github/awesome-kubernetes
1
0
Fork 0
mirror of https://github.com/nubenetes/awesome-kubernetes.git synced 2026-08-19 04:16:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(ai-models): mandate Pro models for grounding requests to avoid empty responses

Browse Source
This commit is contained in:
Inaki Fernandez
2026-07-12 16:02:46 +02:00
parent c9d0a65593
commit 2d5400ab48
1 changed files with 9 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gemini_utils.py
+9 -7
View File
@@ -292,7 +292,9 @@ class GeminiQuotaExhausted(Exception):
def handle_retry_error(retry_state):
import sys
log_event(" [🚨] CIRCUIT BREAKER TRIPPED: Tenacity exhausted retries. Emitting exit code 42.")
exc = retry_state.outcome.exception()
log_event(f" [🚨] CIRCUIT BREAKER TRIPPED: Tenacity exhausted retries. Exception: {exc}")
log_event(" Emitting exit code 42.")
sys.exit(42)
@retry(
@@ -391,16 +393,16 @@ async def call_gemini_with_retry(prompt: str, response_format: str = "json", max
base_wait_time = 2.0
# 1. Smart Filtering and Re-ordering
if prefer_flash:
# Strict filter: Only allow flash/lite models
models = [m for m in models_pool if "flash" in m or "lite" in m]
if not models:
models = ["gemini-1.5-flash", "gemini-1.5-flash-latest"]
elif use_grounding:
if use_grounding:
# For grounding, we MANDATE Pro models as they have superior search/reasoning capabilities
models = [m for m in models_pool if "pro" in m]
if not models:
models = ["gemini-1.5-pro", "gemini-1.5-pro-latest"]
elif prefer_flash:
# Strict filter: Only allow flash/lite models
models = [m for m in models_pool if "flash" in m or "lite" in m]
if not models:
models = ["gemini-1.5-flash", "gemini-1.5-flash-latest"]
else:
models = models_pool