Emil_Shanaty/open-webui-ai4me
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'dev' into main

Browse Source
This commit is contained in:
JT
2025-02-05 15:15:24 -08:00
committed by GitHub
parent 157c781b0a b8fe9497fb
commit 40dea3fbe1
148 changed files with 4380 additions and 1767 deletions
Download Patch File Download Diff File Expand all files Collapse all files
backend/open_webui/config.py
+70 -12
View File
@@ -927,6 +927,12 @@ USER_PERMISSIONS_FEATURES_IMAGE_GENERATION = (
== "true"
)
USER_PERMISSIONS_FEATURES_CODE_INTERPRETER = (
os.environ.get("USER_PERMISSIONS_FEATURES_CODE_INTERPRETER", "True").lower()
== "true"
)
DEFAULT_USER_PERMISSIONS = {
"workspace": {
"models": USER_PERMISSIONS_WORKSPACE_MODELS_ACCESS,
@@ -944,6 +950,7 @@ DEFAULT_USER_PERMISSIONS = {
"features": {
"web_search": USER_PERMISSIONS_FEATURES_WEB_SEARCH,
"image_generation": USER_PERMISSIONS_FEATURES_IMAGE_GENERATION,
"code_interpreter": USER_PERMISSIONS_FEATURES_CODE_INTERPRETER,
},
}
@@ -1094,21 +1101,27 @@ TITLE_GENERATION_PROMPT_TEMPLATE = PersistentConfig(
os.environ.get("TITLE_GENERATION_PROMPT_TEMPLATE", ""),
)
DEFAULT_TITLE_GENERATION_PROMPT_TEMPLATE = """Create a concise, 3-5 word title with an emoji as a title for the chat history, in the given language. Suitable Emojis for the summary can be used to enhance understanding but avoid quotation marks or special formatting. RESPOND ONLY WITH THE TITLE TEXT.
Examples of titles:
📉 Stock Market Trends
🍪 Perfect Chocolate Chip Recipe
Evolution of Music Streaming
Remote Work Productivity Tips
Artificial Intelligence in Healthcare
🎮 Video Game Development Insights
DEFAULT_TITLE_GENERATION_PROMPT_TEMPLATE = """### Task:
Generate a concise, 3-5 word title with an emoji summarizing the chat history.
### Guidelines:
- The title should clearly represent the main theme or subject of the conversation.
- Use emojis that enhance understanding of the topic, but avoid quotation marks or special formatting.
- Write the title in the chat's primary language; default to English if multilingual.
- Prioritize accuracy over excessive creativity; keep it clear and simple.
### Output:
JSON format: { "title": "your concise title here" }
### Examples:
- { "title": "📉 Stock Market Trends" },
- { "title": "🍪 Perfect Chocolate Chip Recipe" },
- { "title": "Evolution of Music Streaming" },
- { "title": "Remote Work Productivity Tips" },
- { "title": "Artificial Intelligence in Healthcare" },
- { "title": "🎮 Video Game Development Insights" }
### Chat History:
<chat_history>
{{MESSAGES:END:2}}
</chat_history>"""
TAGS_GENERATION_PROMPT_TEMPLATE = PersistentConfig(
"TAGS_GENERATION_PROMPT_TEMPLATE",
"task.tags.prompt_template",
@@ -1277,7 +1290,28 @@ TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = PersistentConfig(
)
DEFAULT_TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = """Available Tools: {{TOOLS}}\nReturn an empty string if no tools match the query. If a function tool matches, construct and return a JSON object in the format {\"name\": \"functionName\", \"parameters\": {\"requiredFunctionParamKey\": \"requiredFunctionParamValue\"}} using the appropriate tool and its parameters. Only return the object and limit the response to the JSON object without additional text."""
DEFAULT_TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = """Available Tools: {{TOOLS}}
Your task is to choose and return the correct tool(s) from the list of available tools based on the query. Follow these guidelines:
- Return only the JSON object, without any additional text or explanation.
- If no tools match the query, return an empty array:
{
"tool_calls": []
}
- If one or more tools match the query, construct a JSON response containing a "tool_calls" array with objects that include:
- "name": The tool's name.
- "parameters": A dictionary of required parameters and their corresponding values.
The format for the JSON response is strictly:
{
"tool_calls": [
{"name": "toolName1", "parameters": {"key1": "value1"}},
{"name": "toolName2", "parameters": {"key2": "value2"}}
]
}"""
DEFAULT_EMOJI_GENERATION_PROMPT_TEMPLATE = """Your task is to reflect the speaker's likely facial expression through a fitting emoji. Interpret emotions from the message and reflect their facial expression using fitting, diverse emojis (e.g., 😊, 😢, 😡, 😱).
@@ -1290,6 +1324,24 @@ Your task is to synthesize these responses into a single, high-quality response.
Responses from models: {{responses}}"""
DEFAULT_CODE_INTERPRETER_PROMPT = """
#### Tools Available
1. **Code Interpreter**: `<code_interpreter type="code" lang="python"></code_interpreter>`
- You have access to a Python shell that runs directly in the user's browser, enabling fast execution of code for analysis, calculations, or problem-solving. Use it in this response.
- The Python code you write can incorporate a wide array of libraries, handle data manipulation or visualization, perform API calls for web-related tasks, or tackle virtually any computational challenge. Use this flexibility to **think outside the box, craft elegant solutions, and harness Python's full potential**.
- To use it, **you must enclose your code within `<code_interpreter type="code" lang="python">` XML tags** and stop right away. If you don't, the code won't execute. Do NOT use triple backticks.
- When coding, **always aim to print meaningful outputs** (e.g., results, tables, summaries, or visuals) to better interpret and verify the findings. Avoid relying on implicit outputs; prioritize explicit and clear print statements so the results are effectively communicated to the user.
- After obtaining the printed output, **always provide a concise analysis, interpretation, or next steps to help the user understand the findings or refine the outcome further.**
- If the results are unclear, unexpected, or require validation, refine the code and execute it again as needed. Always aim to deliver meaningful insights from the results, iterating if necessary.
- If a link is provided for an image, audio, or any file, include it in the response exactly as given to ensure the user has access to the original resource.
- All responses should be communicated in the chat's primary language, ensuring seamless understanding. If the chat is multilingual, default to English for clarity.
- **If a link to an image, audio, or any file is provided in markdown format, ALWAYS regurgitate explicitly display it as part of the response to ensure the user can access it easily, do NOT change the link.**
Ensure that the tools are effectively utilized to achieve the highest-quality analysis for the user."""
####################################
# Vector Database
####################################
@@ -1319,6 +1371,7 @@ CHROMA_HTTP_SSL = os.environ.get("CHROMA_HTTP_SSL", "false").lower() == "true"
MILVUS_URI = os.environ.get("MILVUS_URI", f"{DATA_DIR}/vector_db/milvus.db")
MILVUS_DB = os.environ.get("MILVUS_DB", "default")
MILVUS_TOKEN = os.environ.get("MILVUS_TOKEN", None)
# Qdrant
QDRANT_URI = os.environ.get("QDRANT_URI", None)
@@ -1711,6 +1764,11 @@ BING_SEARCH_V7_SUBSCRIPTION_KEY = PersistentConfig(
os.environ.get("BING_SEARCH_V7_SUBSCRIPTION_KEY", ""),
)
EXA_API_KEY = PersistentConfig(
"EXA_API_KEY",
"rag.web.search.exa_api_key",
os.getenv("EXA_API_KEY", ""),
)
RAG_WEB_SEARCH_RESULT_COUNT = PersistentConfig(
"RAG_WEB_SEARCH_RESULT_COUNT",
backend/open_webui/constants.py
+1 -1
View File
@@ -57,7 +57,7 @@ class ERROR_MESSAGES(str, Enum):
)
FILE_NOT_SENT = "FILE_NOT_SENT"
FILE_NOT_SUPPORTED = "Oops! It seems like the file format you're trying to upload is not supported. Please upload a file with a supported format (e.g., JPG, PNG, PDF, TXT) and try again."
FILE_NOT_SUPPORTED = "Oops! It seems like the file format you're trying to upload is not supported. Please upload a file with a supported format and try again."
NOT_FOUND = "We could not find what you're looking for :/"
USER_NOT_FOUND = "We could not find what you're looking for :/"
backend/open_webui/env.py
+14 -6
View File
@@ -356,14 +356,22 @@ WEBUI_SECRET_KEY = os.environ.get(
), # DEPRECATED: remove at next major version
)
WEBUI_SESSION_COOKIE_SAME_SITE = os.environ.get(
"WEBUI_SESSION_COOKIE_SAME_SITE",
os.environ.get("WEBUI_SESSION_COOKIE_SAME_SITE", "lax"),
WEBUI_SESSION_COOKIE_SAME_SITE = os.environ.get("WEBUI_SESSION_COOKIE_SAME_SITE", "lax")
WEBUI_SESSION_COOKIE_SECURE = (
os.environ.get("WEBUI_SESSION_COOKIE_SECURE", "false").lower() == "true"
)
WEBUI_SESSION_COOKIE_SECURE = os.environ.get(
"WEBUI_SESSION_COOKIE_SECURE",
os.environ.get("WEBUI_SESSION_COOKIE_SECURE", "false").lower() == "true",
WEBUI_AUTH_COOKIE_SAME_SITE = os.environ.get(
"WEBUI_AUTH_COOKIE_SAME_SITE", WEBUI_SESSION_COOKIE_SAME_SITE
)
WEBUI_AUTH_COOKIE_SECURE = (
os.environ.get(
"WEBUI_AUTH_COOKIE_SECURE",
os.environ.get("WEBUI_SESSION_COOKIE_SECURE", "false"),
).lower()
== "true"
)
if WEBUI_AUTH and WEBUI_SECRET_KEY == "":
backend/open_webui/functions.py
+1 -1
View File
@@ -250,7 +250,7 @@ async def generate_function_chat_completion(
params = model_info.params.model_dump()
form_data = apply_model_params_to_body_openai(params, form_data)
form_data = apply_model_system_prompt_to_body(params, form_data, user)
form_data = apply_model_system_prompt_to_body(params, form_data, metadata, user)
pipe_id = get_pipe_id(form_data)
function_module = get_function_module_by_id(request, pipe_id)
backend/open_webui/main.py
+23 -6
View File
@@ -177,6 +177,7 @@ from open_webui.config import (
BING_SEARCH_V7_ENDPOINT,
BING_SEARCH_V7_SUBSCRIPTION_KEY,
BRAVE_SEARCH_API_KEY,
EXA_API_KEY,
KAGI_SEARCH_API_KEY,
MOJEEK_SEARCH_API_KEY,
GOOGLE_PSE_API_KEY,
@@ -523,6 +524,7 @@ app.state.config.SEARCHAPI_ENGINE = SEARCHAPI_ENGINE
app.state.config.JINA_API_KEY = JINA_API_KEY
app.state.config.BING_SEARCH_V7_ENDPOINT = BING_SEARCH_V7_ENDPOINT
app.state.config.BING_SEARCH_V7_SUBSCRIPTION_KEY = BING_SEARCH_V7_SUBSCRIPTION_KEY
app.state.config.EXA_API_KEY = EXA_API_KEY
app.state.config.RAG_WEB_SEARCH_RESULT_COUNT = RAG_WEB_SEARCH_RESULT_COUNT
app.state.config.RAG_WEB_SEARCH_CONCURRENT_REQUESTS = RAG_WEB_SEARCH_CONCURRENT_REQUESTS
@@ -859,6 +861,7 @@ async def chat_completion(
if model_id not in request.app.state.MODELS:
raise Exception("Model not found")
model = request.app.state.MODELS[model_id]
model_info = Models.get_model_by_id(model_id)
# Check if user has access to the model
if not BYPASS_MODEL_ACCESS_CONTROL and user.role == "user":
@@ -875,12 +878,25 @@ async def chat_completion(
"tool_ids": form_data.get("tool_ids", None),
"files": form_data.get("files", None),
"features": form_data.get("features", None),
"variables": form_data.get("variables", None),
"model": model_info,
**(
{"function_calling": "native"}
if form_data.get("params", {}).get("function_calling") == "native"
or (
model_info
and model_info.params.model_dump().get("function_calling")
== "native"
)
else {}
),
}
form_data["metadata"] = metadata
form_data, events = await process_chat_payload(
form_data, metadata, events = await process_chat_payload(
request, form_data, metadata, user, model
)
except Exception as e:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
@@ -889,6 +905,7 @@ async def chat_completion(
try:
response = await chat_completion_handler(request, form_data, user)
return await process_chat_response(
request, response, form_data, user, events, metadata, tasks
)
@@ -1007,10 +1024,6 @@ async def get_app_config(request: Request):
else {}
),
},
"google_drive": {
"client_id": GOOGLE_DRIVE_CLIENT_ID.value,
"api_key": GOOGLE_DRIVE_API_KEY.value,
},
**(
{
"default_models": app.state.config.DEFAULT_MODELS,
@@ -1030,6 +1043,10 @@ async def get_app_config(request: Request):
"max_count": app.state.config.FILE_MAX_COUNT,
},
"permissions": {**app.state.config.USER_PERMISSIONS},
"google_drive": {
"client_id": GOOGLE_DRIVE_CLIENT_ID.value,
"api_key": GOOGLE_DRIVE_API_KEY.value,
},
}
if user is not None
else {}
@@ -1063,7 +1080,7 @@ async def get_app_version():
@app.get("/api/version/updates")
async def get_app_latest_release_version():
async def get_app_latest_release_version(user=Depends(get_verified_user)):
if OFFLINE_MODE:
log.debug(
f"Offline mode is enabled, returning current version as latest version"
backend/open_webui/retrieval/utils.py
+57 -19
View File
@@ -15,13 +15,20 @@ from langchain_core.documents import Document
from open_webui.config import VECTOR_DB
from open_webui.retrieval.vector.connector import VECTOR_DB_CLIENT
from open_webui.utils.misc import get_last_user_message
from open_webui.models.users import UserModel
from open_webui.env import SRC_LOG_LEVELS, OFFLINE_MODE
from open_webui.env import (
SRC_LOG_LEVELS,
OFFLINE_MODE,
ENABLE_FORWARD_USER_INFO_HEADERS,
)
from open_webui.config import (
RAG_EMBEDDING_QUERY_PREFIX,
RAG_EMBEDDING_PASSAGE_PREFIX,
RAG_EMBEDDING_PREFIX_FIELD_NAME
)
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
@@ -65,9 +72,7 @@ class VectorSearchRetriever(BaseRetriever):
def query_doc(
collection_name: str,
query_embedding: list[float],
k: int,
collection_name: str, query_embedding: list[float], k: int, user: UserModel = None
):
try:
result = VECTOR_DB_CLIENT.search(
@@ -263,27 +268,31 @@ def get_embedding_function(
embedding_batch_size,
):
if embedding_engine == "":
return lambda query, prefix: embedding_function.encode(query, prompt = prefix if prefix else None).tolist()
return lambda query, prefix, user=None: embedding_function.encode(query, prompt = prefix if prefix else None).tolist()
elif embedding_engine in ["ollama", "openai"]:
func = lambda query, prefix: generate_embeddings(
func = lambda query, prefix, user=None: generate_embeddings(
engine=embedding_engine,
model=embedding_model,
text=query,
prefix=prefix,
url=url,
key=key,
user=user,
)
def generate_multiple(query, prefix, func):
def generate_multiple(query, prefix, user, func):
if isinstance(query, list):
embeddings = []
for i in range(0, len(query), embedding_batch_size):
embeddings.extend(func(query[i : i + embedding_batch_size], prefix))
embeddings.extend(
func(query[i : i + embedding_batch_size], prefix=prefix, user=user)
)
return embeddings
else:
return func(query, prefix)
return func(query, prefix, user)
return lambda query, prefix, user=None: generate_multiple(query, prefix, user, func)
else:
raise ValueError(f"Unknown embedding engine: {embedding_engine}")
return lambda query, prefix: generate_multiple(query, prefix, func)
def get_sources_from_files(
@@ -428,9 +437,13 @@ def get_model_path(model: str, update_model: bool = False):
def generate_openai_batch_embeddings(
model: str, texts: list[str], url: str = "https://api.openai.com/v1", key: str = "", prefix: str = None
model: str,
texts: list[str],
url: str = "https://api.openai.com/v1",
key: str = "",
prefix: str = None,
user: UserModel = None
) -> Optional[list[list[float]]]:
try:
json_data = {
"input": texts,
@@ -444,6 +457,16 @@ def generate_openai_batch_embeddings(
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {key}",
**(
{
"X-OpenWebUI-User-Name": user.name,
"X-OpenWebUI-User-Id": user.id,
"X-OpenWebUI-User-Email": user.email,
"X-OpenWebUI-User-Role": user.role,
}
if ENABLE_FORWARD_USER_INFO_HEADERS and user
else {}
),
},
json=json_data,
)
@@ -459,7 +482,12 @@ def generate_openai_batch_embeddings(
def generate_ollama_batch_embeddings(
model: str, texts: list[str], url: str, key: str = "", prefix: str = None
model: str,
texts: list[str],
url: str,
key: str = "",
prefix: str = None,
user: UserModel = None
) -> Optional[list[list[float]]]:
try:
json_data = {
@@ -474,6 +502,16 @@ def generate_ollama_batch_embeddings(
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {key}",
**(
{
"X-OpenWebUI-User-Name": user.name,
"X-OpenWebUI-User-Id": user.id,
"X-OpenWebUI-User-Email": user.email,
"X-OpenWebUI-User-Role": user.role,
}
if ENABLE_FORWARD_USER_INFO_HEADERS
else {}
),
},
json=json_data,
)
@@ -492,6 +530,7 @@ def generate_ollama_batch_embeddings(
def generate_embeddings(engine: str, model: str, text: Union[str, list[str]], prefix: Union[str , None] = None, **kwargs):
url = kwargs.get("url", "")
key = kwargs.get("key", "")
user = kwargs.get("user")
if prefix is not None and RAG_EMBEDDING_PREFIX_FIELD_NAME is None:
if isinstance(text, list):
@@ -502,19 +541,18 @@ def generate_embeddings(engine: str, model: str, text: Union[str, list[str]], pr
if engine == "ollama":
if isinstance(text, list):
embeddings = generate_ollama_batch_embeddings(
**{"model": model, "texts": text, "url": url, "key": key, "prefix": prefix}
**{"model": model, "texts": text, "url": url, "key": key, "prefix": prefix, "user": user}
)
else:
embeddings = generate_ollama_batch_embeddings(
**{"model": model, "texts": [text], "url": url, "key": key, "prefix": prefix}
**{"model": model, "texts": [text], "url": url, "key": key, "prefix": prefix, "user": user}
)
return embeddings[0] if isinstance(text, str) else embeddings
elif engine == "openai":
if isinstance(text, list):
embeddings = generate_openai_batch_embeddings(model, text, url, key, prefix)
embeddings = generate_openai_batch_embeddings(model, text, url, key, prefix, user)
else:
embeddings = generate_openai_batch_embeddings(model, [text], url, key, prefix)
embeddings = generate_openai_batch_embeddings(model, [text], url, key, prefix, user)
return embeddings[0] if isinstance(text, str) else embeddings
backend/open_webui/retrieval/vector/dbs/milvus.py
+5 -1
View File
@@ -8,13 +8,17 @@ from open_webui.retrieval.vector.main import VectorItem, SearchResult, GetResult
from open_webui.config import (
MILVUS_URI,
MILVUS_DB,
MILVUS_TOKEN,
)
class MilvusClient:
def __init__(self):
self.collection_prefix = "open_webui"
self.client = Client(uri=MILVUS_URI, database=MILVUS_DB)
if MILVUS_TOKEN is None:
self.client = Client(uri=MILVUS_URI, database=MILVUS_DB)
else:
self.client = Client(uri=MILVUS_URI, database=MILVUS_DB, token=MILVUS_TOKEN)
def _result_to_get_result(self, result) -> GetResult:
ids = []
backend/open_webui/retrieval/web/exa.py
+76
View File
@@ -0,0 +1,76 @@
import logging
from dataclasses import dataclass
from typing import Optional
import requests
from open_webui.env import SRC_LOG_LEVELS
from open_webui.retrieval.web.main import SearchResult
log = logging.getLogger(__name__)
log.setLevel(SRC_LOG_LEVELS["RAG"])
EXA_API_BASE = "https://api.exa.ai"
@dataclass
class ExaResult:
url: str
title: str
text: str
def search_exa(
api_key: str,
query: str,
count: int,
filter_list: Optional[list[str]] = None,
) -> list[SearchResult]:
"""Search using Exa Search API and return the results as a list of SearchResult objects.
