[Feature] Gmail Loader (#841)

embedchain/chunkers/gmail.py

@@ -0,0 +1,22 @@
+from typing import Optional
+
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+
+from embedchain.chunkers.base_chunker import BaseChunker
+from embedchain.config.add_config import ChunkerConfig
+from embedchain.helper.json_serializable import register_deserializable
+
+
+@register_deserializable
+class GmailChunker(BaseChunker):
+    """Chunker for gmail."""
+
+    def __init__(self, config: Optional[ChunkerConfig] = None):
+        if config is None:
+            config = ChunkerConfig(chunk_size=1000, chunk_overlap=0, length_function=len)
+        text_splitter = RecursiveCharacterTextSplitter(
+            chunk_size=config.chunk_size,
+            chunk_overlap=config.chunk_overlap,
+            length_function=config.length_function,
+        )
+        super().__init__(text_splitter)

embedchain/data_formatter/data_formatter.py

@@ -1,6 +1,7 @@
 from embedchain.chunkers.base_chunker import BaseChunker
 from embedchain.chunkers.docs_site import DocsSiteChunker
 from embedchain.chunkers.docx_file import DocxFileChunker
+from embedchain.chunkers.gmail import GmailChunker
 from embedchain.chunkers.images import ImagesChunker
 from embedchain.chunkers.json import JSONChunker
 from embedchain.chunkers.mdx import MdxChunker
@@ -22,6 +23,7 @@ from embedchain.loaders.base_loader import BaseLoader
 from embedchain.loaders.csv import CsvLoader
 from embedchain.loaders.docs_site_loader import DocsSiteLoader
 from embedchain.loaders.docx_file import DocxFileLoader
+from embedchain.loaders.gmail import GmailLoader
 from embedchain.loaders.images import ImagesLoader
 from embedchain.loaders.json import JSONLoader
 from embedchain.loaders.local_qna_pair import LocalQnaPairLoader
@@ -84,6 +86,7 @@ class DataFormatter(JSONSerializable):
             DataType.UNSTRUCTURED: UnstructuredLoader,
             DataType.JSON: JSONLoader,
             DataType.OPENAPI: OpenAPILoader,
+            DataType.GMAIL: GmailLoader,
         }
         lazy_loaders = {DataType.NOTION}
         if data_type in loaders:
@@ -128,6 +131,7 @@ class DataFormatter(JSONSerializable):
             DataType.UNSTRUCTURED: UnstructuredFileChunker,
             DataType.JSON: JSONChunker,
             DataType.OPENAPI: OpenAPIChunker,
+            DataType.GMAIL: GmailChunker,
         }
         if data_type in chunker_classes:
             chunker_class: type = chunker_classes[data_type]

