slipstream/discordbot
slipstream/discordbot
1
1
Fork 0
Code Issues Pull Requests 3 Actions Packages Projects Releases Wiki Activity

feat: Use specific model for emoji/sticker descriptions

Browse Source 
Introduces `EMOJI_STICKER_DESCRIPTION_MODEL` to `config.py` and uses it in `generate_image_description` for "emoji" and "sticker" item types. This ensures a dedicated model is used for these specific image description tasks, improving accuracy or efficiency.
This commit is contained in:
Slipstream 2025-05-29 11:41:36 -06:00
parent 4ec866adb6
commit fd13c35afd
Signed by: slipstream
GPG Key ID: 13E498CE010AC6FD
2 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
gurt/api.py
View File

@ -151,7 +151,7 @@ from google.api_core import exceptions as google_exceptions # Keep for retry log
# Relative imports for components within the 'gurt' package # Relative imports for components within the 'gurt' package
from .config import ( from .config import (
PROJECT_ID, LOCATION, DEFAULT_MODEL, FALLBACK_MODEL, CUSTOM_TUNED_MODEL_ENDPOINT, # Import the new endpoint PROJECT_ID, LOCATION, DEFAULT_MODEL, FALLBACK_MODEL, CUSTOM_TUNED_MODEL_ENDPOINT, EMOJI_STICKER_DESCRIPTION_MODEL, # Import the new endpoint and model
API_TIMEOUT, API_RETRY_ATTEMPTS, API_RETRY_DELAY, TOOLS, RESPONSE_SCHEMA, API_TIMEOUT, API_RETRY_ATTEMPTS, API_RETRY_DELAY, TOOLS, RESPONSE_SCHEMA,
PROACTIVE_PLAN_SCHEMA, # Import the new schema PROACTIVE_PLAN_SCHEMA, # Import the new schema
TAVILY_API_KEY, PISTON_API_URL, PISTON_API_KEY, BASELINE_PERSONALITY, TENOR_API_KEY # Import other needed configs TAVILY_API_KEY, PISTON_API_URL, PISTON_API_KEY, BASELINE_PERSONALITY, TENOR_API_KEY # Import other needed configs
@ -1923,9 +1923,8 @@ async def generate_image_description(
# 4. Call AI # 4. Call AI
# Use a multimodal model, e.g., DEFAULT_MODEL if it's Gemini 1.5 Pro or similar # Use a multimodal model, e.g., DEFAULT_MODEL if it's Gemini 1.5 Pro or similar
# If DEFAULT_MODEL is tuned for JSON, we might need to specify a base multimodal model here. # Determine which model to use based on item_type
# For now, assume DEFAULT_MODEL can handle this. model_to_use = EMOJI_STICKER_DESCRIPTION_MODEL if item_type in ["emoji", "sticker"] else DEFAULT_MODEL
model_to_use = DEFAULT_MODEL # Or specify a known multimodal model like "models/gemini-1.5-pro-preview-0409"
print(f"Calling AI for image description ({item_name}) using model: {model_to_use}") print(f"Calling AI for image description ({item_name}) using model: {model_to_use}")
ai_response_obj = await call_google_genai_api_with_retry( ai_response_obj = await call_google_genai_api_with_retry(

1
gurt/config.py
View File

@ -26,6 +26,7 @@ DEFAULT_MODEL = os.getenv("GURT_DEFAULT_MODEL", "gemini-2.5-flash-preview-05-20"
FALLBACK_MODEL = os.getenv("GURT_FALLBACK_MODEL", "gemini-2.5-flash-preview-05-20") FALLBACK_MODEL = os.getenv("GURT_FALLBACK_MODEL", "gemini-2.5-flash-preview-05-20")
CUSTOM_TUNED_MODEL_ENDPOINT = os.getenv("GURT_CUSTOM_TUNED_MODEL", "gemini-2.5-flash-preview-05-20") CUSTOM_TUNED_MODEL_ENDPOINT = os.getenv("GURT_CUSTOM_TUNED_MODEL", "gemini-2.5-flash-preview-05-20")
SAFETY_CHECK_MODEL = os.getenv("GURT_SAFETY_CHECK_MODEL", "gemini-2.5-flash-preview-05-20") # Use a Vertex AI model for safety checks SAFETY_CHECK_MODEL = os.getenv("GURT_SAFETY_CHECK_MODEL", "gemini-2.5-flash-preview-05-20") # Use a Vertex AI model for safety checks
EMOJI_STICKER_DESCRIPTION_MODEL = "gemini-2.0-flash-001" # Hardcoded for emoji/sticker image descriptions
# --- Database Paths --- # --- Database Paths ---
DB_PATH = os.getenv("GURT_DB_PATH", "data/gurt_memory.db") DB_PATH = os.getenv("GURT_DB_PATH", "data/gurt_memory.db")