ling/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-04-28 06:51:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #2182 from NousResearch/hermes/hermes-5d6932ba

Browse Source 
fix: 6 bugs in model metadata, reasoning detection, and delegate tool
This commit is contained in:
Teknium
2026-03-20 08:53:01 -07:00
committed by GitHub
parent a51a767407 55ce601502
commit 5e705bc31b
4 changed files with 50 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
agent/model_metadata.py
View File

@@ -34,17 +34,29 @@ _PROVIDER_PREFIXES: frozenset[str] = frozenset({
})
_OLLAMA_TAG_PATTERN = re.compile(
r"^(\d+\.?\d*b|latest|stable|q\d|fp?\d|instruct|chat|coder|vision|text)",
re.IGNORECASE,
)
def _strip_provider_prefix(model: str) -> str:
"""Strip a recognised provider prefix from a model string.
``"local:my-model"`` → ``"my-model"``
``"qwen3.5:27b"`` → ``"qwen3.5:27b"`` (unchanged — not a provider prefix)
``"qwen:0.5b"`` → ``"qwen:0.5b"`` (unchanged — Ollama model:tag)
``"deepseek:latest"``→ ``"deepseek:latest"``(unchanged — Ollama model:tag)
"""
if ":" not in model or model.startswith("http"):
return model
prefix = model.split(":", 1)[0].strip().lower()
if prefix in _PROVIDER_PREFIXES:
return model.split(":", 1)[1]
prefix, suffix = model.split(":", 1)
prefix_lower = prefix.strip().lower()
if prefix_lower in _PROVIDER_PREFIXES:
# Don't strip if suffix looks like an Ollama tag (e.g. "7b", "latest", "q4_0")
if _OLLAMA_TAG_PATTERN.match(suffix.strip()):
return model
return suffix
return model
_model_metadata_cache: Dict[str, Dict[str, Any]] = {}
@@ -800,7 +812,7 @@ def get_model_context_length(
ctx = _resolve_nous_context_length(model)
if ctx:
return ctx
elif provider:
if provider:
from agent.models_dev import lookup_models_dev_context
ctx = lookup_models_dev_context(provider, model)
if ctx:
@@ -812,10 +824,13 @@ def get_model_context_length(
return metadata[model].get("context_length", 128000)
# 8. Hardcoded defaults (fuzzy match — longest key first for specificity)
# Only check `default_model in model` (is the key a substring of the input).
# The reverse (`model in default_model`) causes shorter names like
# "claude-sonnet-4" to incorrectly match "claude-sonnet-4-6" and return 1M.
for default_model, length in sorted(
DEFAULT_CONTEXT_LENGTHS.items(), key=lambda x: len(x[0]), reverse=True
):
if default_model in model or model in default_model:
if default_model in model:
return length
# 9. Query local server as last resort

5
agent/models_dev.py
View File

@@ -107,11 +107,12 @@ def fetch_models_dev(force_refresh: bool = False) -> Dict[str, Any]:
except Exception as e:
logger.debug("Failed to fetch models.dev: %s", e)
# Fall back to disk cache
# Fall back to disk cache — use a short TTL (5 min) so we retry
# the network fetch soon instead of serving stale data for a full hour.
if not _models_dev_cache:
_models_dev_cache = _load_disk_cache()
if _models_dev_cache:
_models_dev_cache_time = time.time()
_models_dev_cache_time = time.time() - _MODELS_DEV_CACHE_TTL + 300
logger.debug("Loaded models.dev from disk cache (%d providers)", len(_models_dev_cache))
return _models_dev_cache

17
run_agent.py
View File

@@ -1142,23 +1142,24 @@ class AIAgent:
def _has_content_after_think_block(self, content: str) -> bool:
"""
Check if content has actual text after any <think></think> blocks.
Check if content has actual text after any reasoning/thinking blocks.
This detects cases where the model only outputs reasoning but no actual
response, which indicates an incomplete generation that should be retried.
Must stay in sync with _strip_think_blocks() tag variants.
Args:
content: The assistant message content to check
Returns:
True if there's meaningful content after think blocks, False otherwise
"""
if not content:
return False
# Remove all <think>...</think> blocks (including nested ones, non-greedy)
cleaned = re.sub(r'<think>.*?</think>', '', content, flags=re.DOTALL)
# Remove all reasoning tag variants (must match _strip_think_blocks)
cleaned = self._strip_think_blocks(content)
# Check if there's any non-whitespace content remaining
return bool(cleaned.strip())

33
tools/delegate_tool.py
View File

@@ -470,22 +470,25 @@ def delegate_task(
_parent_tool_names = list(_model_tools._last_resolved_tool_names)
# Build all child agents on the main thread (thread-safe construction)
# Wrapped in try/finally so the global is always restored even if a
# child build raises (otherwise _last_resolved_tool_names stays corrupted).
children = []
for i, t in enumerate(task_list):
child = _build_child_agent(
task_index=i, goal=t["goal"], context=t.get("context"),
toolsets=t.get("toolsets") or toolsets, model=creds["model"],
max_iterations=effective_max_iter, parent_agent=parent_agent,
override_provider=creds["provider"], override_base_url=creds["base_url"],
override_api_key=creds["api_key"],
override_api_mode=creds["api_mode"],
)
# Override with correct parent tool names (before child construction mutated global)
child._delegate_saved_tool_names = _parent_tool_names
children.append((i, t, child))
# Authoritative restore: reset global to parent's tool names after all children built
_model_tools._last_resolved_tool_names = _parent_tool_names
try:
for i, t in enumerate(task_list):
child = _build_child_agent(
task_index=i, goal=t["goal"], context=t.get("context"),
toolsets=t.get("toolsets") or toolsets, model=creds["model"],
max_iterations=effective_max_iter, parent_agent=parent_agent,
override_provider=creds["provider"], override_base_url=creds["base_url"],
override_api_key=creds["api_key"],
override_api_mode=creds["api_mode"],
)
# Override with correct parent tool names (before child construction mutated global)
child._delegate_saved_tool_names = _parent_tool_names
children.append((i, t, child))
finally:
# Authoritative restore: reset global to parent's tool names after all children built
_model_tools._last_resolved_tool_names = _parent_tool_names
if n_tasks == 1:
# Single task -- run directly (no thread pool overhead)