Args:
api_key (str): A Exa Search API key
query (str): The query to search for
count (int): Number of results to return
filter_list (Optional[list[str]]): List of domains to filter results by
"""
log.info(f"Searching with Exa for query: {query}")
headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
payload = {
"query": query,
"numResults": count or 5,
"includeDomains": filter_list,
"contents": {"text": True, "highlights": True},
"type": "auto", # Use the auto search type (keyword or neural)
}
try:
response = requests.post(
f"{EXA_API_BASE}/search", headers=headers, json=payload
)
response.raise_for_status()
data = response.json()
results = []
for result in data["results"]:
results.append(
ExaResult(
url=result["url"],
title=result["title"],
text=result["text"],
)
)
log.info(f"Found {len(results)} results")
return [
SearchResult(
link=result.url,
title=result.title,
snippet=result.text,
)
for result in results
]
except Exception as e:
log.error(f"Error searching Exa: {e}")
return []
backend/open_webui/retrieval/web/main.py
+4
View File
@@ -1,3 +1,5 @@
import validators
from typing import Optional
from urllib.parse import urlparse
@@ -10,6 +12,8 @@ def get_filtered_results(results, filter_list):
filtered_results = []
for result in results:
url = result.get("url") or result.get("link", "")
if not validators.url(url):
continue
domain = urlparse(url).netloc
if any(domain.endswith(filtered_domain) for filtered_domain in filter_list):
filtered_results.append(result)
backend/open_webui/retrieval/web/utils.py
+15 -4
View File
@@ -43,6 +43,17 @@ def validate_url(url: Union[str, Sequence[str]]):
return False
def safe_validate_urls(url: Sequence[str]) -> Sequence[str]:
valid_urls = []
for u in url:
try:
if validate_url(u):
valid_urls.append(u)
except ValueError:
continue
return valid_urls
def resolve_hostname(hostname):
# Get address information
addr_info = socket.getaddrinfo(hostname, None)
@@ -86,11 +97,11 @@ def get_web_loader(
verify_ssl: bool = True,
requests_per_second: int = 2,
):
# Check if the URL is valid
if not validate_url(urls):
raise ValueError(ERROR_MESSAGES.INVALID_URL)
# Check if the URLs are valid
safe_urls = safe_validate_urls([urls] if isinstance(urls, str) else urls)
return SafeWebBaseLoader(
urls,
safe_urls,
verify_ssl=verify_ssl,
requests_per_second=requests_per_second,
continue_on_failure=True,
backend/open_webui/routers/auths.py
+14 -9
View File
@@ -25,8 +25,8 @@ from open_webui.env import (
WEBUI_AUTH,
WEBUI_AUTH_TRUSTED_EMAIL_HEADER,
WEBUI_AUTH_TRUSTED_NAME_HEADER,
WEBUI_SESSION_COOKIE_SAME_SITE,
WEBUI_SESSION_COOKIE_SECURE,
WEBUI_AUTH_COOKIE_SAME_SITE,
WEBUI_AUTH_COOKIE_SECURE,
SRC_LOG_LEVELS,
)
from fastapi import APIRouter, Depends, HTTPException, Request, status
@@ -95,8 +95,8 @@ async def get_session_user(
value=token,
expires=datetime_expires_at,
httponly=True, # Ensures the cookie is not accessible via JavaScript
samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
secure=WEBUI_SESSION_COOKIE_SECURE,
samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
secure=WEBUI_AUTH_COOKIE_SECURE,
)
user_permissions = get_permissions(
@@ -164,7 +164,7 @@ async def update_password(
############################
# LDAP Authentication
############################
@router.post("/ldap", response_model=SigninResponse)
@router.post("/ldap", response_model=SessionUserResponse)
async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
ENABLE_LDAP = request.app.state.config.ENABLE_LDAP
LDAP_SERVER_LABEL = request.app.state.config.LDAP_SERVER_LABEL
@@ -288,6 +288,10 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
httponly=True, # Ensures the cookie is not accessible via JavaScript
)
user_permissions = get_permissions(
user.id, request.app.state.config.USER_PERMISSIONS
)
return {
"token": token,
"token_type": "Bearer",
@@ -296,6 +300,7 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
"name": user.name,
"role": user.role,
"profile_image_url": user.profile_image_url,
"permissions": user_permissions,
}
else:
raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
@@ -378,8 +383,8 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
value=token,
expires=datetime_expires_at,
httponly=True, # Ensures the cookie is not accessible via JavaScript
samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
secure=WEBUI_SESSION_COOKIE_SECURE,
samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
secure=WEBUI_AUTH_COOKIE_SECURE,
)
user_permissions = get_permissions(
@@ -473,8 +478,8 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
value=token,
expires=datetime_expires_at,
httponly=True, # Ensures the cookie is not accessible via JavaScript
samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
secure=WEBUI_SESSION_COOKIE_SECURE,
samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
secure=WEBUI_AUTH_COOKIE_SECURE,
)
if request.app.state.config.WEBHOOK_URL:
backend/open_webui/routers/chats.py
+8 -2
View File
@@ -444,15 +444,21 @@ async def pin_chat_by_id(id: str, user=Depends(get_verified_user)):
############################
class CloneForm(BaseModel):
title: Optional[str] = None
@router.post("/{id}/clone", response_model=Optional[ChatResponse])
async def clone_chat_by_id(id: str, user=Depends(get_verified_user)):
async def clone_chat_by_id(
form_data: CloneForm, id: str, user=Depends(get_verified_user)
):
chat = Chats.get_chat_by_id_and_user_id(id, user.id)
if chat:
updated_chat = {
**chat.chat,
"originalChatId": chat.id,
"branchPointMessageId": chat.chat["history"]["currentId"],
"title": f"Clone of {chat.title}",
"title": form_data.title if form_data.title else f"Clone of {chat.title}",
}
chat = Chats.insert_new_chat(user.id, ChatForm(**{"chat": updated_chat}))
backend/open_webui/routers/files.py
+4 -2
View File
@@ -71,7 +71,7 @@ def upload_file(
)
try:
process_file(request, ProcessFileForm(file_id=id))
process_file(request, ProcessFileForm(file_id=id), user=user)
file_item = Files.get_file_by_id(id=id)
except Exception as e:
log.exception(e)
@@ -193,7 +193,9 @@ async def update_file_data_content_by_id(
if file and (file.user_id == user.id or user.role == "admin"):
try:
process_file(
request, ProcessFileForm(file_id=id, content=form_data.content)
request,
ProcessFileForm(file_id=id, content=form_data.content),
user=user,
)
file = Files.get_file_by_id(id=id)
except Exception as e:
backend/open_webui/routers/knowledge.py
+37 -12
View File
@@ -264,7 +264,11 @@ def add_file_to_knowledge_by_id(
detail=ERROR_MESSAGES.NOT_FOUND,
)
if knowledge.user_id != user.id and user.role != "admin":
if (
knowledge.user_id != user.id
and not has_access(user.id, "write", knowledge.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -285,7 +289,9 @@ def add_file_to_knowledge_by_id(
# Add content to the vector database
try:
process_file(
request, ProcessFileForm(file_id=form_data.file_id, collection_name=id)
request,
ProcessFileForm(file_id=form_data.file_id, collection_name=id),
user=user,
)
except Exception as e:
log.debug(e)
@@ -342,7 +348,12 @@ def update_file_from_knowledge_by_id(
detail=ERROR_MESSAGES.NOT_FOUND,
)
if knowledge.user_id != user.id and user.role != "admin":
if (
knowledge.user_id != user.id
and not has_access(user.id, "write", knowledge.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -363,7 +374,9 @@ def update_file_from_knowledge_by_id(
# Add content to the vector database
try:
process_file(
request, ProcessFileForm(file_id=form_data.file_id, collection_name=id)
request,
ProcessFileForm(file_id=form_data.file_id, collection_name=id),
user=user,
)
except Exception as e:
raise HTTPException(
@@ -406,7 +419,11 @@ def remove_file_from_knowledge_by_id(
detail=ERROR_MESSAGES.NOT_FOUND,
)
if knowledge.user_id != user.id and user.role != "admin":
if (
knowledge.user_id != user.id
and not has_access(user.id, "write", knowledge.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -429,10 +446,6 @@ def remove_file_from_knowledge_by_id(
if VECTOR_DB_CLIENT.has_collection(collection_name=file_collection):
VECTOR_DB_CLIENT.delete_collection(collection_name=file_collection)
# Delete physical file
if file.path:
Storage.delete_file(file.path)
# Delete file from database
Files.delete_file_by_id(form_data.file_id)
@@ -484,7 +497,11 @@ async def delete_knowledge_by_id(id: str, user=Depends(get_verified_user)):
detail=ERROR_MESSAGES.NOT_FOUND,
)
if knowledge.user_id != user.id and user.role != "admin":
if (
knowledge.user_id != user.id
and not has_access(user.id, "write", knowledge.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -543,7 +560,11 @@ async def reset_knowledge_by_id(id: str, user=Depends(get_verified_user)):
detail=ERROR_MESSAGES.NOT_FOUND,
)
if knowledge.user_id != user.id and user.role != "admin":
if (
knowledge.user_id != user.id
and not has_access(user.id, "write", knowledge.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -582,7 +603,11 @@ def add_files_to_knowledge_batch(
detail=ERROR_MESSAGES.NOT_FOUND,
)
if knowledge.user_id != user.id and user.role != "admin":
if (
knowledge.user_id != user.id
and not has_access(user.id, "write", knowledge.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
backend/open_webui/routers/memories.py
+6 -4
View File
@@ -57,7 +57,7 @@ async def add_memory(
{
"id": memory.id,
"text": memory.content,
"vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
"vector": request.app.state.EMBEDDING_FUNCTION(memory.content, user),
"metadata": {"created_at": memory.created_at},
}
],
@@ -82,7 +82,7 @@ async def query_memory(
):
results = VECTOR_DB_CLIENT.search(
collection_name=f"user-memory-{user.id}",
vectors=[request.app.state.EMBEDDING_FUNCTION(form_data.content)],
vectors=[request.app.state.EMBEDDING_FUNCTION(form_data.content, user)],
limit=form_data.k,
)
@@ -105,7 +105,7 @@ async def reset_memory_from_vector_db(
{
"id": memory.id,
"text": memory.content,
"vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
"vector": request.app.state.EMBEDDING_FUNCTION(memory.content, user),
"metadata": {
"created_at": memory.created_at,
"updated_at": memory.updated_at,
@@ -160,7 +160,9 @@ async def update_memory_by_id(
{
"id": memory.id,
"text": memory.content,
"vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
"vector": request.app.state.EMBEDDING_FUNCTION(
memory.content, user
),
"metadata": {
"created_at": memory.created_at,
"updated_at": memory.updated_at,
backend/open_webui/routers/models.py
+5 -1
View File
@@ -183,7 +183,11 @@ async def delete_model_by_id(id: str, user=Depends(get_verified_user)):
detail=ERROR_MESSAGES.NOT_FOUND,
)
if model.user_id != user.id and user.role != "admin":
if (
user.role != "admin"
and model.user_id != user.id
and not has_access(user.id, "write", model.access_control)
):
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail=ERROR_MESSAGES.UNAUTHORIZED,
backend/open_webui/routers/ollama.py
+8 -5
View File
@@ -395,7 +395,7 @@ async def get_ollama_tags(
)
if user.role == "user" and not BYPASS_MODEL_ACCESS_CONTROL:
models["models"] = get_filtered_models(models, user)
models["models"] = await get_filtered_models(models, user)
return models
@@ -939,6 +939,7 @@ async def generate_completion(
class ChatMessage(BaseModel):
role: str
content: str
tool_calls: Optional[list[dict]] = None
images: Optional[list[str]] = None
@@ -950,6 +951,7 @@ class GenerateChatCompletionForm(BaseModel):
template: Optional[str] = None
stream: Optional[bool] = True
keep_alive: Optional[Union[int, str]] = None
tools: Optional[list[dict]] = None
async def get_ollama_url(request: Request, model: str, url_idx: Optional[int] = None):
@@ -977,6 +979,7 @@ async def generate_chat_completion(
if BYPASS_MODEL_ACCESS_CONTROL:
bypass_filter = True
metadata = form_data.pop("metadata", None)
try:
form_data = GenerateChatCompletionForm(**form_data)
except Exception as e:
@@ -987,8 +990,6 @@ async def generate_chat_completion(
)
payload = {**form_data.model_dump(exclude_none=True)}
if "metadata" in payload:
del payload["metadata"]
model_id = payload["model"]
model_info = Models.get_model_by_id(model_id)
@@ -1006,7 +1007,7 @@ async def generate_chat_completion(
payload["options"] = apply_model_params_to_body_ollama(
params, payload["options"]
)
payload = apply_model_system_prompt_to_body(params, payload, user)
payload = apply_model_system_prompt_to_body(params, payload, metadata, user)
# Check if user has access to the model
if not bypass_filter and user.role == "user":
@@ -1159,6 +1160,8 @@ async def generate_openai_chat_completion(
url_idx: Optional[int] = None,
user=Depends(get_verified_user),
):
metadata = form_data.pop("metadata", None)
try:
completion_form = OpenAIChatCompletionForm(**form_data)
except Exception as e:
@@ -1185,7 +1188,7 @@ async def generate_openai_chat_completion(
if params:
payload = apply_model_params_to_body_openai(params, payload)
payload = apply_model_system_prompt_to_body(params, payload, user)
payload = apply_model_system_prompt_to_body(params, payload, metadata, user)
# Check if user has access to the model
if user.role == "user":
backend/open_webui/routers/openai.py
+4 -4
View File
@@ -489,7 +489,7 @@ async def get_models(
raise HTTPException(status_code=500, detail=error_detail)
if user.role == "user" and not BYPASS_MODEL_ACCESS_CONTROL:
models["data"] = get_filtered_models(models, user)
models["data"] = await get_filtered_models(models, user)
return models
@@ -551,9 +551,9 @@ async def generate_chat_completion(
bypass_filter = True
idx = 0
payload = {**form_data}
if "metadata" in payload:
del payload["metadata"]
metadata = payload.pop("metadata", None)
model_id = form_data.get("model")
model_info = Models.get_model_by_id(model_id)
@@ -566,7 +566,7 @@ async def generate_chat_completion(
params = model_info.params.model_dump()
payload = apply_model_params_to_body_openai(params, payload)
payload = apply_model_system_prompt_to_body(params, payload, user)
payload = apply_model_system_prompt_to_body(params, payload, metadata, user)
# Check if user has access to the model
if not bypass_filter and user.role == "user":
backend/open_webui/routers/prompts.py
+5 -1
View File
@@ -147,7 +147,11 @@ async def delete_prompt_by_command(command: str, user=Depends(get_verified_user)
detail=ERROR_MESSAGES.NOT_FOUND,
)
if prompt.user_id != user.id and user.role != "admin":
if (
prompt.user_id != user.id
and not has_access(user.id, "write", prompt.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
backend/open_webui/routers/retrieval.py
+43 -9
View File
@@ -55,6 +55,7 @@ from open_webui.retrieval.web.serply import search_serply
from open_webui.retrieval.web.serpstack import search_serpstack
from open_webui.retrieval.web.tavily import search_tavily
from open_webui.retrieval.web.bing import search_bing
from open_webui.retrieval.web.exa import search_exa
from open_webui.retrieval.utils import (
@@ -389,6 +390,7 @@ async def get_rag_config(request: Request, user=Depends(get_admin_user)):
"jina_api_key": request.app.state.config.JINA_API_KEY,
"bing_search_v7_endpoint": request.app.state.config.BING_SEARCH_V7_ENDPOINT,
"bing_search_v7_subscription_key": request.app.state.config.BING_SEARCH_V7_SUBSCRIPTION_KEY,
"exa_api_key": request.app.state.config.EXA_API_KEY,
"result_count": request.app.state.config.RAG_WEB_SEARCH_RESULT_COUNT,
"concurrent_requests": request.app.state.config.RAG_WEB_SEARCH_CONCURRENT_REQUESTS,
},
@@ -437,6 +439,7 @@ class WebSearchConfig(BaseModel):
jina_api_key: Optional[str] = None
bing_search_v7_endpoint: Optional[str] = None
bing_search_v7_subscription_key: Optional[str] = None
exa_api_key: Optional[str] = None
result_count: Optional[int] = None
concurrent_requests: Optional[int] = None
@@ -543,6 +546,8 @@ async def update_rag_config(
form_data.web.search.bing_search_v7_subscription_key
)
request.app.state.config.EXA_API_KEY = form_data.web.search.exa_api_key
request.app.state.config.RAG_WEB_SEARCH_RESULT_COUNT = (
form_data.web.search.result_count
)
@@ -592,6 +597,7 @@ async def update_rag_config(
"jina_api_key": request.app.state.config.JINA_API_KEY,
"bing_search_v7_endpoint": request.app.state.config.BING_SEARCH_V7_ENDPOINT,
"bing_search_v7_subscription_key": request.app.state.config.BING_SEARCH_V7_SUBSCRIPTION_KEY,
"exa_api_key": request.app.state.config.EXA_API_KEY,
"result_count": request.app.state.config.RAG_WEB_SEARCH_RESULT_COUNT,
"concurrent_requests": request.app.state.config.RAG_WEB_SEARCH_CONCURRENT_REQUESTS,
},
@@ -661,6 +667,7 @@ def save_docs_to_vector_db(
overwrite: bool = False,
split: bool = True,
add: bool = False,
user=None,
) -> bool:
def _get_docs_info(docs: list[Document]) -> str:
docs_info = set()
@@ -776,7 +783,7 @@ def save_docs_to_vector_db(
)
embeddings = embedding_function(
list(map(lambda x: x.replace("\n", " "), texts)), RAG_EMBEDDING_PASSAGE_PREFIX
list(map(lambda x: x.replace("\n", " "), texts)), prefix=RAG_EMBEDDING_PASSAGE_PREFIX, user=user
)
items = [
@@ -934,6 +941,7 @@ def process_file(
"hash": hash,
},
add=(True if form_data.collection_name else False),
user=user,
)
if result:
@@ -991,7 +999,7 @@ def process_text(
text_content = form_data.content
log.debug(f"text_content: {text_content}")
result = save_docs_to_vector_db(request, docs, collection_name)
result = save_docs_to_vector_db(request, docs, collection_name, user=user)
if result:
return {
"status": True,
@@ -1024,7 +1032,9 @@ def process_youtube_video(
content = " ".join([doc.page_content for doc in docs])
log.debug(f"text_content: {content}")
save_docs_to_vector_db(request, docs, collection_name, overwrite=True)
save_docs_to_vector_db(
request, docs, collection_name, overwrite=True, user=user
)
return {
"status": True,
@@ -1065,7 +1075,9 @@ def process_web(
content = " ".join([doc.page_content for doc in docs])
log.debug(f"text_content: {content}")
save_docs_to_vector_db(request, docs, collection_name, overwrite=True)
save_docs_to_vector_db(
request, docs, collection_name, overwrite=True, user=user
)
return {
"status": True,
@@ -1100,6 +1112,7 @@ def search_web(request: Request, engine: str, query: str) -> list[SearchResult]:
- SERPER_API_KEY
- SERPLY_API_KEY
- TAVILY_API_KEY
- EXA_API_KEY
- SEARCHAPI_API_KEY + SEARCHAPI_ENGINE (by default `google`)
Args:
query (str): The query to search for
@@ -1234,6 +1247,13 @@ def search_web(request: Request, engine: str, query: str) -> list[SearchResult]:
request.app.state.config.RAG_WEB_SEARCH_RESULT_COUNT,
request.app.state.config.RAG_WEB_SEARCH_DOMAIN_FILTER_LIST,
)
elif engine == "exa":
return search_exa(
request.app.state.config.EXA_API_KEY,
query,
request.app.state.config.RAG_WEB_SEARCH_RESULT_COUNT,
request.app.state.config.RAG_WEB_SEARCH_DOMAIN_FILTER_LIST,
)
else:
raise Exception("No search engine API key found in environment variables")
@@ -1273,7 +1293,9 @@ def process_web_search(
requests_per_second=request.app.state.config.RAG_WEB_SEARCH_CONCURRENT_REQUESTS,
)
docs = loader.load()
save_docs_to_vector_db(request, docs, collection_name, overwrite=True)
save_docs_to_vector_db(
request, docs, collection_name, overwrite=True, user=user
)
return {
"status": True,
@@ -1307,7 +1329,9 @@ def query_doc_handler(
return query_doc_with_hybrid_search(
collection_name=form_data.collection_name,
query=form_data.query,
embedding_function=request.app.state.EMBEDDING_FUNCTION,
embedding_function=lambda query: request.app.state.EMBEDDING_FUNCTION(
query, user=user
),
k=form_data.k if form_data.k else request.app.state.config.TOP_K,
reranking_function=request.app.state.rf,
r=(
@@ -1315,12 +1339,17 @@ def query_doc_handler(
if form_data.r
else request.app.state.config.RELEVANCE_THRESHOLD
),
user=user,
)
else:
return query_doc(
collection_name=form_data.collection_name,
query_embedding=request.app.state.EMBEDDING_FUNCTION(form_data.query, RAG_EMBEDDING_QUERY_PREFIX),
query_embedding=request.app.state.EMBEDDING_FUNCTION(
form_data.query, prefix=RAG_EMBEDDING_QUERY_PREFIX, user=user
),
k=form_data.k if form_data.k else request.app.state.config.TOP_K,
user=user,
)
except Exception as e:
log.exception(e)
@@ -1349,7 +1378,9 @@ def query_collection_handler(
return query_collection_with_hybrid_search(
collection_names=form_data.collection_names,
queries=[form_data.query],
embedding_function=request.app.state.EMBEDDING_FUNCTION,
embedding_function=lambda query: request.app.state.EMBEDDING_FUNCTION(
query, user=user
),
k=form_data.k if form_data.k else request.app.state.config.TOP_K,
reranking_function=request.app.state.rf,
r=(
@@ -1362,7 +1393,9 @@ def query_collection_handler(
return query_collection(
collection_names=form_data.collection_names,
queries=[form_data.query],
embedding_function=request.app.state.EMBEDDING_FUNCTION,
embedding_function=lambda query: request.app.state.EMBEDDING_FUNCTION(
query, user=user
),
k=form_data.k if form_data.k else request.app.state.config.TOP_K,
)
@@ -1510,6 +1543,7 @@ def process_files_batch(
docs=all_docs,
collection_name=collection_name,
add=True,
user=user,
)
# Update all files with collection name
backend/open_webui/routers/tasks.py
+19 -3
View File
@@ -4,6 +4,7 @@ from fastapi.responses import JSONResponse, RedirectResponse
from pydantic import BaseModel
from typing import Optional
import logging
import re
from open_webui.utils.chat import generate_chat_completion
from open_webui.utils.task import (
@@ -89,6 +90,10 @@ async def update_task_config(
form_data.TITLE_GENERATION_PROMPT_TEMPLATE
)
request.app.state.config.IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE = (
form_data.IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE
)
request.app.state.config.ENABLE_AUTOCOMPLETE_GENERATION = (
form_data.ENABLE_AUTOCOMPLETE_GENERATION
)
@@ -161,9 +166,20 @@ async def generate_title(
else:
template = DEFAULT_TITLE_GENERATION_PROMPT_TEMPLATE
messages = form_data["messages"]
# Remove reasoning details from the messages
for message in messages:
message["content"] = re.sub(
r"<details\s+type=\"reasoning\"[^>]*>.*?<\/details>",
"",
message["content"],
flags=re.S,
).strip()
content = title_generation_template(
template,
form_data["messages"],
messages,
{
"name": user.name,
"location": user.info.get("location") if user.info else None,
@@ -175,10 +191,10 @@ async def generate_title(
"messages": [{"role": "user", "content": content}],
"stream": False,
**(
{"max_tokens": 50}
{"max_tokens": 1000}
if models[task_model_id]["owned_by"] == "ollama"
else {
"max_completion_tokens": 50,
"max_completion_tokens": 1000,
}
),
"metadata": {
backend/open_webui/routers/tools.py
+5 -1
View File
@@ -227,7 +227,11 @@ async def delete_tools_by_id(
detail=ERROR_MESSAGES.NOT_FOUND,
)
if tools.user_id != user.id and user.role != "admin":
if (
tools.user_id != user.id
and not has_access(user.id, "write", tools.access_control)
and user.role != "admin"
):
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail=ERROR_MESSAGES.UNAUTHORIZED,
backend/open_webui/routers/users.py
+1
View File
@@ -79,6 +79,7 @@ class ChatPermissions(BaseModel):
class FeaturesPermissions(BaseModel):
web_search: bool = True
image_generation: bool = True
code_interpreter: bool = True
class UserPermissions(BaseModel):
backend/open_webui/socket/main.py
+5 -2
View File
@@ -325,7 +325,7 @@ def get_event_emitter(request_info):
def get_event_call(request_info):
async def __event_call__(event_data):
async def __event_caller__(event_data):
response = await sio.call(
"chat-events",
{
@@ -337,7 +337,10 @@ def get_event_call(request_info):
)
return response
return __event_call__
return __event_caller__
get_event_caller = get_event_call
def get_user_id_from_session_pool(sid):
backend/open_webui/utils/middleware.py
+796 -210
View File
File diff suppressed because it is too large Load Diff
backend/open_webui/utils/misc.py
+66 -5
View File
@@ -7,6 +7,18 @@ from pathlib import Path
from typing import Callable, Optional
import collections.abc
def deep_update(d, u):
for k, v in u.items():
if isinstance(v, collections.abc.Mapping):
d[k] = deep_update(d.get(k, {}), v)
else:
d[k] = v
return d
def get_message_list(messages, message_id):
"""
Reconstructs a list of messages in order up to the specified message_id.
