QDocs

QDocs is a set of standard QBO operations on documents (Attachment records) and scores, with no separate service to call. Each stage of the pipeline is one operation:

Stage	Operation	Module	Implemented in
Split	pdfsplit (and pdfsplitbypage, pdfsplitbyoutline, pdfsplitbysize, pdfsplitbysemantics)	attachment	qbo4.Document (Aspose PDF)
Split, fan-out	splitpdf	attachment	qbo3 attachment
Read	ocrtodescription	attachment	qbo3 attachment
Classify	aiclassify	attachment	qbo4.AI
Label	ailabel	any (here attachment)	qbo4.AI
Transcribe	aitranscribe	attachment	qbo4.AI
Extract	calculate with an AI scoring engine	score	qbo3 score, qbo4.AI engines
Batch	import on an import file whose template uses a batch engine	importfile	qbo4.AI, qbo4.Document

All operations follow the standard route pattern:

• /api/{module}/{operation} runs synchronously on the web server.
• /api/{module}/queue/{operation} queues the operation for an app server. Use it for splitting, extraction, and anything that touches many documents.

Every call runs as the authenticated user, against records that user can reach. Classification, labels, extraction, transcription, agents, and token tracking are documented in full in the AI developer reference; this page covers how QDocs uses them, and documents splitting in full.

Split a PDF

GET /api/attachment/pdfsplit?ID=2207
GET /api/attachment/queue/pdfsplit?ID=2207&Target_Object=Process&Target_ObjectID=42&LedgerID=310
GET /api/attachment/splitpdf?ID=2207,2208,2209&SplitQueueName=DocumentQueue

pdfsplit runs the configured splitters in order and stops at the first one that can split the PDF. The default order is the outline (bookmark) splitter, then the AI splitter, so a PDF with bookmarks is split by them and any other PDF goes to the model. It returns the new documents.

splitpdf is a fan-out: it queues one pdfsplit per ID on SplitQueueName (default: the installation’s default queue), passing every other parameter through. The Documents panel’s Split PDF option calls it through the queue.

Parameter	Purpose
ID	The document to split.
Target_*	The parent of the new documents, e.g. Target_Object=Process&Target_ObjectID=42. Defaults to the package’s own parent.
Options_*	Binds to the splitter options with the prefix removed, e.g. Options_Splitters=Ai, Options_DestinationPath=…, Options_Outline:NestLevel=1.
ImportFileTemplate	Name of an import template whose Description replaces the AI splitter’s instructions. InstructionTemplate is a synonym.
PromptSignature	A method signature whose JSON result is a list of document definitions. QBO builds classify-and-extract instructions from it (see below). Ignored when ImportFileTemplate is set.
SingleSplitDelete	Default true: when the model finds only one document, the original is updated in place and the duplicate removed.
LedgerID	Posts the call’s token usage to this ledger.

Each new document is saved with SetDocumentInfo, linked to the package with the link type Document Package Source, and stamped with the segment’s metadata. Metadata keys that match Attachment columns are set as fields (with MergeXmlData); keys that don’t are written to XmlData.

AI splitter options

Option	Default	Purpose
Ai:AgentName	PdfAnalysis	The agent (IAgentManager) that receives the PDF. It must accept PDF content.
Ai:Chat	Temperature 0, built-in instructions	Chat options forwarded to the agent, including Instructions.
Ai:MaxResolution / Ai:ImageQuality	72 / 80	A downsampled copy is sent to reduce tokens; page numbers are unchanged. 0 disables downsampling.
Ai:MaxPageCount / Ai:MaxChunkBytes	500 / 40 MB	Larger PDFs are cut into size-bounded chunks and sent together in one request, with a map of each chunk’s page range.
Ai:AddOutlineToSource	false	Adds a bookmark per segment to the original package.

The model must return a JSON array of segments:

[
  { "StartPage": 1, "EndPage": 3, "FileName": "FirstNoticeOfLoss", "DocumentType": "First notice of loss" },
  { "StartPage": 4, "EndPage": 9, "FileName": "RepairEstimate", "DocumentType": "Repair estimate",
    "EstimateTotal": "8412.50 :: 0.94" }
]

StartPage and EndPage are required and 1-based. Extracted values use the value :: confidence convention. An empty array means the PDF isn’t compound, and nothing is split.

