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(agent): prefer Ollama Modelfile num_ctx over GGUF training max

Browse source 
_query_local_context_length was checking model_info.context_length
(the GGUF training max) before num_ctx (the Modelfile runtime override),
inverse to query_ollama_num_ctx. The two helpers therefore disagreed on
the same model:

  hermes-brain:qwen3-14b-ctx32k     # Modelfile: num_ctx 32768
  underlying qwen3:14b GGUF         # qwen3.context_length: 40960

query_ollama_num_ctx correctly returned 32768 (the value Ollama will
actually allocate KV cache for). _query_local_context_length returned
40960, which let ContextCompressor grow conversations past 32768 before
triggering compression — at which point Ollama silently truncated the
prefix, corrupting context.

Swap the order so num_ctx is checked first, matching query_ollama_num_ctx.
Adds a parametrized test that seeds both values and asserts num_ctx wins.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
ismell0992-afk 2026-04-13 11:41:45 +02:00 • committed by Teknium
parent 39b83f3443
commit 3e99964789
2 changed files with 49 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

17
agent/model_metadata.py
View file

@ -775,12 +775,12 @@ def _query_local_context_length(model: str, base_url: str) -> Optional[int]:
resp = client.post(f"{server_url}/api/show", json={"name": model})
if resp.status_code == 200:
data = resp.json()
# Check model_info for context length
model_info = data.get("model_info", {})
for key, value in model_info.items():
if "context_length" in key and isinstance(value, (int, float)):
return int(value)
# Check parameters string for num_ctx
# Prefer explicit num_ctx from Modelfile parameters: this is
# the *runtime* context Ollama will actually allocate KV cache
# for. The GGUF model_info.context_length is the training max,
# which can be larger than num_ctx — using it here would let
# Hermes grow conversations past the runtime limit and Ollama
# would silently truncate. Matches query_ollama_num_ctx().
params = data.get("parameters", "")
if "num_ctx" in params:
for line in params.split("\n"):
@ -791,6 +791,11 @@ def _query_local_context_length(model: str, base_url: str) -> Optional[int]:
return int(parts[-1])
except ValueError:
pass
# Fall back to GGUF model_info context_length (training max)
model_info = data.get("model_info", {})
for key, value in model_info.items():
if "context_length" in key and isinstance(value, (int, float)):
return int(value)
# LM Studio native API: /api/v1/models returns max_context_length.
# This is more reliable than the OpenAI-compat /v1/models which