docvision_annotation_add
ChatGPTNON-destructive annotation (highlight/strike/box/comment) tied to a page/area. Never modifies the original.
Productivity
Overview
DocVision AI is a stateless MCP document-intelligence connector for PDFs. From a file you provide (base64 or URL) it opens and inspects the document (pages, geometry, metadata, SHA-256 fingerprint), extracts the real text page by page with x/y provenance, runs OCR on scanned pages, searches, and performs page operations (rotate, delete, extract, reorder, split, merge) that always return a NEW PDF rather than modifying the original. It compares two documents and flags changed amounts, dates and IDs; extracts invoice/form fields where each value is marked SUPPORTED_BY_SOURCE with its provenance and absent values stay UNKNOWN; detects personal data to redact and returns it already masked; and keeps a persistent Document-OS layer (immutable versions, fields, annotations, patches, reviews, batches, RBAC policy, audit). It blocks internal/metadata URLs (anti-SSRF), refuses to empty a document, and does not authenticate signatures or assert a forgery verdict; forensics flags anomalies only.
Tools
docvision_annotations_list
ChatGPTPersisted annotations of a document.
docvision_apercu
ChatGPTDescribes DocVision AI: the 3 layers, what is REAL vs deferred (worker layer), the integrity rules, and the tool list.
docvision_audit_log
ChatGPTAppend-only log of operations on a document (creation, patch, field, annotation, review, quality).
docvision_batch_create
ChatGPTCreates a persistent processing batch. items:[{ref, fields:[{name,type,value,confidence,disagreement?}], signature_present?}]. Each item will be routed by the quality gate.
docvision_batch_list
ChatGPTLists persisted batches.
docvision_batch_process
ChatGPTApplies the quality gate to each item of a batch and routes it → AUTO_APPROVED / REVIEW / FAILED, with metrics (including average confidence). FAILED/REVIEW go to the queue, not to export.
docvision_batch_review_queue
ChatGPTItems of a batch by status (REVIEW by default, or FAILED, or * for all).
docvision_batch_status
ChatGPTMetrics of a batch (processed, auto-approved, review, failed, average confidence).
docvision_diff
ChatGPTCompares two texts/versions: separates CONTENT CHANGE from visual MOVE, and flags changed NUMBERS (amounts/dates/IDs).
docvision_doc_create
ChatGPTPersistently stores a document; its content becomes the IMMUTABLE original version v1 (always preserved). content:{text?,structure?,pdf_sha256?}. Returns document_id.
docvision_doc_get
ChatGPTMetadata of a persistent document.
docvision_doc_history
ChatGPTVersions of a document (original/current/derivations).
docvision_doccanvas
ChatGPTDisplays the DocCanvas widget (original): Document panel (fields + color-coded confidence + provenance), Quality panel (heatmap + review queue), Versions, and Accept/Reject/Retry actions. Pass
document_id (loads the persistent document) and/or fields.docvision_docs_list
ChatGPTLists persisted documents.
docvision_extract_fields
ChatGPTExtracts candidate fields (invoice/contract/form/statement/admin) from the provided pages, with PROVENANCE (page/area) and SUPPORTED_BY_SOURCE status. Absent values stay UNKNOWN.
docvision_extract_text
ChatGPTExtracts the real text of a PDF page by page WITH normalized positions (provenance). Flags pages with no text layer (scanned → raster OCR = worker layer). Fabricates no text.
docvision_field_record
ChatGPTPersistently records a field (value + provenance + confidence + status). Absent value → UNKNOWN.
docvision_fields_list
ChatGPTPersisted extracted fields of a document.
docvision_forensics_inspect
ChatGPTAnalyzes the provided integrity signals (metadata, per-page fonts, images, geometry, incremental updates) and flags ANOMALIES. NEVER concludes forgery (forgery verdict stays NOT_ESTABLISHED). Signature detected is not authenticated.
docvision_inspect
ChatGPTOpens a PDF (base64 or url) and returns pages, geometry, metadata and SHA-256 fingerprint. Real processing (pdf-lib).
docvision_merge
ChatGPTMerges several PDFs (list of base64/url sources) into one (base64), in the order provided.
docvision_ocr_consensus
ChatGPTResolves the text of ONE region from the outputs of several OCR engines (agreement/majority/divergence), with confidence and a review flag. Illegible → UNKNOWN (never invented).