With PromptSignature, QBO invokes the signature, inserts its JSON as Document Definitions, and asks the model to match each segment to a definition’s DocumentTemplate, return its AttachmentTemplateID, and extract each of its ExtractionFields as its ValueType with a confidence. Segments must not overlap and must cover every page.

Other split operations

GET /api/attachment/pdfsplitbypage?ID=17&Pages=[{"startPage":1,"title":"Intro"},{"startPage":6,"title":"Body"}]
GET /api/attachment/pdfsplitbyoutline?ID=17&NestLevel=1
GET /api/attachment/pdfsplitbysize?ID=17&MaxFileSize=5000000
GET /api/attachment/pdfsplitbysemantics?ID=17&MaxFileSize=20000000

All accept Target_*. pdfsplitbypage, pdfsplitbysize, and pdfsplitbysemantics link each new document to the package as Document Package Source; pdfsplitbyoutline does not.

Classify, label, and transcribe

GET /api/attachment/queue/aiclassify?ID=2208,2209&AgentName=Gemini&DefaultTemplateID=4
GET /api/attachment/ailabel?ID=2208&AgentName=Gemini
GET /api/attachment/queue/aitranscribe?ID=2210&ScoreTemplateID=12

• aiclassify chooses among attachment templates that have a Description (narrow with Template_* filters; add Template_nocache=true after editing descriptions). It OCRs first when the document has no text and OCR is configured.
• ailabel applies every label (collection) whose SourceObject is Attachment and whose description fits. Narrow with Label_* filters.
• aitranscribe appends a transcript of an mp3 to Description, then optionally calculates ScoreTemplateID.

Parameters and behavior: AI developer reference → Operations.

Extract fields

Extraction is a score calculated on the document with an AI engine:

GET /api/score/calculate?Template=Repair Estimate&Object=Attachment&ObjectID=2208
GET /api/score/queue/calculate?Template=Repair Estimate&Object=Attachment&ObjectID=2208

Engine	Input	Registry
AIScoreEngine	Attachment.Description (converted to text if empty)	Completion service (ICompleteManager)
AiModalEngine	The original file, sent as document content	Agent (IAgentManager)

Set AgentName, UseConfidence (default on), and, for the modal engine, FileExpirationMinutes in ScoreTemplate.EngineParameters. Each ScoreItem stores the typed value and its Confidence; Score.Confidence is the weighted average. Recalculate confidence with /api/score/calculateconfidence?ID=…. See Scoring developer reference and Scoring and import engines.

Batch jobs

Batch engines are import engines. Create an import file from a template that uses one, then run it:

GET /api/importfile/queue/import?ID=1234

Engine	Instructions	Documents	Each result
PdfSplitterBatchEngine	ImportFileTemplate.Description	ImportStatement, or Attachment/Search with the call’s parameters	Attachment/ProcessPdfSplitterResult: one segment updates the document in place; several create linked child documents
ClassificationBatchEngine	ImportFileTemplate.Description	ImportStatement, or Attachment/Search	One JSON object per document. Top-level properties are applied with SetProperties; the full object is saved to XmlData
AiBatchEngine	Instructions parameter, or ImportFileTemplate.Description	ImportStatement (required)	ResponseMethod, a method signature substituted from each row, unless the statement returns a Response column

Engine options come from configuration (for example AsposePDF:PdfSplitterBatchEngine and GenerativeAI:ClassificationBatchEngine) and are overridden per template in ImportFileTemplate.ParameterString:

Option	Default	Purpose
Agent	default (AI batch only)	Agent name. It must implement IBatchChatClient (Claude, Gemini, OpenAI, or Bedrock with batch enabled).
PollIntervalMinutes	60	Time between status polls.
ProcessPartialResults	true	Recover and apply truncated responses.
QueueResults	true	Queue each result through ImportFileQueue instead of processing inline.
ProgressLogInterval	1000	Results between progress log entries.

Submission saves the provider’s batch ID in ImportFile.ExternalReference and schedules a poll. When a submission spans several provider jobs, each extra job gets a child import file. Per-document errors are written to the import log and counted in ErrorCount.

Token cost

Pass LedgerID to pdfsplit, or Chat_LedgerID to agent-based operations, to post input, cached input, output, and reasoning tokens to a ledger. On Gemini, long instructions are folded into a cachedContent entry keyed by content hash, so a large set of document definitions isn’t re-sent with every request. See Token cost tracking.

Parameters

The parameters shown are the commonly used ones. Check the operation’s method signature in source (paths listed under the page title) before relying on less common options.