chore: Bump version

Signed-off-by: -LAN- <laipz8200@outlook.com>

.markdownlint.json

@@ -1,3 +1,4 @@
 {
-    "MD024": false
+    "MD024": false,
+    "MD013": false
 }

CHANGELOG.md

@@ -5,6 +5,19 @@ All notable changes to Dify will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.15.8] - 2025-05-30
+
+### Added
+
+- Added gunicorn keepalive setting (#19537)
+
+### Fixed
+
+- Fixed database configuration to allow DB_EXTRAS to set search_path via options (#16a4f77)
+- Fixed frontend third-party package security issues (#19655)
+- Updated dependencies: huggingface-hub (~0.16.4 to ~0.31.0), transformers (~4.35.0 to ~4.39.0), and resend (~0.7.0 to ~2.9.0) (#19563)
+- Downgrade boto3 from 1.36 to 1.35 (#19736)
+
 ## [0.15.7] - 2025-04-27
 
 ### Added

api/configs/middleware/__init__.py

@@ -1,5 +1,5 @@
 from typing import Any, Literal, Optional
-from urllib.parse import quote_plus
+from urllib.parse import parse_qsl, quote_plus
 
 from pydantic import Field, NonNegativeInt, PositiveFloat, PositiveInt, computed_field
 from pydantic_settings import BaseSettings
@@ -166,14 +166,28 @@ class DatabaseConfig(BaseSettings):
         default=False,
     )
 
-    @computed_field
+    @computed_field  # type: ignore[misc]
+    @property
     def SQLALCHEMY_ENGINE_OPTIONS(self) -> dict[str, Any]:
+        # Parse DB_EXTRAS for 'options'
+        db_extras_dict = dict(parse_qsl(self.DB_EXTRAS))
+        options = db_extras_dict.get("options", "")
+        # Always include timezone
+        timezone_opt = "-c timezone=UTC"
+        if options:
+            # Merge user options and timezone
+            merged_options = f"{options} {timezone_opt}"
+        else:
+            merged_options = timezone_opt
+
+        connect_args = {"options": merged_options}
+
         return {
             "pool_size": self.SQLALCHEMY_POOL_SIZE,
             "max_overflow": self.SQLALCHEMY_MAX_OVERFLOW,
             "pool_recycle": self.SQLALCHEMY_POOL_RECYCLE,
             "pool_pre_ping": self.SQLALCHEMY_POOL_PRE_PING,
-            "connect_args": {"options": "-c timezone=UTC"},
+            "connect_args": connect_args,
         }
 
 

api/configs/packaging/__init__.py

@@ -9,7 +9,7 @@ class PackagingInfo(BaseSettings):
 
     CURRENT_VERSION: str = Field(
         description="Dify version",
-        default="0.15.7",
+        default="0.15.8",
     )
 
     COMMIT_SHA: str = Field(

api/core/model_runtime/model_providers/openai/llm/llm.py

@@ -1049,6 +1049,9 @@ class OpenAILargeLanguageModel(_CommonOpenAI, LargeLanguageModel):
         """Calculate num tokens for gpt-3.5-turbo and gpt-4 with tiktoken package.
 
         Official documentation: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb"""
+        if not messages and not tools:
+            return 0
+
         if model.startswith("ft:"):
             model = model.split(":")[1]
 
@@ -1058,17 +1061,17 @@ class OpenAILargeLanguageModel(_CommonOpenAI, LargeLanguageModel):
 
         try:
             encoding = tiktoken.get_encoding(model)
-        except KeyError:
+        except (KeyError, ValueError) as e:
             logger.warning("Warning: model not found. Using cl100k_base encoding.")
-            model = "cl100k_base"
-            encoding = tiktoken.get_encoding(model)
+            encoding_name = "cl100k_base"
+            encoding = tiktoken.get_encoding(encoding_name)
 
