lexoid 0.1.8.post1__py3-none-any.whl → 0.1.10__py3-none-any.whl

lexoid/api.py

@@ -19,6 +19,7 @@ from lexoid.core.utils import (
     recursive_read_html,
     router,
     split_pdf,
+    create_sub_pdf,
 )
 
 
@@ -28,20 +29,24 @@ class ParserType(Enum):
     AUTO = "AUTO"
 
 
-def parse_chunk(
-    path: str, parser_type: ParserType, raw: bool, **kwargs
-) -> List[Dict] | str:
+def parse_chunk(path: str, parser_type: ParserType, **kwargs) -> Dict:
     """
     Parses a file using the specified parser type.
 
     Args:
         path (str): The file path or URL.
         parser_type (ParserType): The type of parser to use (LLM_PARSE, STATIC_PARSE, or AUTO).
-        raw (bool): Whether to return raw text or structured data.
         **kwargs: Additional arguments for the parser.
 
     Returns:
-        List[Dict] | str: Parsed document data as a list of dictionaries or raw text.
+        Dict: Dictionary containing:
+            - raw: Full markdown content as string
+            - segments: List of dictionaries with metadata and content
+            - title: Title of the document
+            - url: URL if applicable
+            - parent_title: Title of parent doc if recursively parsed
+            - recursive_docs: List of dictionaries for recursively parsed documents
+            - token_usage: Dictionary containing token usage statistics
     """
     if parser_type == ParserType.AUTO:
         parser_type = ParserType[router(path)]
@@ -52,63 +57,81 @@ def parse_chunk(
     )
     if parser_type == ParserType.STATIC_PARSE:
         logger.debug("Using static parser")
-        return parse_static_doc(path, raw, **kwargs)
+        return parse_static_doc(path, **kwargs)
     else:
         logger.debug("Using LLM parser")
-        return parse_llm_doc(path, raw, **kwargs)
+        return parse_llm_doc(path, **kwargs)
 
 
 def parse_chunk_list(
-    file_paths: List[str], parser_type: ParserType, raw: bool, kwargs: Dict
-) -> List[Dict | str]:
+    file_paths: List[str], parser_type: ParserType, kwargs: Dict
+) -> Dict:
     """
     Parses a list of files using the specified parser type.
 
     Args:
         file_paths (list): List of file paths.
         parser_type (ParserType): The type of parser to use.
-        raw (bool): Whether to return raw text or structured data.
         kwargs (dict): Additional arguments for the parser.
 
     Returns:
-        List[Dict | str]: List of parsed documents with raw text and/or metadata.
+        Dict: Dictionary containing parsed document data
     """
-    local_docs = []
+    combined_segments = []
+    raw_texts = []
+    token_usage = {"input": 0, "output": 0}
     for file_path in file_paths:
-        result = parse_chunk(file_path, parser_type, raw, **kwargs)
-        if isinstance(result, list):
-            local_docs.extend(result)
-        else:
-            local_docs.append(result.replace("<page break>", "\n\n"))
-    return local_docs
+        result = parse_chunk(file_path, parser_type, **kwargs)
+        combined_segments.extend(result["segments"])
+        raw_texts.append(result["raw"])
+        if "token_usage" in result:
+            token_usage["input"] += result["token_usage"]["input"]
+            token_usage["output"] += result["token_usage"]["output"]
+    token_usage["total"] = token_usage["input"] + token_usage["output"]
+
+    return {
+        "raw": "\n\n".join(raw_texts),
+        "segments": combined_segments,
+        "title": kwargs.get("title", ""),
+        "url": kwargs.get("url", ""),
+        "parent_title": kwargs.get("parent_title", ""),
+        "recursive_docs": [],
+        "token_usage": token_usage,
+    }
 
 
 def parse(
     path: str,
     parser_type: Union[str, ParserType] = "LLM_PARSE",
-    raw: bool = False,
     pages_per_split: int = 4,
     max_processes: int = 4,
     **kwargs,
-) -> Union[List[Dict], str]:
+) -> Dict:
     """
     Parses a document or URL, optionally splitting it into chunks and using multiprocessing.
 
     Args:
         path (str): The file path or URL.
-        parser_type (Union[str, ParserType], optional): The type of parser to use ("LLM_PARSE", "STATIC_PARSE", or "AUTO"). Defaults to "LLM_PARSE".
-        raw (bool, optional): Whether to return raw text or structured data. Defaults to False.
-        pages_per_split (int, optional): Number of pages per split for chunking. Defaults to 4.
-        max_processes (int, optional): Maximum number of processes for parallel processing. Defaults to 4.
+        parser_type (Union[str, ParserType], optional): Parser type ("LLM_PARSE", "STATIC_PARSE", or "AUTO").
+        pages_per_split (int, optional): Number of pages per split for chunking.
+        max_processes (int, optional): Maximum number of processes for parallel processing.
         **kwargs: Additional arguments for the parser.
 
