[gemini] feat: Add Claude API fallback tier to cascade.py (#980) (#1119)

Co-authored-by: Google Gemini <gemini@hermes.local>
Co-committed-by: Google Gemini <gemini@hermes.local>
This commit is contained in:
2026-03-23 18:21:18 +00:00
committed by rockachopa
parent 1be1324a0d
commit e8b3d59041
4 changed files with 51 additions and 2 deletions

View File

@@ -114,6 +114,7 @@ class Provider:
type: str # ollama, openai, anthropic
enabled: bool
priority: int
tier: str | None = None # e.g., "local", "standard_cloud", "frontier"
url: str | None = None
api_key: str | None = None
base_url: str | None = None
@@ -267,6 +268,7 @@ class CascadeRouter:
type=p_data["type"],
enabled=p_data.get("enabled", True),
priority=p_data.get("priority", 99),
tier=p_data.get("tier"),
url=p_data.get("url"),
api_key=p_data.get("api_key"),
base_url=p_data.get("base_url"),
@@ -532,6 +534,7 @@ class CascadeRouter:
model: str | None = None,
temperature: float = 0.7,
max_tokens: int | None = None,
cascade_tier: str | None = None,
) -> dict:
"""Complete a chat conversation with automatic failover.
@@ -545,6 +548,8 @@ class CascadeRouter:
model: Preferred model (tries this first, then provider defaults)
temperature: Sampling temperature
max_tokens: Maximum tokens to generate
cascade_tier: If specified, filters providers by this tier.
- "frontier_required": Uses only Anthropic provider for top-tier models.
Returns:
Dict with content, provider_used, and metrics
@@ -558,7 +563,18 @@ class CascadeRouter:
errors = []
for provider in self.providers:
providers = self.providers
if cascade_tier == "frontier_required":
providers = [p for p in self.providers if p.type == "anthropic"]
if not providers:
raise RuntimeError("No Anthropic provider configured for 'frontier_required' tier.")
elif cascade_tier:
providers = [p for p in self.providers if p.tier == cascade_tier]
if not providers:
raise RuntimeError(f"No providers found for tier: {cascade_tier}")
for provider in providers:
if not self._is_provider_available(provider):
continue