         if model.startswith("gpt-3.5-turbo-0301"):
             # every message follows <im_start>{role/name}\n{content}<im_end>\n
             tokens_per_message = 4
             # if there's a name, the role is omitted
             tokens_per_name = -1
-        elif model.startswith("gpt-3.5-turbo") or model.startswith("gpt-4") or model.startswith(("o1", "o3")):
+        elif model.startswith("gpt-3.5-turbo") or model.startswith("gpt-4") or model.startswith(("o1", "o3", "o4")):
             tokens_per_message = 3
             tokens_per_name = 1
         else:

api/core/tools/utils/web_reader_tool.py

@@ -1,21 +1,13 @@
-import hashlib
-import json
 import mimetypes
-import os
 import re
-import site
-import subprocess
-import tempfile
-import unicodedata
-from contextlib import contextmanager
-from pathlib import Path
-from typing import Any, Literal, Optional, cast
+from collections.abc import Sequence
+from dataclasses import dataclass
+from typing import Any, Optional, cast
 from urllib.parse import unquote
 
 import chardet
 import cloudscraper  # type: ignore
-from bs4 import BeautifulSoup, CData, Comment, NavigableString  # type: ignore
-from regex import regex  # type: ignore
+from readabilipy import simple_json_from_html_string  # type: ignore
 
 from core.helper import ssrf_proxy
 from core.rag.extractor import extract_processor
@@ -23,9 +15,7 @@ from core.rag.extractor.extract_processor import ExtractProcessor
 
 FULL_TEMPLATE = """
 TITLE: {title}
-AUTHORS: {authors}
-PUBLISH DATE: {publish_date}
-TOP_IMAGE_URL: {top_image}
+AUTHOR: {author}
 TEXT:
 
 {text}
@@ -73,8 +63,8 @@ def get_url(url: str, user_agent: Optional[str] = None) -> str:
         response = ssrf_proxy.get(url, headers=headers, follow_redirects=True, timeout=(120, 300))
     elif response.status_code == 403:
         scraper = cloudscraper.create_scraper()
-        scraper.perform_request = ssrf_proxy.make_request
-        response = scraper.get(url, headers=headers, follow_redirects=True, timeout=(120, 300))
+        scraper.perform_request = ssrf_proxy.make_request  # type: ignore
+        response = scraper.get(url, headers=headers, follow_redirects=True, timeout=(120, 300))  # type: ignore
 
     if response.status_code != 200:
         return "URL returned status code {}.".format(response.status_code)
@@ -90,273 +80,36 @@ def get_url(url: str, user_agent: Optional[str] = None) -> str:
     else:
         content = response.text
 
-    a = extract_using_readabilipy(content)
+    article = extract_using_readabilipy(content)
 
-    if not a["plain_text"] or not a["plain_text"].strip():
+    if not article.text:
         return ""
 
     res = FULL_TEMPLATE.format(
-        title=a["title"],
-        authors=a["byline"],
-        publish_date=a["date"],
-        top_image="",
-        text=a["plain_text"] or "",
+        title=article.title,
+        author=article.auther,
+        text=article.text,
     )
 