     Returns:
-        Union[List[Dict], str]: Parsed document data as a list of dictionaries or raw text.
+        Dict: Dictionary containing:
+            - raw: Full markdown content as string
+            - segments: List of dictionaries with metadata and content
+            - title: Title of the document
+            - url: URL if applicable
+            - parent_title: Title of parent doc if recursively parsed
+            - recursive_docs: List of dictionaries for recursively parsed documents
+            - token_usage: Dictionary containing token usage statistics
     """
     kwargs["title"] = os.path.basename(path)
     kwargs["pages_per_split_"] = pages_per_split
     as_pdf = kwargs.get("as_pdf", False)
     depth = kwargs.get("depth", 1)
+
     if type(parser_type) == str:
         parser_type = ParserType[parser_type]
 
@@ -120,15 +143,19 @@ def parse(
             as_pdf = True
 
         if path.startswith(("http://", "https://")):
-            download_dir = os.path.join(temp_dir, "downloads/")
+            kwargs["url"] = path
+            download_dir = kwargs.get("save_dir", os.path.join(temp_dir, "downloads/"))
             os.makedirs(download_dir, exist_ok=True)
             if is_supported_url_file_type(path):
                 path = download_file(path, download_dir)
             elif as_pdf:
-                pdf_path = os.path.join(download_dir, f"webpage_{int(time())}.pdf")
+                pdf_filename = kwargs.get("save_filename", f"webpage_{int(time())}.pdf")
+                if not pdf_filename.endswith(".pdf"):
+                    pdf_filename += ".pdf"
+                pdf_path = os.path.join(download_dir, pdf_filename)
                 path = convert_to_pdf(path, pdf_path)
             else:
-                return recursive_read_html(path, depth, raw)
+                return recursive_read_html(path, depth)
 
         assert is_supported_file_type(
             path
@@ -138,11 +165,15 @@ def parse(
             pdf_path = os.path.join(temp_dir, "converted.pdf")
             path = convert_to_pdf(path, pdf_path)
 
+        if "page_nums" in kwargs and path.lower().endswith(".pdf"):
+            sub_pdf_dir = os.path.join(temp_dir, "sub_pdfs")
+            os.makedirs(sub_pdf_dir, exist_ok=True)
+            sub_pdf_path = os.path.join(sub_pdf_dir, f"{os.path.basename(path)}")
+            path = create_sub_pdf(path, sub_pdf_path, kwargs["page_nums"])
+
         if not path.lower().endswith(".pdf") or parser_type == ParserType.STATIC_PARSE:
             kwargs["split"] = False
-            all_docs = parse_chunk(path, parser_type, raw, **kwargs)
-            if raw:
-                all_docs = [all_docs]
+            result = parse_chunk(path, parser_type, **kwargs)
         else:
             kwargs["split"] = True
             split_dir = os.path.join(temp_dir, "splits/")
@@ -156,22 +187,39 @@ def parse(
                 for i in range(0, len(split_files), chunk_size)
             ]
 
-            process_args = [(chunk, parser_type, raw, kwargs) for chunk in file_chunks]
+            process_args = [(chunk, parser_type, kwargs) for chunk in file_chunks]
 
             if max_processes == 1 or len(file_chunks) == 1:
-                all_docs = [parse_chunk_list(*args) for args in process_args]
+                chunk_results = [parse_chunk_list(*args) for args in process_args]
             else:
                 with ProcessPoolExecutor(max_workers=max_processes) as executor:
-                    all_docs = list(executor.map(parse_chunk_list, *zip(*process_args)))
-
-            all_docs = [item for sublist in all_docs for item in sublist]
+                    chunk_results = list(
+                        executor.map(parse_chunk_list, *zip(*process_args))
+                    )
+
+            # Combine results from all chunks
+            result = {
+                "raw": "\n\n".join(r["raw"] for r in chunk_results),
+                "segments": [seg for r in chunk_results for seg in r["segments"]],
+                "title": kwargs["title"],
+                "url": kwargs.get("url", ""),
+                "parent_title": kwargs.get("parent_title", ""),
+                "recursive_docs": [],
+                "token_usage": {
+                    "input": sum(r["token_usage"]["input"] for r in chunk_results),
+                    "output": sum(r["token_usage"]["output"] for r in chunk_results),
+                    "total": sum(r["token_usage"]["total"] for r in chunk_results),
+                },
+            }
+            if as_pdf:
+                result["pdf_path"] = path
 
     if depth > 1:
-        new_docs = all_docs.copy()
-        for doc in all_docs:
+        recursive_docs = []
+        for segment in result["segments"]:
             urls = re.findall(
                 r'https?://[^\s<>"\']+|www\.[^\s<>"\']+(?:\.[^\s<>"\']+)*',
-                doc if raw else doc["content"],
+                segment["content"],
             )
             for url in urls:
                 if "](" in url:
@@ -182,19 +230,16 @@ def parse(
 
                 kwargs_cp = kwargs.copy()
                 kwargs_cp["depth"] = depth - 1
-                res = parse(
+                kwargs_cp["parent_title"] = result["title"]
+                sub_doc = parse(
                     url,
                     parser_type=parser_type,
-                    raw=raw,
                     pages_per_split=pages_per_split,
                     max_processes=max_processes,
                     **kwargs_cp,
                 )
+                recursive_docs.append(sub_doc)
 
