cat-llm 0.0.67__py3-none-any.whl → 0.0.69__py3-none-any.whl

catllm/calls/all_calls.py

@@ -0,0 +1,433 @@
+# openai stepback prompt
+
+def get_stepback_insight_openai(
+    stepback,
+    api_key,
+    user_model,
+    model_source="openai",
+    creativity=None
+):
+    from openai import OpenAI
+    # Conditional base_url setting based on model source
+    base_url = (
+        "https://api.perplexity.ai" if model_source == "perplexity" 
+        else "https://router.huggingface.co/v1" if model_source == "huggingface"
+        else None
+    )
+    
+    client = OpenAI(api_key=api_key, base_url=base_url)
+    
+    try:
+        stepback_response = client.chat.completions.create(
+            model=user_model,
+            messages=[{'role': 'user', 'content': stepback}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        stepback_insight = stepback_response.choices[0].message.content
+        
+        return stepback_insight, True
+        
+    except Exception as e:
+        print(f"An error occurred during step-back prompting: {e}")
+        return None, False
+
+
+# claude stepback prompt
+
+def get_stepback_insight_anthropic(
+    stepback,
+    api_key,
+    user_model,
+    model_source="anthropic",
+    creativity=None
+):
+    import anthropic
+
+    client = anthropic.Anthropic(api_key=api_key)
+    
+    try:
+        stepback_response = client.messages.create(
+            model=user_model,
+            max_tokens=4096,
+            messages=[{'role': 'user', 'content': stepback}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        stepback_insight = stepback_response.content[0].text
+        
+        return stepback_insight, True
+        
+    except Exception as e:
+        print(f"An error occurred during step-back prompting: {e}")
+        return None, False
+    
+# google stepback prompt
+
+def get_stepback_insight_google(
+    stepback,
+    api_key,
+    user_model,
+    model_source="google",
+    creativity=None
+):
+    
+    import requests
+    
+    url = f"https://generativelanguage.googleapis.com/v1beta/models/{user_model}:generateContent?key={api_key}"
+    
+    headers = {
+        "Content-Type": "application/json"
+    }
+    
+    payload = {
+        "contents": [{
+            "parts": [{"text": stepback}],
+
+            **({"generationConfig": {"temperature": creativity}} if creativity is not None else {})
+        }]
+    }
+    
+    try:
+        response = requests.post(url, headers=headers, json=payload)
+        response.raise_for_status()  # Raise error for bad status codes
+        
+        result = response.json()
+        stepback_insight = result['candidates'][0]['content']['parts'][0]['text']
+        
+        return stepback_insight, True
+        
+    except Exception as e:
+        print(f"An error occurred during step-back prompting: {e}")
+        return None, False
+
+# mistral stepback prompt
+
+def get_stepback_insight_mistral(
+    stepback,
+    api_key,
+    user_model,
+    model_source="mistral",
+    creativity=None
+):
+    
+    from mistralai import Mistral
+    
+    client = Mistral(api_key=api_key)
+    
+    try:
+        stepback_response = client.chat.complete(
+            model=user_model,
+            messages=[{'role': 'user', 'content': stepback}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        stepback_insight = stepback_response.choices[0].message.content
+        
+        return stepback_insight, True
+        
+    except Exception as e:
+        print(f"An error occurred during step-back prompting: {e}")
+        return None, False
+    
+# openai chain of verification calls
+
+def chain_of_verification_openai(
+    initial_reply,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    client,
+    user_model,
+    creativity,
+    remove_numbering
+):
+    """
+    Execute Chain of Verification (CoVe) process.
+    Returns the verified reply or initial reply if error occurs.
+    """
+    try:
+        # STEP 2: Generate verification questions
+        step2_filled = step2_prompt.replace('<<INITIAL_REPLY>>', initial_reply)
+        
+        verification_response = client.chat.completions.create(
+            model=user_model,
+            messages=[{'role': 'user', 'content': step2_filled}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        
+        verification_questions = verification_response.choices[0].message.content
+        
+        # STEP 3: Answer verification questions
+        questions_list = [
+            remove_numbering(q) 
+            for q in verification_questions.split('\n') 
+            if q.strip()
+        ]
+        verification_qa = []
+        
+        # Prompting each question individually
+        for question in questions_list:
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+            
+            answer_response = client.chat.completions.create(
+                model=user_model,
+                messages=[{'role': 'user', 'content': step3_filled}],
+                **({"temperature": creativity} if creativity is not None else {})
+            )
+            
+            answer = answer_response.choices[0].message.content
+            verification_qa.append(f"Q: {question}\nA: {answer}")
+        
+        # STEP 4: Final corrected categorization
+        verification_qa_text = "\n\n".join(verification_qa)
+        
+        step4_filled = (step4_prompt
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+        
+        print(f"Final prompt:\n{step4_filled}\n")
+        
+        final_response = client.chat.completions.create(
+            model=user_model,
+            messages=[{'role': 'user', 'content': step4_filled}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        
+        verified_reply = final_response.choices[0].message.content
+        print("Chain of verification completed. Final response generated.\n")
+        
+        return verified_reply
+        
+    except Exception as e:
+        print(f"ERROR in Chain of Verification: {str(e)}")
+        print("Falling back to initial response.\n")
+        return initial_reply
+    
+# anthropic chain of verification calls
+
+def chain_of_verification_anthropic(
+    initial_reply,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    client,
+    user_model,
+    creativity,
+    remove_numbering
+):
+    """
+    Execute Chain of Verification (CoVe) process for Anthropic Claude.