critical:true tightens the threshold (amounts/dates).docvision_ocr_run
ChatGPTREAL OCR of a PDF (including scanned): rasterizes each page and recognizes text with Tesseract, keeping per-word CONFIDENCE (words < 85% are flagged for review, never silently corrected). Execution runs in a CPU Vercel Sandbox (no serverless limit), so the model loads only once — no timeout. langs e.g. 'eng', 'fra', 'fra+eng'. Default 5 pages/call (max 20). If the Sandbox infra is not wired up, returns an honest WORKER_NOT_CONFIGURED status (never a fake result).
docvision_page_diff
ChatGPTCompares two lists of page hashes/summaries (added/removed/unchanged).
docvision_pages_delete
ChatGPTDeletes pages (spec '2,5-6') and returns a NEW PDF (base64). Irreversible operation on the COPY: the provided original stays intact. Refuses to delete everything.
docvision_pages_extract
ChatGPTExtracts pages (spec '1-3,7') into a NEW PDF (base64).
docvision_pages_reorder
ChatGPTReorders pages according to a complete permutation (e.g. [3,1,2]) and returns a NEW PDF (base64).
docvision_pages_rotate
ChatGPTRotates pages of a PDF (spec '1,3-5' or empty=all) and returns a NEW PDF (base64). The original is never modified.
docvision_patch_apply
ChatGPTApplies a proposed patch: creates an immutable derived VERSION; the original v1 and all versions are preserved. confirm:true required.
docvision_patch_propose
ChatGPTProposes a traceable edit (correct_ocr/rewrite/redact/delete…) WITHOUT touching the original. To be confirmed via apply.
docvision_patch_reject
ChatGPTRejects (or retries) a patch without modifying the original.
docvision_patches_list
ChatGPTPatches of a document and their status.
docvision_policy_check
ChatGPTIndicates whether a role can perform a capability (and the required approval). Executes nothing.
docvision_policy_show
ChatGPTReturns the default RBAC schema (viewer/reviewer/editor/admin/auditor roles, rules, approvals).
docvision_quality_gates
ChatGPTApplies configurable quality thresholds to fields {name,type,value,confidence,disagreement?}. Blocks export if a critical field (amount/date/id) is below threshold. Default thresholds (amounts≥99.8, dates≥99…).
docvision_quality_report
ChatGPTConsolidates gates + round-trip + review queue + forensics into one report, and highlights HIGH-CONFIDENCE errors (the most dangerous).
docvision_quality_save
ChatGPTPersistently saves a quality report for a document.
docvision_reconstruct_docx
ChatGPTProduces a REAL EDITABLE .docx binary (Word styles, lists, tables) from a structure (blocks, or
pages from docvision_structure_analyze). Returns the .docx as base64. Truly editable, not an image + text.docvision_redaction_detect
ChatGPTDetects sensitive candidates (PII, financial, identifiers) in the provided pages, with provenance. Custom rules possible.
docvision_redaction_plan
ChatGPTProduces a redaction PLAN from the candidates + policy. PLAN only: the real burn-in (removal of the underlying content) = worker layer.
docvision_relation_add
ChatGPTCreates a relation between two documents (e.g. version/translation/source).
docvision_render_page
ChatGPTRenders a PDF page as a PNG image (base64) — useful for preview, the DocCanvas, or before OCR.
scale (default 2), grayscale for light restoration.docvision_review_decide
ChatGPTRecords a review decision (ACCEPT/REJECT/RETRY) on a target (patch, region, field).
docvision_review_queue
ChatGPTBuilds the review queue from resolved regions (or {engines}): keeps only what needs review, with the heatmap (green/yellow/red/gray).
docvision_roundtrip_validate
ChatGPTCompares the ORIGINAL and the GENERATED rendering (text/tables/images/critical fields) and returns scores + review areas. DocVision checks its own work (the gap is not hidden).
docvision_search
ChatGPTFull-text search within a PDF with location (page + snippet). Real.
docvision_split
ChatGPTSplits a PDF into two parts (after page
at, default middle). Returns two PDFs (base64).docvision_structure_analyze
ChatGPTTransforms extracted text (pages) into a document tree (headings/paragraphs/lists). Honest heuristic; fine tables/figures via provided regions or worker.
docvision_version_get
ChatGPTContent of a specific version (or the current one) of a document.
docvision_workspace
ChatGPTOpens the DocVision Studio workspace — a 3-column creative workspace (pages · canvas · inspector) to read, select, edit a passage (→ new immutable version via patch; the original is preserved) and annotate, plus a visual (non-certified) signature and client-side exports. An ALTERNATIVE presentation to docvision_doccanvas; pass
document_id to load a persistent document.Capabilities
WritesInteractive
Example Prompts
Click any prompt to copy it.