-                if raw:
-                    new_docs.append(res)
-                else:
-                    new_docs.extend(res)
-        all_docs = new_docs
+        result["recursive_docs"] = recursive_docs
 
-    return "\n".join(all_docs) if raw else all_docs
+    return result

lexoid/core/parse_type/llm_parser.py

@@ -18,6 +18,7 @@ from lexoid.core.prompt_templates import (
 from lexoid.core.utils import convert_image_to_pdf
 from loguru import logger
 from openai import OpenAI
+from together import Together
 from huggingface_hub import InferenceClient
 
 
@@ -33,38 +34,36 @@ def retry_on_http_error(func):
                 return func(*args, **kwargs)
             except HTTPError as e:
                 logger.error(f"Retry failed: {e}")
-                if kwargs.get("raw", False):
-                    return ""
-                return [
-                    {
-                        "metadata": {
-                            "title": kwargs["title"],
-                            "page": kwargs.get("start", 0),
-                        },
-                        "content": "",
-                    }
-                ]
+                return {
+                    "raw": "",
+                    "segments": [],
+                    "title": kwargs["title"],
+                    "url": kwargs.get("url", ""),
+                    "parent_title": kwargs.get("parent_title", ""),
+                    "recursive_docs": [],
+                    "error": f"HTTPError encountered on page {kwargs.get('start', 0)}: {e}",
+                }
 
     return wrapper
 
 
 @retry_on_http_error
-def parse_llm_doc(path: str, raw: bool, **kwargs) -> List[Dict] | str:
+def parse_llm_doc(path: str, **kwargs) -> List[Dict] | str:
     if "model" not in kwargs:
         kwargs["model"] = "gemini-1.5-flash"
     model = kwargs.get("model")
     if model.startswith("gemini"):
-        return parse_with_gemini(path, raw, **kwargs)
+        return parse_with_gemini(path, **kwargs)
     if model.startswith("gpt"):
-        return parse_with_api(path, raw, api="openai", **kwargs)
+        return parse_with_api(path, api="openai", **kwargs)
     if model.startswith("meta-llama"):
         if model.endswith("Turbo") or model == "meta-llama/Llama-Vision-Free":
-            return parse_with_together(path, raw, **kwargs)
-        return parse_with_api(path, raw, api="huggingface", **kwargs)
+            return parse_with_api(path, api="together", **kwargs)
+        return parse_with_api(path, api="huggingface", **kwargs)
     raise ValueError(f"Unsupported model: {model}")
 
 
-def parse_with_gemini(path: str, raw: bool, **kwargs) -> List[Dict] | str:
+def parse_with_gemini(path: str, **kwargs) -> List[Dict] | str:
     api_key = os.environ.get("GOOGLE_API_KEY")
     if not api_key:
         raise ValueError("GOOGLE_API_KEY environment variable is not set")
@@ -119,25 +118,33 @@ def parse_with_gemini(path: str, raw: bool, **kwargs) -> List[Dict] | str:
         if "text" in part
     )
 
-    result = ""
+    combined_text = ""
     if "<output>" in raw_text:
-        result = raw_text.split("<output>")[1].strip()
+        combined_text = raw_text.split("<output>")[1].strip()
     if "</output>" in result:
-        result = result.split("</output>")[0].strip()
-
-    if raw:
-        return result
-
-    return [
-        {
-            "metadata": {
-                "title": kwargs["title"],
-                "page": kwargs.get("start", 0) + page_no,
-            },
-            "content": page,
-        }
-        for page_no, page in enumerate(result.split("<page-break>"), start=1)
-    ]
+        combined_text = result.split("</output>")[0].strip()
+
+    token_usage = result["usageMetadata"]
+    input_tokens = token_usage.get("promptTokenCount", 0)
+    output_tokens = token_usage.get("candidatesTokenCount", 0)
+    total_tokens = input_tokens + output_tokens
+
+    return {
+        "raw": combined_text,
+        "segments": [
+            {"metadata": {"page": kwargs.get("start", 0) + page_no}, "content": page}
+            for page_no, page in enumerate(combined_text.split("<page-break>"), start=1)
+        ],
+        "title": kwargs["title"],
+        "url": kwargs.get("url", ""),
+        "parent_title": kwargs.get("parent_title", ""),
+        "recursive_docs": [],
+        "token_usage": {
+            "input": input_tokens,
+            "output": output_tokens,
+            "total": total_tokens,
+        },
+    }
 
 
 def convert_pdf_page_to_base64(
@@ -155,97 +162,17 @@ def convert_pdf_page_to_base64(
     return base64.b64encode(img_byte_arr.getvalue()).decode("utf-8")
 