@@ -20,7 +32,7 @@ def get_message_list(messages, message_id):
current_message = messages.get(message_id)
if not current_message:
return f"Message ID {message_id} not found in the history."
return None
# Reconstruct the chain by following the parentId links
message_list = []
@@ -131,6 +143,44 @@ def add_or_update_system_message(content: str, messages: list[dict]):
return messages
def add_or_update_user_message(content: str, messages: list[dict]):
"""
Adds a new user message at the end of the messages list
or updates the existing user message at the end.
:param msg: The message to be added or appended.
:param messages: The list of message dictionaries.
:return: The updated list of message dictionaries.
"""
if messages and messages[-1].get("role") == "user":
messages[-1]["content"] = f"{messages[-1]['content']}\n{content}"
else:
# Insert at the end
messages.append({"role": "user", "content": content})
return messages
def append_or_update_assistant_message(content: str, messages: list[dict]):
"""
Adds a new assistant message at the end of the messages list
or updates the existing assistant message at the end.
:param msg: The message to be added or appended.
:param messages: The list of message dictionaries.
:return: The updated list of message dictionaries.
"""
if messages and messages[-1].get("role") == "assistant":
messages[-1]["content"] = f"{messages[-1]['content']}\n{content}"
else:
# Insert at the end
messages.append({"role": "assistant", "content": content})
return messages
def openai_chat_message_template(model: str):
return {
"id": f"{model}-{str(uuid.uuid4())}",
@@ -141,13 +191,24 @@ def openai_chat_message_template(model: str):
def openai_chat_chunk_message_template(
model: str, message: Optional[str] = None, usage: Optional[dict] = None
model: str,
content: Optional[str] = None,
tool_calls: Optional[list[dict]] = None,
usage: Optional[dict] = None,
) -> dict:
template = openai_chat_message_template(model)
template["object"] = "chat.completion.chunk"
if message:
template["choices"][0]["delta"] = {"content": message}
else:
template["choices"][0]["index"] = 0
template["choices"][0]["delta"] = {}
if content:
template["choices"][0]["delta"]["content"] = content
if tool_calls:
template["choices"][0]["delta"]["tool_calls"] = tool_calls
if not content and not tool_calls:
template["choices"][0]["finish_reason"] = "stop"
if usage:
backend/open_webui/utils/oauth.py
+14 -8
View File
@@ -35,7 +35,7 @@ from open_webui.config import (
AppConfig,
)
from open_webui.constants import ERROR_MESSAGES, WEBHOOK_MESSAGES
from open_webui.env import WEBUI_SESSION_COOKIE_SAME_SITE, WEBUI_SESSION_COOKIE_SECURE
from open_webui.env import WEBUI_AUTH_COOKIE_SAME_SITE, WEBUI_AUTH_COOKIE_SECURE
from open_webui.utils.misc import parse_duration
from open_webui.utils.auth import get_password_hash, create_token
from open_webui.utils.webhook import post_webhook
@@ -82,7 +82,8 @@ class OAuthManager:
oauth_allowed_roles = auth_manager_config.OAUTH_ALLOWED_ROLES
oauth_admin_roles = auth_manager_config.OAUTH_ADMIN_ROLES
oauth_roles = None
role = "pending" # Default/fallback role if no matching roles are found
# Default/fallback role if no matching roles are found
role = auth_manager_config.DEFAULT_USER_ROLE
# Next block extracts the roles from the user data, accepting nested claims of any depth
if oauth_claim and oauth_allowed_roles and oauth_admin_roles:
@@ -273,11 +274,16 @@ class OAuthManager:
log.error(
f"Error downloading profile image '{picture_url}': {e}"
)
picture_url = ""
picture_url = "/user.png"
if not picture_url:
picture_url = "/user.png"
username_claim = auth_manager_config.OAUTH_USERNAME_CLAIM
name = user_data.get(username_claim)
if not isinstance(user, str):
name = email
role = self.get_user_role(None, user_data)
user = Auths.insert_new_auth(
@@ -285,7 +291,7 @@ class OAuthManager:
password=get_password_hash(
str(uuid.uuid4())
), # Random password, not used
name=user_data.get(username_claim, "User"),
name=name,
profile_image_url=picture_url,
role=role,
oauth_sub=provider_sub,
@@ -323,8 +329,8 @@ class OAuthManager:
key="token",
value=jwt_token,
httponly=True, # Ensures the cookie is not accessible via JavaScript
samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
secure=WEBUI_SESSION_COOKIE_SECURE,
samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
secure=WEBUI_AUTH_COOKIE_SECURE,
)
if ENABLE_OAUTH_SIGNUP.value:
@@ -333,8 +339,8 @@ class OAuthManager:
key="oauth_id_token",
value=oauth_id_token,
httponly=True,
samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
secure=WEBUI_SESSION_COOKIE_SECURE,
samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
secure=WEBUI_AUTH_COOKIE_SECURE,
)
# Redirect back to the frontend with the JWT token
redirect_url = f"{request.base_url}auth#token={jwt_token}"
backend/open_webui/utils/payload.py
+19 -2
View File
@@ -1,4 +1,4 @@
from open_webui.utils.task import prompt_template
from open_webui.utils.task import prompt_template, prompt_variables_template
from open_webui.utils.misc import (
add_or_update_system_message,
)
@@ -7,11 +7,14 @@ from typing import Callable, Optional
# inplace function: form_data is modified
def apply_model_system_prompt_to_body(params: dict, form_data: dict, user) -> dict:
def apply_model_system_prompt_to_body(
params: dict, form_data: dict, metadata: Optional[dict] = None, user=None
) -> dict:
system = params.get("system", None)
if not system:
return form_data
# Legacy (API Usage)
if user:
template_params = {
"user_name": user.name,
@@ -19,7 +22,15 @@ def apply_model_system_prompt_to_body(params: dict, form_data: dict, user) -> di
}
else:
template_params = {}
system = prompt_template(system, **template_params)
# Metadata (WebUI Usage)
if metadata:
variables = metadata.get("variables", {})
if variables:
system = prompt_variables_template(system, variables)
form_data["messages"] = add_or_update_system_message(
system, form_data.get("messages", [])
)
@@ -155,6 +166,9 @@ def convert_payload_openai_to_ollama(openai_payload: dict) -> dict:
)
ollama_payload["stream"] = openai_payload.get("stream", False)
if "tools" in openai_payload:
ollama_payload["tools"] = openai_payload["tools"]
if "format" in openai_payload:
ollama_payload["format"] = openai_payload["format"]
@@ -188,4 +202,7 @@ def convert_payload_openai_to_ollama(openai_payload: dict) -> dict:
if ollama_options:
ollama_payload["options"] = ollama_options
if "metadata" in openai_payload:
ollama_payload["metadata"] = openai_payload["metadata"]
return ollama_payload
backend/open_webui/utils/plugin.py
+9 -4
View File
@@ -167,9 +167,14 @@ def load_function_module_by_id(function_id, content=None):
def install_frontmatter_requirements(requirements):
if requirements:
req_list = [req.strip() for req in requirements.split(",")]
for req in req_list:
log.info(f"Installing requirement: {req}")
subprocess.check_call([sys.executable, "-m", "pip", "install", req])
try:
req_list = [req.strip() for req in requirements.split(",")]
for req in req_list:
log.info(f"Installing requirement: {req}")
subprocess.check_call([sys.executable, "-m", "pip", "install", req])
except Exception as e:
log.error(f"Error installing package: {req}")
raise e
else:
log.info("No requirements found in frontmatter.")