     return res
 
 
-def extract_using_readabilipy(html):
-    with tempfile.NamedTemporaryFile(delete=False, mode="w+") as f_html:
-        f_html.write(html)
-        f_html.close()
-    html_path = f_html.name
-
-    # Call Mozilla's Readability.js Readability.parse() function via node, writing output to a temporary file
-    article_json_path = html_path + ".json"
-    jsdir = os.path.join(find_module_path("readabilipy"), "javascript")
-    with chdir(jsdir):
-        subprocess.check_call(["node", "ExtractArticle.js", "-i", html_path, "-o", article_json_path])
-
-    # Read output of call to Readability.parse() from JSON file and return as Python dictionary
-    input_json = json.loads(Path(article_json_path).read_text(encoding="utf-8"))
-
-    # Deleting files after processing
-    os.unlink(article_json_path)
-    os.unlink(html_path)
-
-    article_json: dict[str, Any] = {
-        "title": None,
-        "byline": None,
-        "date": None,
-        "content": None,
-        "plain_content": None,
-        "plain_text": None,
-    }
-    # Populate article fields from readability fields where present
-    if input_json:
-        if input_json.get("title"):
-            article_json["title"] = input_json["title"]
-        if input_json.get("byline"):
-            article_json["byline"] = input_json["byline"]
-        if input_json.get("date"):
-            article_json["date"] = input_json["date"]
-        if input_json.get("content"):
-            article_json["content"] = input_json["content"]
-            article_json["plain_content"] = plain_content(article_json["content"], False, False)
-            article_json["plain_text"] = extract_text_blocks_as_plain_text(article_json["plain_content"])
-        if input_json.get("textContent"):
-            article_json["plain_text"] = input_json["textContent"]
-            article_json["plain_text"] = re.sub(r"\n\s*\n", "\n", article_json["plain_text"])
-
-    return article_json
+@dataclass
+class Article:
+    title: str
+    auther: str
+    text: Sequence[dict]
 
 
-def find_module_path(module_name):
-    for package_path in site.getsitepackages():
-        potential_path = os.path.join(package_path, module_name)
-        if os.path.exists(potential_path):
-            return potential_path
-
-    return None
-
-
-@contextmanager
-def chdir(path):
-    """Change directory in context and return to original on exit"""
-    # From https://stackoverflow.com/a/37996581, couldn't find a built-in
-    original_path = os.getcwd()
-    os.chdir(path)
-    try:
-        yield
-    finally:
-        os.chdir(original_path)
-
-
-def extract_text_blocks_as_plain_text(paragraph_html):
-    # Load article as DOM
-    soup = BeautifulSoup(paragraph_html, "html.parser")
-    # Select all lists
-    list_elements = soup.find_all(["ul", "ol"])
-    # Prefix text in all list items with "* " and make lists paragraphs
-    for list_element in list_elements:
-        plain_items = "".join(
-            list(filter(None, [plain_text_leaf_node(li)["text"] for li in list_element.find_all("li")]))
-        )
-        list_element.string = plain_items
-        list_element.name = "p"
-    # Select all text blocks
-    text_blocks = [s.parent for s in soup.find_all(string=True)]
-    text_blocks = [plain_text_leaf_node(block) for block in text_blocks]
-    # Drop empty paragraphs
-    text_blocks = list(filter(lambda p: p["text"] is not None, text_blocks))
-    return text_blocks
-
-
-def plain_text_leaf_node(element):
-    # Extract all text, stripped of any child HTML elements and normalize it
-    plain_text = normalize_text(element.get_text())
-    if plain_text != "" and element.name == "li":
-        plain_text = "* {}, ".format(plain_text)
-    if plain_text == "":
-        plain_text = None
-    if "data-node-index" in element.attrs:
-        plain = {"node_index": element["data-node-index"], "text": plain_text}
-    else:
-        plain = {"text": plain_text}
-    return plain
-
-
-def plain_content(readability_content, content_digests, node_indexes):
-    # Load article as DOM
-    soup = BeautifulSoup(readability_content, "html.parser")
-    # Make all elements plain
-    elements = plain_elements(soup.contents, content_digests, node_indexes)
-    if node_indexes:
-        # Add node index attributes to nodes
-        elements = [add_node_indexes(element) for element in elements]
-    # Replace article contents with plain elements
-    soup.contents = elements
-    return str(soup)
-
-
-def plain_elements(elements, content_digests, node_indexes):
-    # Get plain content versions of all elements
-    elements = [plain_element(element, content_digests, node_indexes) for element in elements]
-    if content_digests:
-        # Add content digest attribute to nodes
-        elements = [add_content_digest(element) for element in elements]
-    return elements
-
-
-def plain_element(element, content_digests, node_indexes):
-    # For lists, we make each item plain text
-    if is_leaf(element):
-        # For leaf node elements, extract the text content, discarding any HTML tags
-        # 1. Get element contents as text
-        plain_text = element.get_text()
-        # 2. Normalize the extracted text string to a canonical representation
-        plain_text = normalize_text(plain_text)
-        # 3. Update element content to be plain text
-        element.string = plain_text
-    elif is_text(element):
-        if is_non_printing(element):
-            # The simplified HTML may have come from Readability.js so might
-            # have non-printing text (e.g. Comment or CData). In this case, we
-            # keep the structure, but ensure that the string is empty.
-            element = type(element)("")
-        else:
-            plain_text = element.string
-            plain_text = normalize_text(plain_text)
-            element = type(element)(plain_text)
-    else:
-        # If not a leaf node or leaf type call recursively on child nodes, replacing
-        element.contents = plain_elements(element.contents, content_digests, node_indexes)
-    return element
-
-
-def add_node_indexes(element, node_index="0"):
-    # Can't add attributes to string types
-    if is_text(element):
-        return element
-    # Add index to current element
-    element["data-node-index"] = node_index
-    # Add index to child elements
-    for local_idx, child in enumerate([c for c in element.contents if not is_text(c)], start=1):
-        # Can't add attributes to leaf string types
-        child_index = "{stem}.{local}".format(stem=node_index, local=local_idx)
-        add_node_indexes(child, node_index=child_index)
-    return element
-
-
-def normalize_text(text):
-    """Normalize unicode and whitespace."""
-    # Normalize unicode first to try and standardize whitespace characters as much as possible before normalizing them
-    text = strip_control_characters(text)
-    text = normalize_unicode(text)
-    text = normalize_whitespace(text)
-    return text
-
-
-def strip_control_characters(text):
-    """Strip out unicode control characters which might break the parsing."""
-    # Unicode control characters
-    #   [Cc]: Other, Control [includes new lines]
-    #   [Cf]: Other, Format
-    #   [Cn]: Other, Not Assigned
-    #   [Co]: Other, Private Use
-    #   [Cs]: Other, Surrogate
-    control_chars = {"Cc", "Cf", "Cn", "Co", "Cs"}
-    retained_chars = ["\t", "\n", "\r", "\f"]
-
-    # Remove non-printing control characters
-    return "".join(
-        [
-            "" if (unicodedata.category(char) in control_chars) and (char not in retained_chars) else char
-            for char in text
-        ]
+def extract_using_readabilipy(html: str):
+    json_article: dict[str, Any] = simple_json_from_html_string(html, use_readability=True)
+    article = Article(
+        title=json_article.get("title") or "",
+        auther=json_article.get("byline") or "",
+        text=json_article.get("plain_text") or [],
     )
 
-
-def normalize_unicode(text):
-    """Normalize unicode such that things that are visually equivalent map to the same unicode string where possible."""
-    normal_form: Literal["NFC", "NFD", "NFKC", "NFKD"] = "NFKC"
-    text = unicodedata.normalize(normal_form, text)
-    return text
-
-
-def normalize_whitespace(text):
-    """Replace runs of whitespace characters with a single space as this is what happens when HTML text is displayed."""
-    text = regex.sub(r"\s+", " ", text)
-    # Remove leading and trailing whitespace
-    text = text.strip()
-    return text
-
-
-def is_leaf(element):
-    return element.name in {"p", "li"}
-
-
-def is_text(element):
-    return isinstance(element, NavigableString)
-
-
-def is_non_printing(element):
-    return any(isinstance(element, _e) for _e in [Comment, CData])
-
-
-def add_content_digest(element):
-    if not is_text(element):
-        element["data-content-digest"] = content_digest(element)
-    return element
-
-
-def content_digest(element):
-    digest: Any
-    if is_text(element):
-        # Hash
-        trimmed_string = element.string.strip()
-        if trimmed_string == "":
-            digest = ""
-        else:
-            digest = hashlib.sha256(trimmed_string.encode("utf-8")).hexdigest()
-    else:
-        contents = element.contents
-        num_contents = len(contents)
-        if num_contents == 0:
-            # No hash when no child elements exist
-            digest = ""
-        elif num_contents == 1:
-            # If single child, use digest of child
-            digest = content_digest(contents[0])
-        else:
-            # Build content digest from the "non-empty" digests of child nodes
-            digest = hashlib.sha256()
-            child_digests = list(filter(lambda x: x != "", [content_digest(content) for content in contents]))
-            for child in child_digests:
-                digest.update(child.encode("utf-8"))
-            digest = digest.hexdigest()
-    return digest
+    return article
 
 
 def get_image_upload_file_ids(content):

api/docker/entrypoint.sh

@@ -35,6 +35,7 @@ else
       --worker-class ${SERVER_WORKER_CLASS:-gevent} \
       --worker-connections ${SERVER_WORKER_CONNECTIONS:-10} \
       --timeout ${GUNICORN_TIMEOUT:-200} \
+      --keep-alive ${GUNICORN_KEEP_ALIVE:-2} \
       app:app
   fi
 fi

api/pyproject.toml

@@ -21,7 +21,7 @@ azure-ai-inference = "~1.0.0b8"
 azure-ai-ml = "~1.20.0"
 azure-identity = "1.16.1"
 beautifulsoup4 = "4.12.2"
-boto3 = "1.36.12"
+boto3 = "~1.35.0"
 bs4 = "~0.0.1"
 cachetools = "~5.3.0"
 celery = "~5.4.0"
@@ -48,7 +48,7 @@ google-generativeai = "0.8.1"
 googleapis-common-protos = "1.63.0"
 gunicorn = "~23.0.0"
 httpx = { version = "~0.27.0", extras = ["socks"] }
-huggingface-hub = "~0.16.4"
+huggingface-hub = "~0.31.0"
 jieba = "0.42.1"
 langfuse = "~2.51.3"
 langsmith = "~0.1.77"
@@ -78,7 +78,7 @@ pyyaml = "~6.0.1"
 readabilipy = "0.2.0"
 redis = { version = "~5.0.3", extras = ["hiredis"] }
 replicate = "~0.22.0"
-resend = "~0.7.0"
+resend = "~2.9.0"
 sagemaker = "~2.231.0"
 scikit-learn = "~1.5.1"
 sentry-sdk = { version = "~1.44.1", extras = ["flask"] }
@@ -87,9 +87,9 @@ starlette = "0.41.0"
 tencentcloud-sdk-python-hunyuan = "~3.0.1294"
 tiktoken = "^0.9.0"
 tokenizers = "~0.15.0"
-transformers = "~4.35.0"
+transformers = "~4.39.0"
 unstructured = { version = "~0.16.1", extras = ["docx", "epub", "md", "msg", "ppt", "pptx"] }
-validators = "0.21.0"
+validators = "0.22.0"
 volcengine-python-sdk = {extras = ["ark"], version = "~1.0.98"}
 websocket-client = "~1.7.0"
 xinference-client = "0.15.2"
@@ -112,7 +112,7 @@ safetensors = "~0.4.3"
 # [ Tools ] dependency group
 ############################################################
 [tool.poetry.group.tools.dependencies]
-arxiv = "2.1.0"
+arxiv = "2.2.0"
 cloudscraper = "1.2.71"
 duckduckgo-search = "~6.3.0"
 jsonpath-ng = "1.6.1"
@@ -166,7 +166,7 @@ tcvectordb = "1.3.2"
 tidb-vector = "0.0.9"
 upstash-vector = "0.6.0"
 volcengine-compat = "~1.0.156"
-weaviate-client = "~3.21.0"
+weaviate-client = "~3.26.0"
 
