retrieval
Retrieval, RAG configuration, and web search models.
Classes
Bases: BaseModel
Form for specifying a collection name.
- Code Reference models retrieval Classes ProcessUrlForm
Bases: CollectionNameForm
Form for processing a URL.
- Code Reference routers retrieval Classes RetrievalClient Functions
Bases: BaseModel
Form for search queries.
- Code Reference routers retrieval Classes RetrievalClient Functions process_web_search
Bases: BaseModel
Configuration for OpenAI embedding model.
- Code Reference models retrieval Classes EmbeddingModelUpdateForm Attributes openai_config
Bases: BaseModel
Configuration for Ollama embedding model.
- Code Reference models retrieval Classes EmbeddingModelUpdateForm Attributes ollama_config
Bases: BaseModel
Configuration for Azure OpenAI embedding model.
- Code Reference models retrieval Classes EmbeddingModelUpdateForm Attributes azure_openai_config
Bases: BaseModel
Form for updating the embedding model configuration.
- Code Reference routers retrieval Classes RetrievalClient Functions update_embedding_config
Attributes
openai_config: Optional[OpenAIConfigForm] = None
Configuration for OpenAI embedding model.
ollama_config: Optional[OllamaConfigForm] = None
Configuration for Ollama embedding model.
azure_openai_config: Optional[AzureOpenAIConfigForm] = None
Configuration for Azure OpenAI embedding model.
The embedding engine to use (e.g., 'ollama', 'openai').
The batch size for embedding generation.
Whether to enable asynchronous embedding generation.
Bases: BaseModel
Configuration for web search and retrieval.
- Code Reference models retrieval Classes ConfigForm Attributes web
Attributes
Whether users must confirm before a web search runs. When enabled, the client
UI shows the WEB_SEARCH_CONFIRMATION_CONTENT message and requires acknowledgement
before the search proceeds.
Confirmation message shown to users before a web search runs, when
ENABLE_WEB_SEARCH_CONFIRMATION is enabled. Defaults to
'Your query will be sent to the configured web search provider.'.
Whether to trust the environment variables for web search.
The number of web search results to retrieve.
The number of concurrent web search requests.
The number of concurrent web loader requests.
List of domains to filter from web search results.
Maximum content length in characters for web fetch results. Content exceeding this is truncated.
Whether to bypass embedding and retrieval for web search results.
Whether to bypass the web loader for web search results.
API key for Ollama Cloud web search.
API key for Google Programmable Search Engine.
Engine ID for Google Programmable Search Engine.
Maximum number of context tokens returned by Brave Search. Defaults to 8192.
API key for SERPHouse Search (used when WEB_SEARCH_ENGINE == 'serphouse').
Search domain passed to SERPHouse as the domain query parameter, e.g.
'google.com' or 'bing.com'. Defaults to 'google.com' if empty.
The endpoint for Bing Search V7.
The subscription key for Bing Search V7.
The search context usage for Perplexity.
The search API URL for Perplexity.
Base URL for the Microsoft Web IQ API (used for both search and page browsing),
selected when WEB_SEARCH_ENGINE == 'microsoft_web_iq' or
WEB_LOADER_ENGINE == 'microsoft_web_iq'. Defaults to 'https://api.microsoft.ai/v3'.
API key for the Microsoft Web IQ API (sent as the x-apikey header).
Language code forwarded to the Microsoft Web IQ API as the language field, e.g.
'en'. Defaults to 'en'.
Whether to enable SSL verification for the web loader.
The URL for external web search.
The API key for external web search.
The URL for external web loader.
The API key for external web loader.
List of languages for YouTube loader.
The proxy URL for YouTube loader.
The translation language for YouTube loader.
API key for Yandex Search.
JSON configuration string for Yandex search.
Dict Fields (when parsed as JSON):
- query (dict, optional): Query configuration options.
- searchType (str, optional): Search type, e.g., 'SEARCH_TYPE_COM'.
- Additional Yandex API parameters may be included.
Defaults to '{"query": {"searchType": "SEARCH_TYPE_COM"}}' if not specified.
Parameters for Linkup search.
Dict Fields
url(str, optional): Override endpoint URL. Defaults to 'https://api.linkup.so/v1/search'.depth(str, optional): Search depth. Typical values: 'standard', 'deep'. Defaults to 'standard'.outputType(str, optional): Output type. Typical values: 'sourcedAnswer', 'searchResults'. Defaults to 'sourcedAnswer'.- Additional Linkup API parameters may be included.
The dictionary is forwarded to the Linkup Search API as the JSON body (with q and maxResults
injected automatically). The special url key, if present, is popped and used as the request
endpoint instead of the default. See the Linkup API documentation for additional supported parameters.
Bases: BaseModel
Configuration form for retrieval settings.
- Code Reference routers retrieval Classes RetrievalClient Functions update_config
Attributes
Whether to bypass embedding and retrieval.
Whether to enable hybrid search.
Whether to enable enriched texts for hybrid search.
Relevance threshold for search results.
Engine for content extraction.
