fix(honcho): surface honcho_reasoning backend failures instead of 'No result'

dialectic_query collapsed every backend failure to an empty string, so
the explicit honcho_reasoning tool rendered timeouts, server errors,
and genuinely-empty answers identically as 'No result from Honcho.'
(#36098 issue 4). Operators debugging 'search works but reasoning does
not' were sent down representation/observation rabbit holes when the
real cause was a 30s timeout on a medium-reasoning dialectic call.

Add raise_errors to dialectic_query (default false — automatic
injection keeps its fail-quiet behavior and cadence backoff) and pass
it from the explicit tool call, returning a tool error that names the
failure and points at the timeout knob. Auth errors keep their
dedicated handler.
This commit is contained in:
Erosika
2026-08-10 19:12:52 -04:00
committed by kshitij
parent 606481586a
commit c28c434706
2 changed files with 33 additions and 7 deletions
+25 -7
View File
@@ -1527,13 +1527,31 @@ class HonchoMemoryProvider(MemoryProvider):
return tool_error("Missing required parameter: query")
peer = args.get("peer", "user")
reasoning_level = args.get("reasoning_level")
result = self._manager.dialectic_query(
self._session_key, query,
reasoning_level=reasoning_level,
peer=peer,
# Explicit reasoning bypasses the automatic-injection cap.
apply_injection_cap=False,
)
try:
result = self._manager.dialectic_query(
self._session_key, query,
reasoning_level=reasoning_level,
peer=peer,
# Explicit reasoning bypasses the automatic-injection cap.
apply_injection_cap=False,
# Explicit tool call: surface timeouts/server errors as
# errors instead of collapsing them into "no result",
# which is indistinguishable from an empty answer.
raise_errors=True,
)
except HonchoAuthError:
# Let the outer dispatch's auth-specific handler render this.
raise
except Exception as e:
logger.warning("honcho_reasoning failed: %s", e)
return tool_error(
f"Honcho reasoning query failed ({e}). This is a backend "
"error, not an empty result — the peer may still have "
"relevant context. Slow dialectic calls at higher "
"reasoning levels can exceed the configured timeout; "
"consider a lower reasoning_level or raising the "
"'timeout' value in honcho.json."
)
# Update cadence tracker so auto-injection respects the gap after an explicit call
self._last_dialectic_turn = self._turn_count
return json.dumps({"result": result or "No result from Honcho."})
+8
View File
@@ -844,6 +844,7 @@ class HonchoSessionManager:
reasoning_level: str | None = None,
peer: str = "user",
apply_injection_cap: bool = True,
raise_errors: bool = False,
) -> str:
"""
Query Honcho's dialectic endpoint about a peer.
@@ -862,6 +863,11 @@ class HonchoSessionManager:
apply_injection_cap: Clip automatic injections to
``dialecticMaxChars``. Explicit ``honcho_reasoning`` calls pass
False because Honcho already bounds their output.
raise_errors: Re-raise backend failures instead of returning "".
Explicit tool calls pass True so a timeout or server error
surfaces as an error, not as "no result" (#36098 issue 4:
collapsing failures to "" made auth errors, timeouts, and
genuinely-empty answers indistinguishable).
Returns:
Honcho's synthesized answer, or empty string on failure.
@@ -917,6 +923,8 @@ class HonchoSessionManager:
raise
except Exception as e:
logger.warning("Honcho dialectic query failed: %s", e)
if raise_errors:
raise
return ""
def prefetch_context(self, session_key: str, user_message: str | None = None) -> None: