cat-stack 0.1.0__py3-none-any.whl

cat_stack/calls/pdf_CoVe.py

@@ -0,0 +1,386 @@
+# PDF-aware Chain of Verification (CoVe) functions for various LLM providers
+# These functions include the PDF page in verification steps for accurate document-based categorization
+
+
+def pdf_chain_of_verification_openai(
+    initial_reply,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    client,
+    user_model,
+    creativity,
+    remove_numbering,
+    pdf_content
+):
+    """
+    Execute Chain of Verification (CoVe) process for PDF pages with OpenAI.
+    The PDF page (as image) is included in verification steps for accurate assessment.
+    Returns the verified reply or initial reply if error occurs.
+
+    Args:
+        pdf_content: The PDF page content in OpenAI format (as image_url dict after conversion)
+    """
+    try:
+        # STEP 2: Generate verification questions (text only - questions about the categorization)
+        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 WITH the PDF page
+        questions_list = [
+            remove_numbering(q)
+            for q in verification_questions.split('\n')
+            if q.strip()
+        ]
+        verification_qa = []
+
+        for question in questions_list:
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+
+            # Include PDF page in the verification question
+            message_content = [
+                {"type": "text", "text": step3_filled},
+                pdf_content
+            ]
+
+            answer_response = client.chat.completions.create(
+                model=user_model,
+                messages=[{'role': 'user', 'content': message_content}],
+                **({"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 WITH the PDF page
+        verification_qa_text = "\n\n".join(verification_qa)
+
+        step4_filled = (step4_prompt
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+
+        # Include PDF page in final categorization
+        final_message_content = [
+            {"type": "text", "text": step4_filled},
+            pdf_content
+        ]
+
+        final_response = client.chat.completions.create(
+            model=user_model,
+            messages=[{'role': 'user', 'content': final_message_content}],
+            response_format={"type": "json_object"},
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+
+        verified_reply = final_response.choices[0].message.content
+
+        return verified_reply
+
+    except Exception as e:
+        return initial_reply
+
+
+def pdf_chain_of_verification_anthropic(
+    initial_reply,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    client,  # Deprecated, kept for backward compatibility
+    user_model,
+    creativity,
+    remove_numbering,
+    pdf_content,
+    api_key=None
+):
+    """
+    Execute Chain of Verification (CoVe) process for PDF pages with Anthropic Claude.
+    The PDF page is included in verification steps for accurate assessment.
+    Returns the verified reply or initial reply if error occurs.
+
+    Uses direct HTTP requests instead of Anthropic SDK.
+
+    Args:
+        pdf_content: The PDF page content in Anthropic format (dict with type: "document")
+        api_key: Anthropic API key for authentication
+    """
+    import requests
+
+    if api_key is None:
+        return initial_reply
+
+    endpoint = "https://api.anthropic.com/v1/messages"
+    headers = {
+        "Content-Type": "application/json",
+        "x-api-key": api_key,
+        "anthropic-version": "2023-06-01"
+    }
+
+    def make_anthropic_request(messages, max_tokens=4096):
+        """Helper to make Anthropic API requests."""
+        payload = {
+            "model": user_model,
+            "max_tokens": max_tokens,
+            "messages": messages,
+        }
+        if creativity is not None:
+            payload["temperature"] = creativity
+
+        response = requests.post(endpoint, headers=headers, json=payload, timeout=120)
+        response.raise_for_status()
+        result = response.json()
+
+        content = result.get("content", [])
+        if content and content[0].get("type") == "text":
+            return content[0].get("text", "")
+        return ""
+
+    try:
+        # STEP 2: Generate verification questions (text only)
+        step2_filled = step2_prompt.replace('<<INITIAL_REPLY>>', initial_reply)
+
+        verification_questions = make_anthropic_request(
+            [{'role': 'user', 'content': step2_filled}]
+        )
+
+        # STEP 3: Answer verification questions WITH the PDF page
+        questions_list = [
+            remove_numbering(q)
+            for q in verification_questions.split('\n')
+            if q.strip()
+        ]
+        verification_qa = []
+
+        for question in questions_list:
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+
+            # Include PDF page in the verification question
+            message_content = [
+                {"type": "text", "text": step3_filled},
+                pdf_content
+            ]
+
+            answer = make_anthropic_request(
+                [{'role': 'user', 'content': message_content}]
+            )
+            verification_qa.append(f"Q: {question}\nA: {answer}")
+
+        # STEP 4: Final corrected categorization WITH the PDF page
+        verification_qa_text = "\n\n".join(verification_qa)
+
+        step4_filled = (step4_prompt
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+
+        # Include PDF page in final categorization
+        final_message_content = [
+            {"type": "text", "text": step4_filled},
+            pdf_content
+        ]
+
+        verified_reply = make_anthropic_request(
+            [{'role': 'user', 'content': final_message_content}]
+        )
+
+        return verified_reply
+
+    except Exception as e:
+        return initial_reply
+
+
+def pdf_chain_of_verification_google(
+    initial_reply,
+    prompt,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    url,
+    headers,
+    creativity,
+    remove_numbering,
+    make_google_request,
+    pdf_data,
+    mime_type
+):
+    """
+    Execute Chain of Verification (CoVe) process for PDF pages with Google Gemini.
