mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-04-25 00:51:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(gateway): read custom_providers context_length in hygiene compression (#4085)

Browse source 
Gateway hygiene pre-compression only checked model.context_length from
the top-level config, missing per-model context_length defined in
custom_providers entries. This caused premature compression for custom
provider users (e.g. 128K default instead of 200K configured).

The AIAgent's own compressor already reads custom_providers correctly
(run_agent.py lines 1171-1189). This adds the same fallback to the
gateway hygiene path, running after runtime provider resolution so
the base_url is available for matching.
This commit is contained in:
Teknium 2026-03-30 17:04:31 -07:00 • committed by GitHub
parent f007284d05
commit 4a7c17fca5
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 23 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

23
gateway/run.py
View file

@ -2284,6 +2284,29 @@ class GatewayRunner:
_hyg_api_key = _hyg_runtime.get("api_key")
except Exception:
pass
# Check custom_providers per-model context_length
# (same fallback as run_agent.py lines 1171-1189).
# Must run after runtime resolution so _hyg_base_url is set.
if _hyg_config_context_length is None and _hyg_base_url:
try:
_hyg_custom_providers = _hyg_data.get("custom_providers")
if isinstance(_hyg_custom_providers, list):
for _cp in _hyg_custom_providers:
if not isinstance(_cp, dict):
continue
_cp_url = (_cp.get("base_url") or "").rstrip("/")
if _cp_url and _cp_url == _hyg_base_url.rstrip("/"):
_cp_models = _cp.get("models", {})
if isinstance(_cp_models, dict):
_cp_model_cfg = _cp_models.get(_hyg_model, {})
if isinstance(_cp_model_cfg, dict):
_cp_ctx = _cp_model_cfg.get("context_length")
if _cp_ctx is not None:
_hyg_config_context_length = int(_cp_ctx)
break
except (TypeError, ValueError):
pass
except Exception:
pass