 
-def parse_with_together(path: str, raw: bool, **kwargs) -> List[Dict] | str:
-    api_key = os.environ.get("TOGETHER_API_KEY")
-    if not api_key:
-        raise ValueError("TOGETHER_API_KEY environment variable is not set")
-
-    url = "https://api.together.xyz/v1/chat/completions"
-    headers = {
-        "Authorization": f"Bearer {api_key}",
-        "Content-Type": "application/json",
-    }
-
-    mime_type, _ = mimetypes.guess_type(path)
-    if mime_type and mime_type.startswith("image"):
-        with open(path, "rb") as img_file:
-            image_base64 = base64.b64encode(img_file.read()).decode("utf-8")
-            images = [(0, f"data:{mime_type};base64,{image_base64}")]
-    else:
-        pdf_document = pdfium.PdfDocument(path)
-        images = [
-            (
-                page_num,
-                f"data:image/png;base64,{convert_pdf_page_to_base64(pdf_document, page_num)}",
-            )
-            for page_num in range(len(pdf_document))
-        ]
-
-    all_results = []
-    for page_num, image_url in images:
-        messages = [
-            {
-                "role": "user",
-                "content": [
-                    {"type": "text", "text": LLAMA_PARSER_PROMPT},
-                    {"type": "image_url", "image_url": {"url": image_url}},
-                ],
-            }
-        ]
-
-        payload = {
-            "model": kwargs["model"],
-            "messages": messages,
-            "max_tokens": kwargs.get("max_tokens", 1024),
-            "temperature": kwargs.get("temperature", 0.7),
-        }
-
-        response = requests.post(url, json=payload, headers=headers)
-        response.raise_for_status()
-        response_data = response.json()
-
-        page_text = response_data["choices"][0]["message"]["content"]
-        if kwargs.get("verbose", None):
-            logger.debug(f"Page {page_num + 1} response: {page_text}")
-
-        result = page_text
-        if "<output>" in page_text:
-            result = page_text.split("<output>")[1].strip()
-        if "</output>" in result:
-            result = result.split("</output>")[0].strip()
-        all_results.append((page_num, result))
-
-    all_results.sort(key=lambda x: x[0])
-    all_texts = [text for _, text in all_results]
-    combined_text = "<page-break>".join(all_texts)
-
-    if raw:
-        return combined_text
-
-    return [
-        {
-            "metadata": {
-                "title": kwargs["title"],
-                "page": kwargs.get("start", 0) + page_no,
-            },
-            "content": page,
-        }
-        for page_no, page in enumerate(all_texts, start=1)
-    ]
-
-
-def parse_with_api(path: str, raw: bool, api: str, **kwargs) -> List[Dict] | str:
+def parse_with_api(path: str, api: str, **kwargs) -> List[Dict] | str:
     """
     Parse documents (PDFs or images) using various vision model APIs.
 
     Args:
         path (str): Path to the document to parse
-        raw (bool): If True, return raw text; if False, return structured data
-        api (str): Which API to use ("openai" or "huggingface")
+        api (str): Which API to use ("openai", "huggingface", or "together")
         **kwargs: Additional arguments including model, temperature, title, etc.
 
     Returns:
-        List[Dict] | str: Parsed content either as raw text or structured data
+        Dict: Dictionary containing parsed document data
     """
     # Initialize appropriate client
     clients = {
@@ -253,6 +180,7 @@ def parse_with_api(path: str, raw: bool, api: str, **kwargs) -> List[Dict] | str
         "huggingface": lambda: InferenceClient(
             token=os.environ["HUGGINGFACEHUB_API_TOKEN"]
         ),
+        "together": lambda: Together(),
     }
     assert api in clients, f"Unsupported API: {api}"
     logger.debug(f"Parsing with {api} API and model {kwargs['model']}")
@@ -329,6 +257,7 @@ def parse_with_api(path: str, raw: bool, api: str, **kwargs) -> List[Dict] | str
 
         # Get completion from selected API
         response = client.chat.completions.create(**completion_params)
+        token_usage = response.usage
 
         # Extract the response text
         page_text = response.choices[0].message.content
@@ -341,23 +270,44 @@ def parse_with_api(path: str, raw: bool, api: str, **kwargs) -> List[Dict] | str
             result = page_text.split("<output>")[1].strip()
         if "</output>" in result:
             result = result.split("</output>")[0].strip()
-        all_results.append((page_num, result))
+        all_results.append(
+            (
+                page_num,
+                result,
+                token_usage.prompt_tokens,
+                token_usage.completion_tokens,
+                token_usage.total_tokens,
+            )
+        )
 
     # Sort results by page number and combine
     all_results.sort(key=lambda x: x[0])
-    all_texts = [text for _, text in all_results]
+    all_texts = [text for _, text, _, _, _ in all_results]
     combined_text = "<page-break>".join(all_texts)
 
