Error message if processing had failed (snapshot; see :attr:`status`).

* Convenience methods

Get cached settings instance.

Make HTTP request

Make HTTP request

Construct a SQL WHERE clause from a metadata filter dictionary.

List accessible documents. Args: skip: Number of documents to skip limit: Maximum number of documents to ret

Normalize user-provided type labels to canonical metadata type names.

Ingest a text document into Morphik. Args: content: Text content to ingest metadata: Optional metadata dicti

Parse document and return page-wise XML chunks with bbox metadata. Uses API if configured, otherwise falls back to local parsing. Ar

Close the HTTP client

Delete a document and all its associated data. This method deletes a document and all its associated data, including: - Docu

Normalize a folder path into canonical form (leading slash, no trailing slash). Rules: - Trim outer whitespace. - Strip leading/trai

Update document metadata if user has write access. Args: document_id: The document ID to update updates: Dictionary o

Returns a UserScope object scoped to this folder and the end user. Args: end_user_id: The ID of the end user Re

Get a PostgreSQL connection with retry logic. Yields: A PostgreSQL connection object Raises: psycopg.Operati

Retrieve document metadata by ID if user has access.

Escape single quotes for SQL literals.

Initialize database tables for the user limits system if not already done.

Parse document response

Check or record usage against organization limits. When *verify_only* is **True** the function **only** validates that the operation **would*

Resolve requested API fields to the document table columns needed to serve them. Lightweight status keys (``status``, ``error``, timestamps)

Normalize metadata values and build a type map for top-level fields.

Get document metadata by filename. If multiple documents have the same filename, returns the most recently updated one. Args

Retrieve relevant chunks. Args: query: Search query text (mutually exclusive with query_image) filters: Opti

Bulk-insert embeddings in one go instead of row-by-row ORM adds. This avoids per-row overhead and cuts the SQL round-trips from *n*

Build a Chunk for a page image. Raw bytes are attached only in local ColPali mode, where the in-process model reads them; in API mode

Detect file type from content string and return appropriate extension. Content can be either base64 encoded or plain text.

Download file from S3 asynchronously using a dedicated thread pool. Uses a 64-thread executor to support parallel downloads of 50-75 multivec

Encode bytes to base64 string. Uses Rust SIMD-optimized implementation when available (2-3x faster).

Create a folder to scope operations. Args: name: The name of the folder (leaf segment when using nested paths)

Update a document by replacing its content and/or metadata. Args: document_id: ID of the document to update

Close the HTTP client

Generate embedding for a query. Args: text: Query text to embed Returns: Embedding vector

Generate completion using relevant chunks as context. Args: query: Query text filters: Optional metadata fil

Call a specific ColPali API endpoint, retrying transient failures. 429/5xx responses, timeouts and transport errors are retried with

Check if content field contains a storage key rather than actual content.

True when the error (or its direct cause) is transient infrastructure failure.

Get user limits information.

Ingests file content from bytes. Saves to storage, creates document record, and then enqueues a background job for chunking and embed

Upload a file to S3 using a shared executor for true concurrency.

Upload base64-encoded content to S3. Accepts either a raw base64 string or a data URI (e.g. "data:image/png;base64,..."). Does not pr

Resolve a folder identifier that might be either an ID or a name.

Recursively serialize datetime objects to ISO format strings.

Add a sub-operation timing that will be displayed under its parent phase

Delete file from S3.

Generate embeddings for chunks to be ingested into the vector store. Args: chunks: Single chunk or list of chunks to emb

Move and/or rename a folder by setting its new canonical path. Args: folder_id_or_name: Folder ID, name, or full path

Return JSON-serializable metadata plus a parallel type map. Args: metadata: Original metadata dictionary. type_hints: Optional ex

Rerank chunks based on their relevance to the query

Check if the user has access to the folder.

Combine multiple SQL clauses with a logical operator.

Convert bytes to a data URI string. Uses Rust implementation when available for faster base64 encoding.

Delete chunks for a document, scoped by app_id to prevent cross-tenant deletion.

* Make a fetch request with proper authentication based on connection type