+    The PDF page is included in verification steps for accurate assessment.
+    Returns the verified reply or initial reply if error occurs.
+
+    Args:
+        pdf_data: Base64 encoded PDF page data
+        mime_type: MIME type of the content (e.g., "application/pdf")
+    """
+    import time
+
+    try:
+        # STEP 2: Generate verification questions (text only)
+        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 WITH the PDF page
+        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)  # Rate limit handling
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+
+            # Include PDF page in the verification question
+            payload_step3 = {
+                "contents": [{
+                    "parts": [
+                        {"text": step3_filled},
+                        {
+                            "inline_data": {
+                                "mime_type": mime_type,
+                                "data": pdf_data
+                            }
+                        }
+                    ]
+                }],
+                **({"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 WITH the PDF page
+        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))
+
+        # Include PDF page in final categorization
+        payload_step4 = {
+            "contents": [{
+                "parts": [
+                    {"text": step4_filled},
+                    {
+                        "inline_data": {
+                            "mime_type": mime_type,
+                            "data": pdf_data
+                        }
+                    }
+                ]
+            }],
+            "generationConfig": {
+                "responseMimeType": "application/json",
+                **({"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"]
+
+        return verified_reply
+
+    except Exception as e:
+        return initial_reply
+
+
+def pdf_chain_of_verification_mistral(
+    initial_reply,
+    step2_prompt,
+    step3_prompt,
+    step4_prompt,
+    client,
+    user_model,
+    creativity,
+    remove_numbering,
+    pdf_content
+):
+    """
+    Execute Chain of Verification (CoVe) process for PDF pages with Mistral AI.
+    The PDF page (as image) is included in verification steps for accurate assessment.
+    Returns the verified reply or initial reply if error occurs.
+
+    Args:
+        pdf_content: The PDF page content in Mistral format (dict with image_url after conversion)
+    """
+    try:
+        # STEP 2: Generate verification questions (text only)
+        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 WITH the PDF page
+        questions_list = [
+            remove_numbering(q)
+            for q in verification_questions.split('\n')
+            if q.strip()
+        ]
+        verification_qa = []
+
+        for question in questions_list:
+            step3_filled = step3_prompt.replace('<<QUESTION>>', question)
+
+            # Include PDF page in the verification question
+            message_content = [
+                {"type": "text", "text": step3_filled},
+                pdf_content
+            ]
+
+            answer_response = client.chat.complete(
+                model=user_model,
+                messages=[{'role': 'user', 'content': message_content}],
+                **({"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 WITH the PDF page
+        verification_qa_text = "\n\n".join(verification_qa)
+
+        step4_filled = (step4_prompt
+            .replace('<<INITIAL_REPLY>>', initial_reply)
+            .replace('<<VERIFICATION_QA>>', verification_qa_text))
+
+        # Include PDF page in final categorization
+        final_message_content = [
+            {"type": "text", "text": step4_filled},
+            pdf_content
+        ]
+
+        final_response = client.chat.complete(
+            model=user_model,
+            messages=[{'role': 'user', 'content': final_message_content}],
+            response_format={"type": "json_object"},
+            **({"temperature": creativity} if creativity is not None else {})
+        )
+
+        verified_reply = final_response.choices[0].message.content
+
+        return verified_reply
+
+    except Exception as e:
+        return initial_reply

cat_stack/calls/pdf_stepback.py

@@ -0,0 +1,210 @@
+# PDF-aware Stepback prompting functions for various LLM providers
+# These functions generate abstract insights about PDF document categorization tasks
+
+import requests
+
+
+def get_pdf_stepback_insight_openai(
+    stepback,
+    api_key,
+    user_model,
+    model_source="openai",
+    creativity=None
+):
+    """
+    Get stepback insight for PDF categorization from OpenAI-compatible APIs.
+    Supports OpenAI, Perplexity, Huggingface, and xAI.
+
+    The stepback prompt asks for abstract thinking about document categorization
+    before analyzing specific PDF pages.
+
+    Uses direct HTTP requests instead of OpenAI SDK for lighter dependencies.
+    """
+    # Determine the base URL based on model source
+    if model_source == "huggingface":
+        from cat_stack._providers import _detect_huggingface_endpoint
+        base_url = _detect_huggingface_endpoint(api_key, user_model)
+    elif model_source == "huggingface-together":
+        base_url = "https://router.huggingface.co/together/v1"
+    elif model_source == "perplexity":
+        base_url = "https://api.perplexity.ai"
+    elif model_source == "xai":
+        base_url = "https://api.x.ai/v1"
+    else:
+        base_url = "https://api.openai.com/v1"
+
+    endpoint = f"{base_url}/chat/completions"
+
+    headers = {
+        "Content-Type": "application/json",
+        "Authorization": f"Bearer {api_key}"
+    }
+
+    payload = {
+        "model": user_model,
+        "messages": [{"role": "user", "content": stepback}],
+    }
+
+    if creativity is not None:
+        payload["temperature"] = creativity
+
+    try:
+        response = requests.post(endpoint, headers=headers, json=payload, timeout=120)
+        response.raise_for_status()
+        result = response.json()
+        stepback_insight = result["choices"][0]["message"]["content"]
+
+        return stepback_insight, True
+
+    except Exception as e:
+        return None, False
+
+
+def get_pdf_stepback_insight_anthropic(
+    stepback,
+    api_key,
+    user_model,
+    model_source="anthropic",
+    creativity=None
+):
+    """
+    Get stepback insight for PDF categorization from Anthropic Claude.
+
+    Uses direct HTTP requests instead of Anthropic SDK for lighter dependencies.
+    """
+    import requests
+
+    endpoint = "https://api.anthropic.com/v1/messages"
+
+    headers = {
+        "Content-Type": "application/json",
+        "x-api-key": api_key,
+        "anthropic-version": "2023-06-01"
+    }
+
+    payload = {
+        "model": user_model,
+        "max_tokens": 4096,
+        "messages": [{"role": "user", "content": stepback}],
+    }
+
+    if creativity is not None:
+        payload["temperature"] = creativity
+
+    try:
+        response = requests.post(endpoint, headers=headers, json=payload, timeout=120)
+        response.raise_for_status()
+        result = response.json()
+
+        # Parse response - Anthropic returns content as a list
+        content = result.get("content", [])
+        if content and content[0].get("type") == "text":
+            stepback_insight = content[0].get("text", "")
+            return stepback_insight, True
+
+        return None, False
+
+    except Exception as e:
+        return None, False
+
+
+def get_pdf_stepback_insight_google(
+    stepback,
+    api_key,
+    user_model,
+    model_source="google",
+    creativity=None
+):
+    """
+    Get stepback insight for PDF categorization from Google Gemini.
+    """
+    import requests
+
+    url = f"https://generativelanguage.googleapis.com/v1beta/models/{user_model}:generateContent"
+
+    headers = {
+        "x-goog-api-key": api_key,
+        "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()
+
+        result = response.json()
+        stepback_insight = result['candidates'][0]['content']['parts'][0]['text']
+
+        return stepback_insight, True
+
+    except Exception as e:
+        return None, False
+
+
+def get_pdf_stepback_insight_mistral(
+    stepback,
+    api_key,
+    user_model,
+    model_source="mistral",
+    creativity=None
+):
+    """
+    Get stepback insight for PDF categorization from Mistral AI.
+    """
+    import requests
+
+    endpoint = "https://api.mistral.ai/v1/chat/completions"
+    headers = {
+        "Content-Type": "application/json",
+        "Authorization": f"Bearer {api_key}"
+    }
+
+    payload = {
+        "model": user_model,
+        "messages": [{'role': 'user', 'content': stepback}],
+    }
+    if creativity is not None:
+        payload["temperature"] = creativity
+
+    try:
+        response = requests.post(endpoint, headers=headers, json=payload, timeout=120)
+        response.raise_for_status()
+        result = response.json()
+        stepback_insight = result["choices"][0]["message"]["content"]
+
+        return stepback_insight, True
+
+    except Exception as e:
+        return None, False
+
+
+def get_pdf_stepback_insight(model_source, stepback, api_key, user_model, creativity):
+    """Get step-back insight using the appropriate provider for PDF tasks."""
+    stepback_functions = {
+        "openai": get_pdf_stepback_insight_openai,
+        "perplexity": get_pdf_stepback_insight_openai,
+        "huggingface": get_pdf_stepback_insight_openai,
+        "huggingface-together": get_pdf_stepback_insight_openai,
+        "xai": get_pdf_stepback_insight_openai,
+        "anthropic": get_pdf_stepback_insight_anthropic,
+        "google": get_pdf_stepback_insight_google,
+        "mistral": get_pdf_stepback_insight_mistral,
+    }
+
+    func = stepback_functions.get(model_source)
+    if func is None:
+        return None, False
+
+    return func(
+        stepback=stepback,
+        api_key=api_key,
+        user_model=user_model,
+        model_source=model_source,
+        creativity=creativity
+    )