-    if raw:
-        return combined_text
-
-    return [
-        {
-            "metadata": {
-                "title": kwargs["title"],
-                "page": kwargs.get("start", 0) + page_no,
-            },
-            "content": page,
-        }
-        for page_no, page in enumerate(all_texts, start=1)
-    ]
+    return {
+        "raw": combined_text,
+        "segments": [
+            {
+                "metadata": {
+                    "page": kwargs.get("start", 0) + page_no + 1,
+                    "token_usage": {
+                        "input": input_tokens,
+                        "output": output_tokens,
+                        "total": total_tokens,
+                    },
+                },
+                "content": page,
+            }
+            for page_no, page, input_tokens, output_tokens, total_tokens in all_results
+        ],
+        "title": kwargs["title"],
+        "url": kwargs.get("url", ""),
+        "parent_title": kwargs.get("parent_title", ""),
+        "recursive_docs": [],
+        "token_usage": {
+            "input": sum(input_tokens for _, _, input_tokens, _, _ in all_results),
+            "output": sum(output_tokens for _, _, _, output_tokens, _ in all_results),
+            "total": sum(total_tokens for _, _, _, _, total_tokens in all_results),
+        },
+    }

lexoid/core/parse_type/static_parser.py

@@ -9,73 +9,89 @@ from pdfplumber.utils import get_bbox_overlap, obj_to_bbox
 from docx import Document
 
 
-def parse_static_doc(path: str, raw: bool, **kwargs) -> List[Dict] | str:
+def parse_static_doc(path: str, **kwargs) -> Dict:
+    """
+    Parses a document using static parsing methods.
+
+    Args:
+        path (str): The file path.
+        **kwargs: Additional arguments for parsing.
+
+    Returns:
+        Dict: Dictionary containing parsed document data
+    """
     framework = kwargs.get("framework", "pdfplumber")
 
     file_type = get_file_type(path)
     if file_type == "application/pdf":
         if framework == "pdfplumber":
-            return parse_with_pdfplumber(path, raw, **kwargs)
+            return parse_with_pdfplumber(path, **kwargs)
         elif framework == "pdfminer":
-            return parse_with_pdfminer(path, raw, **kwargs)
+            return parse_with_pdfminer(path, **kwargs)
         else:
             raise ValueError(f"Unsupported framework: {framework}")
     elif "wordprocessing" in file_type:
-        return parse_with_docx(path, raw, **kwargs)
+        return parse_with_docx(path, **kwargs)
     elif file_type == "text/html":
         with open(path, "r") as f:
             html_content = f.read()
-            return html_to_markdown(html_content, raw, kwargs["title"])
+            return html_to_markdown(html_content, kwargs["title"])
     elif file_type == "text/plain":
         with open(path, "r") as f:
             content = f.read()
-            if raw:
-                return content
-            else:
-                return [
-                    {
-                        "metadata": {"title": kwargs["title"], "page": 1},
-                        "content": content,
-                    }
-                ]
+            return {
+                "raw": content,
+                "segments": [{"metadata": {"page": 1}, "content": content}],
+                "title": kwargs["title"],
+                "url": kwargs.get("url", ""),
+                "parent_title": kwargs.get("parent_title", ""),
+                "recursive_docs": [],
+            }
     elif file_type == "text/csv":
         df = pd.read_csv(path)
         content = df.to_markdown(index=False)
-        if raw:
-            return content
-        else:
-            return [
-                {
-                    "metadata": {"title": kwargs["title"], "page": 1},
-                    "content": content,
-                }
-            ]
+        return {
+            "raw": content,
+            "segments": [{"metadata": {"page": 1}, "content": content}],
+            "title": kwargs["title"],
+            "url": kwargs.get("url", ""),
+            "parent_title": kwargs.get("parent_title", ""),
+            "recursive_docs": [],
+        }
     else:
         raise ValueError(f"Unsupported file type: {file_type}")
 
 
-def parse_with_pdfminer(path: str, raw: bool, **kwargs) -> List[Dict] | str:
+def parse_with_pdfminer(path: str, **kwargs) -> Dict:
+    """
+    Parse PDF using pdfminer.
+
+    Returns:
+        Dict: Dictionary containing parsed document data
+    """
     pages = list(extract_pages(path))
-    docs = []
+    segments = []
+    raw_texts = []
+
     for page_num, page_layout in enumerate(pages, start=1):
         page_text = "".join(
             element.get_text()
             for element in page_layout
             if isinstance(element, LTTextContainer)
         )
-        if raw:
-            docs.append(page_text)
-        else:
-            docs.append(
-                {
-                    "metadata": {
-                        "title": kwargs["title"],
-                        "page": kwargs["start"] + page_num,
-                    },
-                    "content": page_text,
-                }
-            )
-    return "\n".join(docs) if raw else docs
+        raw_texts.append(page_text)
+        segments.append(
+            {"metadata": {"page": kwargs["start"] + page_num}, "content": page_text}
+        )
+
+    return {
+        "raw": "\n".join(raw_texts),
+        "segments": segments,
+        "title": kwargs["title"],
+        "url": kwargs.get("url", ""),
+        "parent_title": kwargs.get("parent_title", ""),
+        "recursive_docs": [],
+    }
 
 
 def process_table(table) -> str:
@@ -359,44 +375,44 @@ def process_pdf_with_pdfplumber(path: str, **kwargs) -> List[str]:
     return page_texts
 
 
-def parse_with_pdfplumber(path: str, raw: bool, **kwargs) -> List[Dict] | str:
+def parse_with_pdfplumber(path: str, **kwargs) -> Dict:
     """
-    Parse PDF and return either raw text or structured data.
-
-    Args:
-        path (str): Path to the PDF file
-        raw (bool): If True, return raw text with page breaks; if False, return structured data
-        **kwargs: Additional arguments including 'title' and 'start' page number
+    Parse PDF using pdfplumber.
 