+    Returns the verified reply or initial reply if error occurs.
+    """
+    try:
+        # STEP 2: Generate verification questions
+        step2_filled = step2_prompt.replace('<<INITIAL_REPLY>>', initial_reply)
+        
+        verification_response = client.messages.create(
+            model=user_model,
+            messages=[{'role': 'user', 'content': step2_filled}],
+            max_tokens=4096,
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        
+        verification_questions = verification_response.content[0].text
+        
+        # STEP 3: Answer verification questions
+        questions_list = [
+            remove_numbering(q) 
+            for q in verification_questions.split('\n') 
+            if q.strip()
+        ]
+        print(f"Verification questions:\n{questions_list}\n")
+        verification_qa = []
+        
+        # Prompting each question individually
+        for question in questions_list:
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+            
+            answer_response = client.messages.create(
+                model=user_model,
+                messages=[{'role': 'user', 'content': step3_filled}],
+                max_tokens=4096,
+                **({"temperature": creativity} if creativity is not None else {})
+            )
+            
+            answer = answer_response.content[0].text
+            verification_qa.append(f"Q: {question}\nA: {answer}")
+        
+        # STEP 4: Final corrected categorization
+        verification_qa_text = "\n\n".join(verification_qa)
+        
+        step4_filled = (step4_prompt
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+        
+        print(f"Final prompt:\n{step4_filled}\n")
+        
+        final_response = client.messages.create(
+            model=user_model,
+            messages=[{'role': 'user', 'content': step4_filled}],
+            max_tokens=4096,
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        
+        verified_reply = final_response.content[0].text
+        print("Chain of verification completed. Final response generated.\n")
+        
+        return verified_reply
+        
+    except Exception as e:
+        print(f"ERROR in Chain of Verification: {str(e)}")
+        print("Falling back to initial response.\n")
+        return initial_reply
+
+# google chain of verification calls
+def chain_of_verification_google(
+    initial_reply,
+    prompt,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    url,
+    headers,
+    creativity,
+    remove_numbering,
+    make_google_request
+):
+    import time
+    """
+    Execute Chain of Verification (CoVe) process for Google Gemini.
+    Returns the verified reply or initial reply if error occurs.
+    """
+    try:
+        # STEP 2: Generate verification questions
+        step2_filled = step2_prompt.replace('<<INITIAL_REPLY>>', initial_reply)
+        
+        payload_step2 = {
+            "contents": [{
+                "parts": [{"text": step2_filled}]
+            }],
+            **({"generationConfig": {"temperature": creativity}} if creativity is not None else {})
+        }
+        
+        result_step2 = make_google_request(url, headers, payload_step2)
+        verification_questions = result_step2["candidates"][0]["content"]["parts"][0]["text"]
+        
+        # STEP 3: Answer verification questions
+        questions_list = [
+            remove_numbering(q) 
+            for q in verification_questions.split('\n') 
+            if q.strip()
+        ]
+        verification_qa = []
+        
+        for question in questions_list:
+            time.sleep(2)  # temporary rate limit handling
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+            
+            payload_step3 = {
+                "contents": [{
+                    "parts": [{"text": step3_filled}]
+                }],
+                **({"generationConfig": {"temperature": creativity}} if creativity is not None else {})
+            }
+            
+            result_step3 = make_google_request(url, headers, payload_step3)
+            answer = result_step3["candidates"][0]["content"]["parts"][0]["text"]
+            verification_qa.append(f"Q: {question}\nA: {answer}")
+        
+        # STEP 4: Final corrected categorization
+        verification_qa_text = "\n\n".join(verification_qa)
+        
+        step4_filled = (step4_prompt
+            .replace('<<PROMPT>>', prompt)
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+        
+        payload_step4 = {
+            "contents": [{
+                "parts": [{"text": step4_filled}]
+            }],
+            **({"generationConfig": {"temperature": creativity}} if creativity is not None else {})
+        }
+        
+        result_step4 = make_google_request(url, headers, payload_step4)
+        verified_reply = result_step4["candidates"][0]["content"]["parts"][0]["text"]
+        
+        print("Chain of verification completed. Final response generated.\n")
+        return verified_reply
+        
+    except Exception as e:
+        print(f"ERROR in Chain of Verification: {str(e)}")
+        print("Falling back to initial response.\n")
+        return initial_reply
+    
+# mistral chain of verification calls
+
+def chain_of_verification_mistral(
+    initial_reply,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    client,
+    user_model,
+    creativity,
+    remove_numbering
+):
+    """
+    Execute Chain of Verification (CoVe) process for Mistral AI.
