26.1 changelog

This changelog covers the changes in ADM 26.1 since 0.1.6, the previous tagged release. Starting with this release, version numbers use YY.N: the year followed by the release number within that year.

Added

File browser

• A new file-browser region with table and grid views, a folder tree, breadcrumbs and saved view preferences.
• Drag-and-drop upload of files and directories, with chunked transfers, progress reporting, cancellation and retry.
• Document and folder actions directly in the browser, including create, rename, move, delete, restore, ZIP and unzip.
• File details on hover, with optional recursive folder sizes and document counts.
• Responsive layouts, keyboard navigation and screen-reader support.
• A reusable document-actions component for pages outside the file browser.

See Working with documents.

ZIP archives

• Create an archive from a document, a selection of documents or a folder tree with adm_zip_api.
• Extract archives into nested folders while preserving entry names, including spaces.
• ZIP operations respect document permissions and support both database and object storage.

See ZIP and unzip.

File metadata

• Automatic extraction of file metadata into searchable file.* annotations, including information such as titles, page counts and image dimensions where available.
• A scheduled backfill for existing documents, controlled by the metadata extraction settings.

Integration tools

• Six session-scoped adm_my_*_v views for querying accessible documents, folders, versions, document tags, and document and folder annotations from your own application.
• Granular embed permissions for individual actions and upload limits, alongside existing read-only embedding. Embed sessions can use the new file browser.
• Support for an embedded session to act as a user identified by the host application.
• Nine customer agent skills covering the ADM APIs and integration conventions.

See Querying ADM from SQL, Embedding and AI coding agents.

New: AI Pack

The AI Pack is introduced in 26.1 as an optional, separately licensed add-on. It adds AI search, answers and agents over the documents stored in ADM.

• RAG collections: select documents by folder or annotation, extract their text, split it into chunks and generate embeddings.
• Search and answers: semantic retrieval, source preparation and generated answers, with optional query rewriting, hybrid keyword/vector search and surrounding-chunk context.
• Vector storage: Qdrant or native Oracle AI Vector Search on a supported database, with configurable vector indexes and retrieval settings.
• Text extraction: database text extraction and optional extraction through a model that accepts file input.
• Scheduled processing: automatic collection synchronization and a queue worker for text extraction, chunking, embedding and removal, with retries and configurable processing limits.
• Document-aware agents: bundled agents for a document or the current browser selection, with tools to find and read documents. Providers, models and credentials are configurable.
• Access control: retrieval and agent tools enforce ADM document permissions and exclude inaccessible, trashed and archived documents.

Install the AI Pack after the base product. See AI Pack installation, RAG configuration and Agent tools.

Changed

• Document names preserve spaces and case. Upload and rename no longer replace spaces with underscores; rename no longer lowercases the name. Leading and trailing spaces are trimmed. Existing names are not automatically changed.
• Documents retain their names in the trash. Uniqueness is enforced for live documents, allowing trashed documents with the same name to coexist. Restore resolves name conflicts.
• Folder deletion applies to the whole subtree. Trash state now distinguishes the item explicitly deleted from its descendants, so restoring a folder restores its contents together.
• Text extraction has a retry budget. Files that repeatedly fail extraction are marked failed, release their queued BLOB copy and can be explicitly requeued.
• Generated document and related IDs fit within 32 decimal digits. Existing IDs are retained.
• Bundled dependencies: APEX Office Print and APEX Office Edit updated to 26.1.

Security

• Share-link passwords use salted PBKDF2-HMAC-SHA512 hashes. Existing password hashes upgrade when a link is successfully opened with its password.
• Failed share-password attempts are recorded for lockout without holding database sessions in a deliberate sleep.
• Permission checks strengthened for file access, sharing, revocation, comments, token creation and administrative operations.
• Share and embed access respects asset scope, expiry, archive and trash state. Moving or deleting assets invalidates affected path-based tokens and shares where required.
• Sensitive credential settings are restricted to administrators through the settings API.
• Session context is cleared before establishing a new identity.
• Name validation, escaped path matching and sanitized Oracle Text queries protect operations involving user-supplied names and search terms.

Fixed

• Trash and restore inconsistencies involving nested folders, duplicate names, legal holds, retention rules and permanently deleted original folders.
• Share links created without a password incorrectly requiring one.
• Comment operations trusting caller-supplied document and version identifiers.
• MIME-type detection for filenames containing multiple dots and handling of attachments with incorrectly reported MIME types.
• Tag duplication, blank tag names and facet parsing that consumed parts of tag values.
• Audit entries lost through truncation or incorrect session attribution, including embed access.
• Office editor/PDF refresh after saving and construction of the Office Edit REST URL.
• Abandoned embed tokens left by background operations.
• Fresh-install compilation of the bundled licensing package. Dependency cleanup now preserves packages when AOP synonyms or their targets are incomplete.

Errors

Compatibility change for custom integrations

ADM error codes and messages have changed. Code that catches numeric error codes or matches message text must be reviewed before upgrading. Use the named exceptions and category helpers in adm_error; the new AI Pack provides its own errors through adm_ai_error.

Custom hook errors should use the reserved -20700 to -20999 range.

See Error handling for the exception names and category helpers. Review custom integrations against the strengthened permission checks and name validation as well.

Performance and documentation

• Improved hierarchy traversal and access-control query filtering, with additional indexes on sharing and membership tables.
• Expanded PL/SQL and browser test coverage, including installation and migration checks.
• Updated user, administration and integration guides, generated API and view references, and documentation for the new AI Pack.

Upgrading from 0.1.6

Follow the upgrade guide in order. The base upgrade uses 26.1_pre.sql, install.sql, 26.1_post.sql and 26.1_deps_post.sql. Review any migration warnings about existing names or data.

To add the new AI Pack, follow its installation guide after upgrading the base product. The AI migration scripts are for installations that already contain a development version of the pack.