mirror of
https://github.com/JamesTheGiblet/BuddAI.git
synced 2026-01-08 21:58:40 +00:00
48716ff271
- Implemented tests for confidence scoring logic in `test_buddai_confidence.py` and `test_confidence.py`, covering high and low confidence scenarios, escalation thresholds, and validation scoring penalties. - Created tests for fallback logging functionality in `test_fallback_logging.py`, ensuring fallback prompts are logged correctly and the `/logs` command retrieves log content. - Developed tests for fallback prompts in `test_fallback_prompts.py`, verifying that specific prompts are used for different models based on confidence levels. - Generated detailed test reports for multiple test runs, confirming all tests passed successfully.
94 lines
No EOL
3.6 KiB
Python
94 lines
No EOL
3.6 KiB
Python
import os
|
|
import logging
|
|
|
|
# Optional import for Google Generative AI
|
|
try:
|
|
import google.generativeai as genai
|
|
HAS_GEMINI = True
|
|
except ImportError:
|
|
HAS_GEMINI = False
|
|
|
|
# Optional import for OpenAI
|
|
try:
|
|
from openai import OpenAI
|
|
HAS_OPENAI = True
|
|
except ImportError:
|
|
HAS_OPENAI = False
|
|
|
|
class FallbackClient:
|
|
"""
|
|
Handles escalation to external AI models (Gemini, OpenAI) when local confidence is low.
|
|
"""
|
|
def __init__(self):
|
|
self.gemini_key = os.getenv("GEMINI_API_KEY")
|
|
self.openai_key = os.getenv("OPENAI_API_KEY")
|
|
|
|
self.gemini_client = None
|
|
self.openai_client = None
|
|
|
|
# Initialize Gemini
|
|
if self.gemini_key and HAS_GEMINI:
|
|
try:
|
|
genai.configure(api_key=self.gemini_key)
|
|
self.gemini_client = genai.GenerativeModel('gemini-1.5-flash')
|
|
except Exception as e:
|
|
print(f"⚠️ Failed to initialize Gemini client: {e}")
|
|
|
|
# Initialize OpenAI
|
|
if self.openai_key and HAS_OPENAI:
|
|
try:
|
|
self.openai_client = OpenAI(api_key=self.openai_key)
|
|
except Exception as e:
|
|
print(f"⚠️ Failed to initialize OpenAI client: {e}")
|
|
|
|
def escalate(self, model_alias: str, original_prompt: str, buddai_attempt: str, confidence: int) -> str:
|
|
"""
|
|
Routes the escalation request to the appropriate provider.
|
|
"""
|
|
if model_alias == 'gemini':
|
|
return self._call_gemini(original_prompt, buddai_attempt, confidence)
|
|
elif model_alias in ['gpt4', 'chatgpt']:
|
|
return self._call_openai(model_alias, original_prompt, buddai_attempt, confidence)
|
|
|
|
return f"⚠️ Fallback model '{model_alias}' not supported for active escalation."
|
|
|
|
def _call_gemini(self, original_prompt: str, buddai_attempt: str, confidence: int) -> str:
|
|
if not self.gemini_client:
|
|
return f"⚠️ Gemini fallback unavailable (Key missing or init failed)."
|
|
|
|
try:
|
|
prompt = self._build_prompt(original_prompt, buddai_attempt, confidence)
|
|
response = self.gemini_client.generate_content(prompt)
|
|
return f"✨ **Gemini Fallback (Confidence: {confidence}%)**\n\n{response.text}"
|
|
except Exception as e:
|
|
return f"❌ Error calling Gemini API: {str(e)}"
|
|
|
|
def _call_openai(self, model_alias: str, original_prompt: str, buddai_attempt: str, confidence: int) -> str:
|
|
if not self.openai_client:
|
|
return f"⚠️ OpenAI fallback unavailable (Key missing or init failed)."
|
|
|
|
model_map = {
|
|
'gpt4': 'gpt-4',
|
|
'chatgpt': 'gpt-3.5-turbo'
|
|
}
|
|
target_model = model_map.get(model_alias, 'gpt-3.5-turbo')
|
|
|
|
try:
|
|
prompt = self._build_prompt(original_prompt, buddai_attempt, confidence)
|
|
response = self.openai_client.chat.completions.create(
|
|
model=target_model,
|
|
messages=[
|
|
{"role": "system", "content": "You are a high-precision coding assistant acting as a fallback validator."},
|
|
{"role": "user", "content": prompt}
|
|
]
|
|
)
|
|
return f"✨ **{model_alias.upper()} Fallback (Confidence: {confidence}%)**\n\n{response.choices[0].message.content}"
|
|
except Exception as e:
|
|
return f"❌ Error calling OpenAI API: {str(e)}"
|
|
|
|
def _build_prompt(self, original, attempt, confidence):
|
|
return f"""
|
|
[USER REQUEST]: {original}
|
|
[LOCAL ATTEMPT ({confidence}% confidence)]: {attempt}
|
|
[TASK]: Fix issues, apply best practices, and return corrected code.
|
|
""" |