     Returns:
-        Union[List[Dict], str]: Either a list of dictionaries containing page metadata and content,
-                               or a string of raw text with page breaks
+        Dict: Dictionary containing parsed document data
     """
     page_texts = process_pdf_with_pdfplumber(path)
-    if raw:
-        return "<page-break>".join(page_texts)
-    return [
-        {
-            "metadata": {"title": kwargs["title"], "page": kwargs["start"] + page_num},
-            "content": page_text,
-        }
+    segments = [
+        {"metadata": {"page": kwargs["start"] + page_num}, "content": page_text}
         for page_num, page_text in enumerate(page_texts, start=1)
     ]
 
+    return {
+        "raw": "<page-break>".join(page_texts),
+        "segments": segments,
+        "title": kwargs["title"],
+        "url": kwargs.get("url", ""),
+        "parent_title": kwargs.get("parent_title", ""),
+        "recursive_docs": [],
+    }
+
 
-def parse_with_docx(path: str, raw: bool, **kwargs) -> List[Dict] | str:
+def parse_with_docx(path: str, **kwargs) -> Dict:
+    """
+    Parse DOCX document.
+
+    Returns:
+        Dict: Dictionary containing parsed document data
+    """
     doc = Document(path)
     full_text = "\n".join([paragraph.text for paragraph in doc.paragraphs])
 
-    if raw:
-        return full_text
-
-    return [
-        {
-            "metadata": {
-                "title": kwargs["title"],
-                "page": kwargs["start"] + 1,
-            },
-            "content": full_text,
-        }
-    ]
+    return {
+        "raw": full_text,
+        "segments": [{"metadata": {"page": kwargs["start"] + 1}, "content": full_text}],
+        "title": kwargs["title"],
+        "url": kwargs.get("url", ""),
+        "parent_title": kwargs.get("parent_title", ""),
+        "recursive_docs": [],
+    }

lexoid/core/utils.py

@@ -5,7 +5,8 @@ import os
 import re
 import sys
 from difflib import SequenceMatcher
-from typing import Dict, List, Union
+from hashlib import md5
+from typing import Dict, List, Optional
 from urllib.parse import urlparse
 
 import nest_asyncio
@@ -44,6 +45,20 @@ def split_pdf(input_path: str, output_dir: str, pages_per_split: int):
     return paths
 
 
+def create_sub_pdf(
+    input_path: str, output_path: str, page_nums: Optional[tuple[int, ...]|int] = None
+) -> str:
+    if isinstance(page_nums, int):
+        page_nums = (page_nums,)
+    page_nums = tuple(sorted(set(page_nums)))
+    with pikepdf.open(input_path) as pdf:
+        indices = page_nums if page_nums else range(len(pdf.pages))
+        with pikepdf.new() as new_pdf:
+            new_pdf.pages.extend([pdf.pages[i - 1] for i in indices])
+            new_pdf.save(output_path)
+    return output_path
+
+
 def convert_image_to_pdf(image_path: str) -> bytes:
     with Image.open(image_path) as img:
         img_rgb = img.convert("RGB")
@@ -184,14 +199,11 @@ def find_dominant_heading_level(markdown_content: str) -> str:
     return min(heading_counts.keys(), key=len)
 
 
-def split_md_by_headings(
-    markdown_content: str, heading_pattern: str, title: str
-) -> List[Dict]:
+def split_md_by_headings(markdown_content: str, heading_pattern: str) -> List[Dict]:
     """
     Splits markdown content by the specified heading pattern and structures it.
 
     Args:
-        url (str): The URL of the HTML page
         markdown_content (str): The markdown content to split
         heading_pattern (str): The heading pattern to split on (e.g., '##' or 'underline')
 
@@ -211,7 +223,7 @@ def split_md_by_headings(
         if sections and not re.match(r"^[^\n]+\n-+$", sections[0], re.MULTILINE):
             structured_content.append(
                 {
-                    "metadata": {"title": title, "page": "Introduction"},
+                    "metadata": {"page": "Introduction"},
                     "content": sections.pop(0),
                 }
             )
@@ -221,7 +233,7 @@ def split_md_by_headings(
             if i + 1 < len(sections):
                 structured_content.append(
                     {
-                        "metadata": {"title": title, "page": sections[i]},
+                        "metadata": {"page": sections[i]},
                         "content": sections[i + 1],
                     }
                 )
@@ -238,7 +250,7 @@ def split_md_by_headings(
         if len(sections) > len(headings):
             structured_content.append(
                 {
-                    "metadata": {"title": title, "page": "Introduction"},
+                    "metadata": {"page": "Introduction"},
                     "content": sections.pop(0),
                 }
             )
@@ -248,7 +260,7 @@ def split_md_by_headings(
             clean_heading = heading.replace(heading_pattern, "").strip()
             structured_content.append(
                 {
-                    "metadata": {"title": title, "page": clean_heading},
+                    "metadata": {"page": clean_heading},
                     "content": content,
                 }
             )
@@ -256,39 +268,47 @@ def split_md_by_headings(
     return structured_content
 
 
-def html_to_markdown(html: str, raw: bool, title: str) -> str:
+def html_to_markdown(html: str, title: str, url: str) -> str:
     """
     Converts HTML content to markdown.
 