+    Returns the verified reply or initial reply if error occurs.
+    """
+    try:
+        # STEP 2: Generate verification questions
+        step2_filled = step2_prompt.replace('<<INITIAL_REPLY>>', initial_reply)
+        
+        verification_response = client.chat.complete(
+            model=user_model,
+            messages=[{'role': 'user', 'content': step2_filled}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        
+        verification_questions = verification_response.choices[0].message.content
+        
+        # STEP 3: Answer verification questions
+        questions_list = [
+            remove_numbering(q) 
+            for q in verification_questions.split('\n') 
+            if q.strip()
+        ]
+        verification_qa = []
+        
+        # Prompting each question individually
+        for question in questions_list:
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+            
+            answer_response = client.chat.complete(
+                model=user_model,
+                messages=[{'role': 'user', 'content': step3_filled}],
+                **({"temperature": creativity} if creativity is not None else {})
+            )
+            
+            answer = answer_response.choices[0].message.content
+            verification_qa.append(f"Q: {question}\nA: {answer}")
+        
+        # STEP 4: Final corrected categorization
+        verification_qa_text = "\n\n".join(verification_qa)
+        
+        step4_filled = (step4_prompt
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+        
+        final_response = client.chat.complete(
+            model=user_model,
+            messages=[{'role': 'user', 'content': step4_filled}],
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+        
+        verified_reply = final_response.choices[0].message.content
+        print("Chain of verification completed. Final response generated.\n")
+        
+        return verified_reply
+        
+    except Exception as e:
+        print(f"ERROR in Chain of Verification: {str(e)}")
+        print("Falling back to initial response.\n")
+        return initial_reply

catllm/model_reference_list.py

@@ -0,0 +1,94 @@
+# openai list of models
+openai_models = [
+    "gpt-5",
+    "gpt-5-mini",
+    "gpt-5-nano",
+    "gpt-4o",
+    "gpt-4o-mini",
+    "gpt-4.1",
+    "gpt-4.1-mini",
+    "gpt-4.1-nano",
+    "gpt-3.5-turbo",
+    "text-davinci-003",
+    "text-davinci-002"
+]
+
+#  anthropic list of models
+anthropic_models = [
+    "claude-opus-4-20250514-v1:0",
+    "claude-opus-4-1-20250805-v1:0",
+    "claude-sonnet-4-5-20250929-v1:0",
+    "claude-sonnet-4-20250514-v1:0",
+    "claude-3-7-sonnet-20250219-v1:0",
+    "claude-3-5-sonnet-20240620-v1:0",
+    "claude-3-5-haiku-20241022-v1:0",
+    "claude-3-opus-20240229-v1:0",
+    "claude-3-sonnet-20240229-v1:0",
+    "claude-haiku-4-5-20251001-v1:0",
+    "claude-sonnet-4-5-20250929",
+    "claude-haiku-4-5-20251001",
+    "claude-opus-4-1-20250805"
+]
+
+# google list of models
+
+google_models = [
+    "gemini-2.5-flash",
+    "gemini-2.5-flash-lite",
+    "gemini-2.5-pro",
+    "gemini-2.0-flash",
+    "gemini-2.0-flash-lite",
+    "gemini-2.0-pro",
+    "gemini-2.5",
+    "gemini-2.0",
+    "gemini-2.5-flash-preview",
+    "gemini-2.5-pro-preview"
+]
+
+# perplexity list of models
+
+perplexity_models = [
+    "sonar",
+    "sonar-pro",
+    "sonar-reasoning",
+    "sonar-reasoning-pro",
+    "sonar-deep-research",
+    "r1-1776"
+]
+
+# mistral list of models
+
+mistral_models = [
+    "mistral-large-latest",
+    "mistral-medium-2505",
+    "mistral-large-2411",
+    "codestral-2501",
+    "pixtral-large-2411",
+    "mistral-small-2407",
+    "mistral-embed",
+    "codestral-embed",
+    "mistral-moderation-2411",
+    "ministral-3b-2410",
+    "ministral-8b-2410"
+]
+
+# meta list of models
+# list can be found here: https://huggingface.co/collections/meta-llama/llama-4-67f0c30d9fe03840bc9d0164
+meta_llama_models = [
+    "meta/llama-3.1-8b-instruct",
+    "meta/llama-3.1-70b-instruct",
+    "meta/llama-3.1-405b-instruct",
+    "meta/llama-3.2-11b-vision-instruct",
+    "meta/llama-3.2-90b-vision-instruct",
+    "meta/llama-3.3-70b-instruct",
+    "meta/llama-4-scout-17b-16e-instruct",
+    "meta/llama-4-maverick-17b-128e-instruct",
+    "llama-4-maverick-17b-128e-instruct-maas",
+    "llama-4-scout-17b-16e-instruct-maas",
+    "llama-3.3-70b-instruct-maas",
+    "llama-3.2-90b-vision-instruct-maas",
+    "llama-3.1-405b-instruct-maas",
+    "llama-3.1-70b-instruct-maas",
+    "llama-3.1-8b-instruct-maas",
+
+]