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

fix: update Gemini model catalog + wire models.dev as live model source

Browse Source 
Follow-up for salvaged PR #5494:
- Update model catalog to Gemini 3.x + Gemma 4 (drop deprecated 2.0)
- Add list_agentic_models() to models_dev.py with noise filter
- Wire models.dev into _model_flow_api_key_provider as primary source
  (static curated list serves as offline fallback)
- Add gemini -> google mapping in PROVIDER_TO_MODELS_DEV
- Fix Gemma 4 context lengths to 256K (models.dev values)
- Update auxiliary model to gemini-3-flash-preview
- Expand tests: 3.x catalog, context lengths, models.dev integration
This commit is contained in:
Teknium
2026-04-06 10:19:19 -07:00
committed by Teknium
parent 6dfab35501
commit cc7136b1ac
7 changed files with 147 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
agent/auxiliary_client.py
View File

@@ -55,7 +55,7 @@ logger = logging.getLogger(__name__)
# Default auxiliary models for direct API-key providers (cheap/fast for side tasks)
_API_KEY_PROVIDER_AUX_MODELS: Dict[str, str] = {
"gemini": "gemini-2.5-flash",
"gemini": "gemini-3-flash-preview",
"zai": "glm-4.5-flash",
"kimi-coding": "kimi-k2-turbo-preview",
"minimax": "MiniMax-M2.7-highspeed",

6
agent/model_metadata.py
View File

@@ -103,10 +103,8 @@ DEFAULT_CONTEXT_LENGTHS = {
# Google
"gemini": 1048576,
# Gemma (open models served via AI Studio)
"gemma-4-31b": 262144,
"gemma-4-26b": 262144,
"gemma-4-e4b": 131072,
"gemma-4-e2b": 131072,
"gemma-4-31b": 256000,
"gemma-4-26b": 256000,
"gemma-3": 131072,
"gemma": 8192, # fallback for older gemma models
# DeepSeek

34
agent/models_dev.py
View File

@@ -160,6 +160,7 @@ PROVIDER_TO_MODELS_DEV: Dict[str, str] = {
"kilocode": "kilo",
"fireworks": "fireworks-ai",
"huggingface": "huggingface",
"gemini": "google",
"google": "google",
"xai": "xai",
"nvidia": "nvidia",
@@ -422,6 +423,39 @@ def list_provider_models(provider: str) -> List[str]:
return list(models.keys())
# Patterns that indicate non-agentic or noise models (TTS, embedding,
# dated preview snapshots, live/streaming-only, image-only).
import re
_NOISE_PATTERNS: re.Pattern = re.compile(
r"-tts\b|embedding|live-|-(preview|exp)-\d{2,4}[-_]|"
r"-image\b|-image-preview\b|-customtools\b",
re.IGNORECASE,
)
def list_agentic_models(provider: str) -> List[str]:
"""Return model IDs suitable for agentic use from models.dev.
Filters for tool_call=True and excludes noise (TTS, embedding,
dated preview snapshots, live/streaming, image-only models).
Returns an empty list on any failure.
"""
models = _get_provider_models(provider)
if models is None:
return []
result = []
for mid, entry in models.items():
if not isinstance(entry, dict):
continue
if not entry.get("tool_call", False):
continue
if _NOISE_PATTERNS.search(mid):
continue
result.append(mid)
return result
def search_models_dev(
query: str, provider: str = None, limit: int = 5
) -> List[Dict[str, Any]]:

39
hermes_cli/main.py
View File

@@ -2211,24 +2211,37 @@ def _model_flow_api_key_provider(config, provider_id, current_model=""):
save_env_value(base_url_env, override)
effective_base = override
# Model selection — try live /models endpoint first, fall back to defaults.
# Providers with large live catalogs (100+ models) use a curated list instead
# so users see familiar model names rather than an overwhelming dump.
# Model selection — resolution order:
# 1. models.dev registry (cached, filtered for agentic/tool-capable models)
# 2. Curated static fallback list (offline insurance)
# 3. Live /models endpoint probe (small providers without models.dev data)
curated = _PROVIDER_MODELS.get(provider_id, [])
if curated and len(curated) >= 8:
# Try models.dev first — returns tool-capable models, filtered for noise
mdev_models: list = []
try:
from agent.models_dev import list_agentic_models
mdev_models = list_agentic_models(provider_id)
except Exception:
pass
if mdev_models:
model_list = mdev_models
print(f" Found {len(model_list)} model(s) from models.dev registry")
elif curated and len(curated) >= 8:
# Curated list is substantial — use it directly, skip live probe
live_models = None
model_list = curated
print(f" Showing {len(model_list)} curated models — use \"Enter custom model name\" for others.")
else:
api_key_for_probe = existing_key or (get_env_value(key_env) if key_env else "")
live_models = fetch_api_models(api_key_for_probe, effective_base)
if live_models and len(live_models) >= len(curated):
model_list = live_models
print(f" Found {len(model_list)} model(s) from {pconfig.name} API")
else:
model_list = curated
if model_list:
print(f" Showing {len(model_list)} curated models — use \"Enter custom model name\" for others.")
if live_models and len(live_models) >= len(curated):
model_list = live_models
print(f" Found {len(model_list)} model(s) from {pconfig.name} API")
else:
model_list = curated
if model_list:
print(f" Showing {len(model_list)} curated models — use \"Enter custom model name\" for others.")
# else: no defaults either, will fall through to raw input
if provider_id in {"opencode-zen", "opencode-go"}:

10
hermes_cli/models.py
View File

@@ -112,15 +112,15 @@ _PROVIDER_MODELS: dict[str, list[str]] = {
"grok-code-fast-1",
],
"gemini": [
"gemini-3.1-pro-preview",
"gemini-3-flash-preview",
"gemini-3.1-flash-lite-preview",
"gemini-2.5-pro",
"gemini-2.5-flash",
"gemini-2.0-flash",
"gemini-2.0-flash-lite",
"gemini-2.5-flash-lite",
# Gemma open models (also served via AI Studio)
"gemma-4-31b-it",
"gemma-4-26b-a4b-it",
"gemma-4-e4b-it",
"gemma-4-e2b-it",
"gemma-4-26b-it",
],
"zai": [
"glm-5",

5
hermes_cli/setup.py
View File

@@ -112,8 +112,9 @@ _DEFAULT_PROVIDER_MODELS = {
"grok-code-fast-1",
],
"gemini": [
"gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.0-flash", "gemini-2.0-flash-lite",
"gemma-4-31b-it", "gemma-4-26b-a4b-it", "gemma-4-e4b-it", "gemma-4-e2b-it",
"gemini-3.1-pro-preview", "gemini-3-flash-preview", "gemini-3.1-flash-lite-preview",
"gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite",
"gemma-4-31b-it", "gemma-4-26b-it",
],
"zai": ["glm-5", "glm-4.7", "glm-4.5", "glm-4.5-flash"],
"kimi-coding": ["kimi-k2.5", "kimi-k2-thinking", "kimi-k2-turbo-preview"],

