_result_unguarded(tool_name: str, tool_args: str, tool_conten
args = {}
if not isinstance(args, dict):
args = {}
content = tool_content or ""
content_len = len(content)
line_count = content.count("\n") + 1 if content.strip() else 0
summarizer = _TOOL_RESULT_SUMMARIZERS.get(tool_name)
if summarizer is not None:
return summarizer(tool_name, args, content, content_len, line_count)
first_arg = ""
for k, v in list(args.items())[:2]:
first_arg += f" {k}={str(v)[:40]}"
first_arg = "".join(f" {k}={str(v)[:40]}" for k, v in list(args.items())[:2])
return f"[{tool_name}]{first_arg} ({content_len:,} chars result)"
@@ -1655,13 +1650,8 @@ def resolve_model_threshold(model: str, model_thresholds: dict[str, float] | Non
plugin context engines can reuse it."""
if not model_thresholds or not model:
return default
best_key = ""
for key in model_thresholds:
if key in model and len(key) > len(best_key):
best_key = key
if best_key:
return float(model_thresholds[best_key])
return default
best_key = max((key for key in model_thresholds if key in model), key=len, default="")
return float(model_thresholds[best_key]) if best_key else default
def _memory_provider_section(memory_context: str) -> str:
@@ -3588,32 +3578,23 @@ This compaction should PRIORITISE preserving all information related to the focu
_SUMMARY_FAILURE_COOLDOWN_SECONDS, "no auxiliary LLM provider configured",
)
self._last_summary_error = "no auxiliary LLM provider configured"
logger.warning("Context compression: no provider available for "
"summary. Middle turns will be dropped without summary "
"for %d seconds.",
_SUMMARY_FAILURE_COOLDOWN_SECONDS)
logger.warning(
"Context compression: no provider available for summary. Middle turns will be dropped without "
"summary for %d seconds.",
_SUMMARY_FAILURE_COOLDOWN_SECONDS,
)
return None
kind = _classify_summary_failure(e)
_is_model_not_found = kind.model_not_found
_is_timeout = kind.timeout
_is_json_decode = kind.json_decode
_is_streaming_closed = kind.streaming_closed
_is_empty_content = kind.empty_content
_is_truncated_summary = kind.truncated
# Auth/permission/quota failures are not retryable: flag so compress() preserves the
# session. A distinct summary_model still gets the one-shot main-model fallback.
if _is_summary_access_or_quota_error(e):
# Field name kept for caller compatibility; now covers the whole access/quota class.
self._last_summary_auth_failure = True
if _is_json_decode and not _is_model_not_found and not _is_timeout:
if kind.json_decode and not kind.model_not_found and not kind.timeout:
logger.error(
"Context compression failed: auxiliary LLM returned a non-JSON response. provider=%s "
"summary_model=%s main_model=%s base_url=%s err=%s",
self.provider or "auto",
self.summary_model or "(main)",
self.model,
self.base_url or "default",
e,
self.provider or "auto", self.summary_model or "(main)", self.model, self.base_url or "default", e,
)
# A distinct summary model gets ONE main-model retry: a specific reason for known transient classes,
# else a best-effort "failed" retry — losing N turns is worse than one extra summary attempt.
@@ -3623,17 +3604,14 @@ This compaction should PRIORITISE preserving all information related to the focu
and not getattr(self, "_summary_model_fallen_back", False)
):
self._fallback_to_main_for_compression(e, kind.fallback_reason())
return self._generate_summary(
turns_to_summarize,
focus_topic=focus_topic,
memory_context=memory_context,
) # retry immediately
# Retry immediately on the main model.
return self._generate_summary(turns_to_summarize, focus_topic=focus_topic, memory_context=memory_context)
# Transient errors: short cooldown for JSON-decode/streaming-closed. Timeouts escalate
# 60s→300s→900s (structural repeat offenders) and take precedence over the short rung.
if _is_timeout:
if kind.timeout:
_transient_cooldown = _next_timeout_cooldown(self)
elif _is_json_decode or _is_streaming_closed or _is_empty_content or _is_truncated_summary:
elif kind.json_decode or kind.streaming_closed or kind.empty_content or kind.truncated:
_transient_cooldown = 30
else:
_transient_cooldown = 60
@@ -3642,11 +3620,11 @@ This compaction should PRIORITISE preserving all information related to the focu
self._last_summary_error = err_text
# Terminal network/empty-content failure after any fallback: flag so compress() ABORTS
# and preserves the session; independent of abort_on_summary_failure.
if _is_streaming_closed:
if kind.streaming_closed:
self._last_summary_network_failure = True
elif _is_truncated_summary:
elif kind.truncated:
self._last_summary_truncated_failure = True
elif _is_empty_content:
elif kind.empty_content:
self._last_summary_empty_content_failure = True
logger.warning(
"Failed to generate context summary: %s. Further summary attempts paused for %d seconds.", e,