From c6974fd10851a6c8dd9aba1df5fd700d673c3861 Mon Sep 17 00:00:00 2001 From: Teknium Date: Thu, 9 Apr 2026 13:23:43 -0700 Subject: [PATCH] fix: allow custom endpoint users to use main model for auxiliary tasks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Step 1 of _resolve_auto() explicitly excluded 'custom' providers, forcing custom endpoint users through the fragile fallback chain instead of using their known-working main model credentials. This caused silent compression failures for users on local OpenAI- compatible endpoints — the summary generation would fail, middle turns would be silently dropped, and the agent would lose all conversation context. Remove 'custom' from the exclusion list so custom endpoint users get the same main-model-first treatment as DeepSeek, Anthropic, Gemini, and other direct providers. --- agent/auxiliary_client.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/agent/auxiliary_client.py b/agent/auxiliary_client.py index 2f3a64a6..a757f426 100644 --- a/agent/auxiliary_client.py +++ b/agent/auxiliary_client.py @@ -1137,7 +1137,7 @@ def _resolve_auto() -> Tuple[Optional[OpenAI], Optional[str]]: main_model = _read_main_model() if (main_provider and main_model and main_provider not in _AGGREGATOR_PROVIDERS - and main_provider not in ("auto", "custom", "")): + and main_provider not in ("auto", "")): client, resolved = resolve_provider_client(main_provider, main_model) if client is not None: logger.info("Auxiliary auto-detect: using main provider %s (%s)",