backend/open_webui/utils/response.py
+61 -2
View File
@@ -1,4 +1,5 @@
import json
from uuid import uuid4
from open_webui.utils.misc import (
openai_chat_chunk_message_template,
openai_chat_completion_message_template,
@@ -9,7 +10,48 @@ def convert_response_ollama_to_openai(ollama_response: dict) -> dict:
model = ollama_response.get("model", "ollama")
message_content = ollama_response.get("message", {}).get("content", "")
response = openai_chat_completion_message_template(model, message_content)
data = ollama_response
usage = {
"response_token/s": (
round(
(
(
data.get("eval_count", 0)
/ ((data.get("eval_duration", 0) / 10_000_000))
)
* 100
),
2,
)
if data.get("eval_duration", 0) > 0
else "N/A"
),
"prompt_token/s": (
round(
(
(
data.get("prompt_eval_count", 0)
/ ((data.get("prompt_eval_duration", 0) / 10_000_000))
)
* 100
),
2,
)
if data.get("prompt_eval_duration", 0) > 0
else "N/A"
),
"total_duration": data.get("total_duration", 0),
"load_duration": data.get("load_duration", 0),
"prompt_eval_count": data.get("prompt_eval_count", 0),
"prompt_eval_duration": data.get("prompt_eval_duration", 0),
"eval_count": data.get("eval_count", 0),
"eval_duration": data.get("eval_duration", 0),
"approximate_total": (lambda s: f"{s // 3600}h{(s % 3600) // 60}m{s % 60}s")(
(data.get("total_duration", 0) or 0) // 1_000_000_000
),
}
response = openai_chat_completion_message_template(model, message_content, usage)
return response
@@ -19,6 +61,23 @@ async def convert_streaming_response_ollama_to_openai(ollama_streaming_response)
model = data.get("model", "ollama")
message_content = data.get("message", {}).get("content", "")
tool_calls = data.get("message", {}).get("tool_calls", None)
openai_tool_calls = None
if tool_calls:
openai_tool_calls = []
for tool_call in tool_calls:
openai_tool_call = {
"index": tool_call.get("index", 0),
"id": tool_call.get("id", f"call_{str(uuid4())}"),
"type": "function",
"function": {
"name": tool_call.get("function", {}).get("name", ""),
"arguments": f"{tool_call.get('function', {}).get('arguments', {})}",
},
}
openai_tool_calls.append(openai_tool_call)
done = data.get("done", False)
usage = None
@@ -64,7 +123,7 @@ async def convert_streaming_response_ollama_to_openai(ollama_streaming_response)
}
data = openai_chat_chunk_message_template(
model, message_content if not done else None, usage
model, message_content if not done else None, openai_tool_calls, usage
)
line = f"data: {json.dumps(data)}\n\n"
backend/open_webui/utils/task.py
+6
View File
@@ -32,6 +32,12 @@ def get_task_model_id(
return task_model_id
def prompt_variables_template(template: str, variables: dict[str, str]) -> str:
for variable, value in variables.items():
template = template.replace(variable, value)
return template
def prompt_template(
template: str, user_name: Optional[str] = None, user_location: Optional[str] = None
) -> str:
backend/open_webui/utils/tools.py
+13
View File
@@ -61,6 +61,12 @@ def get_tools(
)
for spec in tools.specs:
# TODO: Fix hack for OpenAI API
# Some times breaks OpenAI but others don't. Leaving the comment
for val in spec.get("parameters", {}).get("properties", {}).values():
if val["type"] == "str":
val["type"] = "string"
# Remove internal parameters
spec["parameters"]["properties"] = {
key: val
@@ -73,6 +79,13 @@ def get_tools(
# convert to function that takes only model params and inserts custom params
original_func = getattr(module, function_name)
callable = apply_extra_params_to_tool_function(original_func, extra_params)
if callable.__doc__ and callable.__doc__.strip() != "":
s = re.split(":(param|return)", callable.__doc__, 1)
spec["description"] = s[0]
else:
spec["description"] = function_name
# TODO: This needs to be a pydantic model
tool_dict = {
"toolkit_id": tool_id,
backend/requirements.txt
+8 -8
View File
@@ -1,4 +1,4 @@
fastapi==0.111.0
fastapi==0.115.7
uvicorn[standard]==0.30.6
pydantic==2.9.2
python-multipart==0.0.18
@@ -11,7 +11,7 @@ python-jose==3.3.0
passlib[bcrypt]==1.7.4
requests==2.32.3
aiohttp==3.11.8
aiohttp==3.11.11
async-timeout
aiocache
aiofiles
@@ -57,10 +57,10 @@ einops==0.8.0
ftfy==6.2.3
pypdf==4.3.1
fpdf2==2.8.2
pymdown-extensions==10.11.2
pymdown-extensions==10.14.2
docx2txt==0.8
python-pptx==1.0.0
unstructured==0.15.9
unstructured==0.16.11
nltk==3.9.1
Markdown==3.7
pypandoc==1.13
@@ -71,16 +71,16 @@ xlrd==2.0.1
validators==0.34.0
psutil
sentencepiece
soundfile==0.12.1
soundfile==0.13.1
opencv-python-headless==4.10.0.84
opencv-python-headless==4.11.0.86
rapidocr-onnxruntime==1.3.24
rank-bm25==0.2.2
faster-whisper==1.0.3
PyJWT[crypto]==2.10.1
authlib==1.3.2
authlib==1.4.1
black==24.8.0
langfuse==2.44.0
@@ -89,7 +89,7 @@ pytube==15.0.0
extract_msg
pydub
duckduckgo-search~=7.2.1
duckduckgo-search~=7.3.0
## Google Drive
google-api-python-client