80
tests/test_gemini_provider.py
View File

@@ -8,6 +8,7 @@ from hermes_cli.auth import PROVIDER_REGISTRY, resolve_provider, resolve_api_key
from hermes_cli.models import _PROVIDER_MODELS, _PROVIDER_LABELS, _PROVIDER_ALIASES, normalize_provider
from hermes_cli.model_normalize import normalize_model_for_provider, detect_vendor
from agent.model_metadata import get_model_context_length
from agent.models_dev import PROVIDER_TO_MODELS_DEV, list_agentic_models, _NOISE_PATTERNS
# ── Provider Registry ──
@@ -131,6 +132,12 @@ class TestGeminiModelCatalog:
assert "gemini-2.5-flash" in models
assert "gemma-4-31b-it" in models
def test_provider_models_has_3x(self):
models = _PROVIDER_MODELS["gemini"]
assert "gemini-3.1-pro-preview" in models
assert "gemini-3-flash-preview" in models
assert "gemini-3.1-flash-lite-preview" in models
def test_provider_label(self):
assert "gemini" in _PROVIDER_LABELS
assert _PROVIDER_LABELS["gemini"] == "Google AI Studio"
@@ -165,11 +172,15 @@ class TestGeminiModelNormalization:
class TestGeminiContextLength:
def test_gemma_4_31b_context(self):
ctx = get_model_context_length("gemma-4-31b-it", provider="gemini")
assert ctx == 262144
assert ctx == 256000
def test_gemma_4_e4b_context(self):
ctx = get_model_context_length("gemma-4-e4b-it", provider="gemini")
assert ctx == 131072
def test_gemma_4_26b_context(self):
ctx = get_model_context_length("gemma-4-26b-it", provider="gemini")
assert ctx == 256000
def test_gemini_3_context(self):
ctx = get_model_context_length("gemini-3.1-pro-preview", provider="gemini")
assert ctx == 1048576
# ── Agent Init (no SyntaxError) ──
@@ -195,3 +206,64 @@ class TestGeminiAgentInit:
)
assert agent.api_mode == "chat_completions"
assert agent.provider == "gemini"
# ── models.dev Integration ──
class TestGeminiModelsDev:
def test_gemini_mapped_to_google(self):
assert PROVIDER_TO_MODELS_DEV.get("gemini") == "google"
def test_noise_filter_excludes_tts(self):
assert _NOISE_PATTERNS.search("gemini-2.5-pro-preview-tts")
def test_noise_filter_excludes_dated_preview(self):
assert _NOISE_PATTERNS.search("gemini-2.5-flash-preview-04-17")
def test_noise_filter_excludes_embedding(self):
assert _NOISE_PATTERNS.search("gemini-embedding-001")
def test_noise_filter_excludes_live(self):
assert _NOISE_PATTERNS.search("gemini-live-2.5-flash")
def test_noise_filter_excludes_image(self):
assert _NOISE_PATTERNS.search("gemini-2.5-flash-image")
def test_noise_filter_excludes_customtools(self):
assert _NOISE_PATTERNS.search("gemini-3.1-pro-preview-customtools")
def test_noise_filter_passes_stable(self):
assert not _NOISE_PATTERNS.search("gemini-2.5-flash")
def test_noise_filter_passes_preview(self):
# Non-dated preview (e.g. gemini-3-flash-preview) should pass
assert not _NOISE_PATTERNS.search("gemini-3-flash-preview")
def test_noise_filter_passes_gemma(self):
assert not _NOISE_PATTERNS.search("gemma-4-31b-it")
def test_list_agentic_models_with_mock_data(self):
"""list_agentic_models filters correctly from mock models.dev data."""
mock_data = {
"google": {
"models": {
"gemini-3-flash-preview": {"tool_call": True},
"gemini-2.5-pro": {"tool_call": True},
"gemini-embedding-001": {"tool_call": False},
"gemini-2.5-flash-preview-tts": {"tool_call": False},
"gemini-live-2.5-flash": {"tool_call": True},
"gemini-2.5-flash-preview-04-17": {"tool_call": True},
"gemma-4-31b-it": {"tool_call": True},
}
}
}
with patch("agent.models_dev.fetch_models_dev", return_value=mock_data):
result = list_agentic_models("gemini")
assert "gemini-3-flash-preview" in result
assert "gemini-2.5-pro" in result
assert "gemma-4-31b-it" in result
# Filtered out:
assert "gemini-embedding-001" not in result # no tool_call
assert "gemini-2.5-flash-preview-tts" not in result # no tool_call
assert "gemini-live-2.5-flash" not in result # noise: live-
assert "gemini-2.5-flash-preview-04-17" not in result # noise: dated preview