     Args:
         html (str): The HTML content to convert.
-        raw (bool): Whether to return raw markdown text or structured data.
+        title (str): The title of the HTML page
+        url (str): The URL of the HTML page
 
     Returns:
-        Union[str, List[Dict]]: Either raw markdown content or structured data with metadata and content sections.
+        Dict: Dictionary containing parsed document data
     """
     markdown_content = md(html)
 
-    if raw:
-        return markdown_content
-
     # Find the dominant heading level
     heading_pattern = find_dominant_heading_level(markdown_content)
 
     # Split content by headings and structure it
-    return split_md_by_headings(markdown_content, heading_pattern, title)
+    split_md = split_md_by_headings(markdown_content, heading_pattern)
+
+    content = {
+        "raw": markdown_content,
+        "segments": split_md,
+        "title": title,
+        "url": url,
+        "parent_title": "",
+        "recursive_docs": [],
+    }
+
+    return content
 
 
-def read_html_content(url: str, raw: bool = False) -> Union[str, List[Dict]]:
+def read_html_content(url: str) -> Dict:
     """
     Reads the content of an HTML page from the given URL and converts it to markdown or structured content.
 
     Args:
         url (str): The URL of the HTML page.
-        raw (bool): Whether to return raw markdown text or structured data.
 
     Returns:
-        Union[str, List[Dict]]: Either raw markdown content or structured data with metadata and content sections.
+        Dict: Dictionary containing parsed document data
     """
 
     try:
@@ -351,7 +371,10 @@ def read_html_content(url: str, raw: bool = False) -> Union[str, List[Dict]]:
         soup = BeautifulSoup(
             response.content, "html.parser", from_encoding="iso-8859-1"
         )
-    return html_to_markdown(str(soup), raw, title=url)
+    title = soup.title.string.strip() if soup.title else "No title"
+    url_hash = md5(url.encode("utf-8")).hexdigest()[:8]
+    full_title = f"{title} - {url_hash}"
+    return html_to_markdown(str(soup), title=full_title, url=url)
 
 
 def extract_urls_from_markdown(content: str) -> List[str]:
@@ -378,61 +401,60 @@ def extract_urls_from_markdown(content: str) -> List[str]:
     return list(set(urls))  # Remove duplicates
 
 
-def recursive_read_html(
-    url: str, depth: int, raw: bool, visited_urls: set = None
-) -> Union[str, List[Dict]]:
+def recursive_read_html(url: str, depth: int, visited_urls: set = None) -> Dict:
     """
     Recursively reads HTML content from URLs up to specified depth.
 
     Args:
         url (str): The URL to parse
         depth (int): How many levels deep to recursively parse
-        raw (bool): Whether to return raw text or structured data
         visited_urls (set): Set of already visited URLs to prevent cycles
 
     Returns:
-        Union[str, List[Dict]]: Combined content from all parsed URLs
+        Dict: Dictionary containing parsed document data
     """
     if visited_urls is None:
         visited_urls = set()
 
     if url in visited_urls:
-        return "" if raw else []
+        return {
+            "raw": "",
+            "segments": [],
+            "title": "",
+            "url": url,
+            "parent_title": "",
+            "recursive_docs": [],
+        }
 
     visited_urls.add(url)
 
     try:
-        content = read_html_content(url, raw)
+        content = read_html_content(url)
     except Exception as e:
         print(f"Error processing URL {url}: {str(e)}")
-        return "" if raw else []
+        return {
+            "raw": "",
+            "segments": [],
+            "title": "",
+            "url": url,
+            "parent_title": "",
+            "recursive_docs": [],
+        }
 
     if depth <= 1:
         return content
 
-    # Extract URLs from the content
-    if raw:
-        urls = extract_urls_from_markdown(content)
-    else:
-        # Extract URLs from all content sections
-        urls = []
-        for doc in content:
-            urls.extend(extract_urls_from_markdown(doc["content"]))
+    # Extract URLs from all content sections
+    urls = extract_urls_from_markdown(content["raw"])
 
     # Recursively process each URL
+    recursive_docs = []
     for sub_url in urls:
         if sub_url not in visited_urls:
-            sub_content = recursive_read_html(sub_url, depth - 1, raw, visited_urls)
-
-            if raw:
-                if sub_content:
-                    content += f"\n\n--- Begin content from {sub_url} ---\n\n"
-                    content += sub_content
-                    content += f"\n\n--- End content from {sub_url} ---\n\n"
-            else:
-                if isinstance(sub_content, list):
-                    content.extend(sub_content)
+            sub_content = recursive_read_html(sub_url, depth - 1, visited_urls)
+            recursive_docs.append(sub_content)
 