Mode for PDF loading. 'page' creates one document per page, 'single' combines all pages into one document for better chunking across page boundaries.
Base URL for DataLab Marker API.
Additional configuration for DataLab Marker.
Whether to skip cache for DataLab Marker.
Whether to force OCR for DataLab Marker.
Whether to paginate results for DataLab Marker.
Whether to strip existing OCR for DataLab Marker.
Whether to disable image extraction for DataLab Marker.
Whether to format lines for DataLab Marker.
Whether to use LLM for DataLab Marker.
Output format for DataLab Marker.
URL for external document loader.
API key for external document loader.
Extra HTTP headers appended to requests sent to the external document loader
server, in addition to the auto-injected Content-Type, Authorization (built
from EXTERNAL_DOCUMENT_LOADER_API_KEY), and X-Filename. Only used when
CONTENT_EXTRACTION_ENGINE == 'external'.
Values are strings (any non-string value is coerced to a string at request time) and may contain template tokens that are substituted per uploaded file before the request is sent.
Dict Fields
<header-name>(str, optional): Any HTTP header name mapped to a string value. Values support case-sensitive template tokens that are replaced at request time, including{{FILE_ID}},{{FILE_NAME}},{{FILE_CONTENT_TYPE}},{{CHAT_ID}},{{MESSAGE_ID}},{{USER_MESSAGE_ID}},{{USER_MESSAGE_PARENT_ID}},{{USER_ID}},{{USER_NAME}},{{USER_EMAIL}},{{USER_ROLE}},{{USER_AGENT}}, and{{TASK}}.
Defaults to {} when unset. Example:
{"X-OpenWebUI-File-Id": "{{FILE_ID}}"}.
Parameters for Docling.
Dict Fields
image_export_mode(str, optional): How images should be exported. Defaults to "placeholder" if not specified.- Additional VLM (Vision Language Model) pipeline parameters may be supported by the Docling API.
This dictionary is passed directly to the Docling API's /v1/convert/file endpoint. See the Docling API documentation for additional supported parameters.
Endpoint for Document Intelligence.
Key for Document Intelligence.
Model for Document Intelligence.
Base URL for Mistral OCR API.
When True (and CONTENT_EXTRACTION_ENGINE == 'mistral_ocr'), send the PDF as a
base64 data URL inline instead of first uploading it to Mistral and referencing
the uploaded file. Defaults to False.
Base URL for PaddleOCR VL service. Defaults to 'http://localhost:8080'.
Authentication token for PaddleOCR VL service.
Parameters for MinerU.
Dict Fields
enable_ocr(bool, optional): Enable OCR processing. Defaults to False.enable_formula(bool, optional): Enable formula processing. Defaults to True.enable_table(bool, optional): Enable table processing. Defaults to True.language(str, optional): Language code for processing. Defaults to "en".model_version(str, optional): Model version to use. Defaults to "pipeline".page_ranges(str, optional): Page ranges to process. Defaults to empty string.
This dictionary is passed directly to the MinerU API for document parsing configuration.
List of file extensions that MinerU is allowed to process (e.g., ['pdf']).
Files uploaded to the system with extensions in this list are routed through the MinerU
content extraction engine when CONTENT_EXTRACTION_ENGINE is set to mineru. Frontend
typically accepts a comma-separated string (e.g., 'pdf') and splits it into a list.
Defaults to ['pdf'] if not specified.
Batch size for reranking operations. Defaults to 32.
URL for external reranker.
API key for external reranker.
The timeout for the external reranker.
HuggingFace tokenizer model id (or local path) used by the 'token_transformers'
text splitter (TEXT_SPLITTER == 'token_transformers') for token-based chunking.
A bare name with no path/slash is prefixed with 'sentence-transformers/'. When
empty, the backend falls back to the configured embedding model's tokenizer and
raises if none is available.
Whether to enable markdown header text splitter.
Minimum target size for text chunks.
Width for image compression.
Height for image compression.
List of allowed file extensions.
Whether to enable Google Drive integration.
Whether to enable OneDrive integration.
Bases: BaseModel
Form for processing a file.
- Code Reference routers retrieval Classes RetrievalClient Functions process_file
Bases: BaseModel
Form for processing text.
- Code Reference routers retrieval Classes RetrievalClient Functions process_text
Bases: BaseModel
Form for querying a document.
- Code Reference routers retrieval Classes RetrievalClient Functions query_doc
Attributes
Bases: BaseModel
Form for querying multiple collections.
- Code Reference routers retrieval Classes RetrievalClient Functions query_collection
Attributes
Bases: BaseModel
Form for deleting a file from a collection.
- Code Reference routers retrieval Classes RetrievalClient Functions delete
Bases: BaseModel
Form for batch processing files.
- Code Reference routers retrieval Classes RetrievalClient Functions process_files_batch
Attributes
Bases: BaseModel
Result of a batch file processing operation.
- Code Reference models retrieval Classes BatchProcessFilesResponse Attributes
Bases: BaseModel
Response for batch process files request.
- Code Reference routers retrieval Classes RetrievalClient Functions process_files_batch