 ############################################################
 # [ Dev ] dependency group

docker/.env.example

@@ -142,6 +142,9 @@ CELERY_WORKER_CLASS=
 # it is recommended to set it to 360 to support a longer sse connection time.
 GUNICORN_TIMEOUT=360
 
+# The number of seconds to wait for requests on a Keep-Alive connection, default to 2
+GUNICORN_KEEP_ALIVE=2
+
 # The number of Celery workers. The default is 1, and can be set as needed.
 CELERY_WORKER_AMOUNT=
 

docker/docker-compose-template.yaml

@@ -2,7 +2,7 @@ x-shared-env: &shared-api-worker-env
 services:
   # API service
   api:
-    image: langgenius/dify-api:0.15.7
+    image: langgenius/dify-api:0.15.8
     restart: always
     environment:
       # Use the shared environment variables.
@@ -25,7 +25,7 @@ services:
   # worker service
   # The Celery worker for processing the queue.
   worker:
-    image: langgenius/dify-api:0.15.7
+    image: langgenius/dify-api:0.15.8
     restart: always
     environment:
       # Use the shared environment variables.
@@ -47,7 +47,7 @@ services:
 
   # Frontend web application.
   web:
-    image: langgenius/dify-web:0.15.7
+    image: langgenius/dify-web:0.15.8
     restart: always
     environment:
       CONSOLE_API_URL: ${CONSOLE_API_URL:-}

docker/docker-compose.yaml

@@ -37,6 +37,7 @@ x-shared-env: &shared-api-worker-env
   SERVER_WORKER_CONNECTIONS: ${SERVER_WORKER_CONNECTIONS:-10}
   CELERY_WORKER_CLASS: ${CELERY_WORKER_CLASS:-}
   GUNICORN_TIMEOUT: ${GUNICORN_TIMEOUT:-360}
+  GUNICORN_KEEP_ALIVE: ${GUNICORN_KEEP_ALIVE:-2}
   CELERY_WORKER_AMOUNT: ${CELERY_WORKER_AMOUNT:-}
   CELERY_AUTO_SCALE: ${CELERY_AUTO_SCALE:-false}
   CELERY_MAX_WORKERS: ${CELERY_MAX_WORKERS:-}
@@ -394,7 +395,7 @@ x-shared-env: &shared-api-worker-env
 services:
   # API service
   api:
-    image: langgenius/dify-api:0.15.7
+    image: langgenius/dify-api:0.15.8
     restart: always
     environment:
       # Use the shared environment variables.
@@ -417,7 +418,7 @@ services:
   # worker service
   # The Celery worker for processing the queue.
   worker:
-    image: langgenius/dify-api:0.15.7
+    image: langgenius/dify-api:0.15.8
     restart: always
     environment:
       # Use the shared environment variables.
@@ -439,7 +440,7 @@ services:
 
   # Frontend web application.
   web:
-    image: langgenius/dify-web:0.15.7
+    image: langgenius/dify-web:0.15.8
     restart: always
     environment:
       CONSOLE_API_URL: ${CONSOLE_API_URL:-}

web/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dify-web",
-  "version": "0.15.7",
+  "version": "0.15.8",
   "private": true,
   "engines": {
     "node": ">=18.17.0"
@@ -161,7 +161,7 @@
     "jest-environment-jsdom": "^29.7.0",
     "lint-staged": "^13.2.2",
     "magicast": "^0.3.4",
-    "postcss": "^8.4.31",
+    "postcss": "^8.4.47",
     "sass": "^1.61.0",
     "storybook": "^8.3.5",
     "tailwindcss": "^3.4.4",
@@ -172,7 +172,10 @@
   "resolutions": {
     "@types/react": "~18.2.0",
     "@types/react-dom": "~18.2.0",
-    "string-width": "4.2.3"
+    "string-width": "4.2.3",
+    "nanoid": "~3.3.8",
+    "esbuild": "~0.25.0",
+    "serialize-javascript": "~6.0.2"
   },
   "lint-staged": {
     "**/*.js?(x)": [