+    content["recursive_docs"] = recursive_docs
     return content
 
 

{lexoid-0.1.8.post1.dist-info → lexoid-0.1.10.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: lexoid
-Version: 0.1.8.post1
+Version: 0.1.10
 Summary: 
 Requires-Python: >=3.10,<4.0
 Classifier: Programming Language :: Python :: 3
@@ -28,6 +28,7 @@ Requires-Dist: pyqtwebengine (>=5.15.7,<6.0.0) ; platform_system != "debian"
 Requires-Dist: python-docx (>=1.1.2,<2.0.0)
 Requires-Dist: python-dotenv (>=1.0.0,<2.0.0)
 Requires-Dist: tabulate (>=0.9.0,<0.10.0)
+Requires-Dist: together (>=1.4.0,<2.0.0)
 Description-Content-Type: text/markdown
 
 # Lexoid
@@ -93,10 +94,10 @@ Here's a quick example to parse documents using Lexoid:
 from lexoid.api import parse
 from lexoid.api import ParserType
 
-parsed_md = parse("https://www.justice.gov/eoir/immigration-law-advisor", parser_type="LLM_PARSE", raw=True)
+parsed_md = parse("https://www.justice.gov/eoir/immigration-law-advisor", parser_type="LLM_PARSE")["raw"]
 # or
 pdf_path = "path/to/immigration-law-advisor.pdf"
-parsed_md = parse(pdf_path, parser_type="LLM_PARSE", raw=True)
+parsed_md = parse(pdf_path, parser_type="LLM_PARSE")["raw"]
 
 print(parsed_md)
 ```
@@ -104,7 +105,6 @@ print(parsed_md)
 ### Parameters
 - path (str): The file path or URL.
 - parser_type (str, optional): The type of parser to use ("LLM_PARSE" or "STATIC_PARSE"). Defaults to "AUTO".
-- raw (bool, optional): Return raw text or structured data. Defaults to False.
 - pages_per_split (int, optional): Number of pages per split for chunking. Defaults to 4.
 - max_threads (int, optional): Maximum number of threads for parallel processing. Defaults to 4.
 - **kwargs: Additional arguments for the parser.

lexoid-0.1.10.dist-info/RECORD

@@ -0,0 +1,9 @@
+lexoid/api.py,sha256=45nkTuQcxdppeUiRsiyioJtvlVeWeoq_WgKtGCthIBY,9193
+lexoid/core/parse_type/llm_parser.py,sha256=tH19B0w78OowkDdqJg3rom0kQmyuTaTfDP98Qnwufo0,10625
+lexoid/core/parse_type/static_parser.py,sha256=j3khirFnXq2j3IFEu0TsYWA5sHMpe_oQLFM9Uv3hScM,14100
+lexoid/core/prompt_templates.py,sha256=svSMH0yhm6ZjtOeTtUUEiCYi81ebVY9EZKPKP0Q921o,6311
+lexoid/core/utils.py,sha256=HT37qmdhPpUNN6O571G7ItE5K2Mv8SreBHmxrhdiXA8,18951
+lexoid-0.1.10.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+lexoid-0.1.10.dist-info/METADATA,sha256=4uhJ_IaHEKPl9lxKg8RRrBQ5dn7oB23XCnJNG5sNpH4,4576
+lexoid-0.1.10.dist-info/WHEEL,sha256=Nq82e9rUAnEjt98J6MlVmMCZb-t9cYE2Ir1kpBmnWfs,88
+lexoid-0.1.10.dist-info/RECORD,,

lexoid-0.1.8.post1.dist-info/RECORD

@@ -1,9 +0,0 @@
-lexoid/api.py,sha256=0we4A-U_nWMg57ZusthN27f1zXde2igi9jQujrjXthw,7120
-lexoid/core/parse_type/llm_parser.py,sha256=JsrVALlK4h2j8URSgNIhdWPB6chWXrNrMlImtxVTyyU,11833
-lexoid/core/parse_type/static_parser.py,sha256=NlAE_WMMNvNnVo2aQrA9mN3fwJ6ZrshMC8S9kG0h8CA,13772
-lexoid/core/prompt_templates.py,sha256=svSMH0yhm6ZjtOeTtUUEiCYi81ebVY9EZKPKP0Q921o,6311
-lexoid/core/utils.py,sha256=peWuMVTk90-j0aSDaRnwigpoAz_Q5y8vSosCDc6Zl3g,18642
-lexoid-0.1.8.post1.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-lexoid-0.1.8.post1.dist-info/METADATA,sha256=mz8A_92-GrLfOmT8UYcIxWIEkcskad_9vSnNnlbE4dI,4625
-lexoid-0.1.8.post1.dist-info/WHEEL,sha256=Nq82e9rUAnEjt98J6MlVmMCZb-t9cYE2Ir1kpBmnWfs,88
-lexoid-0.1.8.post1.dist-info/RECORD,,