embedchain/loaders/gmail.py

@@ -0,0 +1,124 @@
+import hashlib
+import logging
+import os
+import quopri
+from textwrap import dedent
+
+from bs4 import BeautifulSoup
+
+try:
+    from llama_index import download_loader
+except ImportError:
+    raise ImportError("Notion requires extra dependencies. Install with `pip install embedchain[community]`") from None
+
+from embedchain.loaders.base_loader import BaseLoader
+from embedchain.utils import clean_string
+
+
+def get_header(text: str, header: str) -> str:
+    start_string_position = text.find(header)
+    pos_start = text.find(":", start_string_position) + 1
+    pos_end = text.find("\n", pos_start)
+    header = text[pos_start:pos_end]
+    return header.strip()
+
+
+class GmailLoader(BaseLoader):
+    def load_data(self, query):
+        """Load data from gmail."""
+        if not os.path.isfile("credentials.json"):
+            raise FileNotFoundError(
+                "You must download the valid credentials file from your google \
+                dev account. Refer this `https://cloud.google.com/docs/authentication/api-keys`"
+            )
+
+        GmailReader = download_loader("GmailReader")
+        loader = GmailReader(query=query, service=None, results_per_page=20)
+        documents = loader.load_data()
+        logging.info(f"Gmail Loader: {len(documents)} mails found for query- {query}")
+
+        data = []
+        data_contents = []
+        logging.info(f"Gmail Loader: {len(documents)} mails found")
+        for document in documents:
+            original_size = len(document.text)
+
+            snippet = document.metadata.get("snippet")
+            meta_data = {
+                "url": document.metadata.get("id"),
+                "date": get_header(document.text, "Date"),
+                "subject": get_header(document.text, "Subject"),
+                "from": get_header(document.text, "From"),
+                "to": get_header(document.text, "To"),
+                "search_query": query,
+            }
+
+            # Decode
+            decoded_bytes = quopri.decodestring(document.text)
+            decoded_str = decoded_bytes.decode("utf-8", errors="replace")
+
+            # Slice
+            mail_start = decoded_str.find("<!DOCTYPE")
+            email_data = decoded_str[mail_start:]
+
+            # Web Page HTML Processing
+            soup = BeautifulSoup(email_data, "html.parser")
+
+            tags_to_exclude = [
+                "nav",
+                "aside",
+                "form",
+                "header",
+                "noscript",
+                "svg",
+                "canvas",
+                "footer",
+                "script",
+                "style",
+            ]
+
+            for tag in soup(tags_to_exclude):
+                tag.decompose()
+
+            ids_to_exclude = ["sidebar", "main-navigation", "menu-main-menu"]
+            for id in ids_to_exclude:
+                tags = soup.find_all(id=id)
+                for tag in tags:
+                    tag.decompose()
+
+            classes_to_exclude = [
+                "elementor-location-header",
+                "navbar-header",
+                "nav",
+                "header-sidebar-wrapper",
+                "blog-sidebar-wrapper",
+                "related-posts",
+            ]
+
+            for class_name in classes_to_exclude:
+                tags = soup.find_all(class_=class_name)
+                for tag in tags:
+                    tag.decompose()
+
+            content = soup.get_text()
+            content = clean_string(content)
+
+            cleaned_size = len(content)
+            if original_size != 0:
+                logging.info(
+                    f"[{id}] Cleaned page size: {cleaned_size} characters, down from {original_size} (shrunk: {original_size-cleaned_size} chars, {round((1-(cleaned_size/original_size)) * 100, 2)}%)"  # noqa:E501
+                )
+
+            result = f"""
+            email from '{meta_data.get('from')}' to '{meta_data.get('to')}'
+            subject: {meta_data.get('subject')}
+            date: {meta_data.get('date')}
+            preview: {snippet}
+            content: f{content}
+            """
+            data_content = dedent(result)
+            data.append({"content": data_content, "meta_data": meta_data})
+            data_contents.append(data_content)
+        doc_id = hashlib.sha256((query + ", ".join(data_contents)).encode()).hexdigest()
+        response_data = {"doc_id": doc_id, "data": data}
+        return response_data

embedchain/models/data_type.py

@@ -28,6 +28,7 @@ class IndirectDataType(Enum):
     UNSTRUCTURED = "unstructured"
     JSON = "json"
     OPENAPI = "openapi"
+    GMAIL = "gmail"
 
 
 class SpecialDataType(Enum):
@@ -55,3 +56,4 @@ class DataType(Enum):
     UNSTRUCTURED = IndirectDataType.UNSTRUCTURED.value
     JSON = IndirectDataType.JSON.value
     OPENAPI = IndirectDataType.OPENAPI.value
+    GMAIL = IndirectDataType.GMAIL.value

embedchain/utils.py

@@ -259,6 +259,8 @@ def detect_datatype(source: Any) -> DataType:
     else:
         # Source is not a URL.
 
+        # TODO: check if source is gmail query
+
         # Use text as final fallback.
         logging.debug(f"Source of `{formatted_source}` detected as `text`.")
         return DataType.TEXT