0.10.0-beta.0 - Reset beta baseline
Reset beta baseline
- Resets the connection status and retry schedule together when a connection is disabled.
- Supports direct Ollama responses without an optional thinking phase and local request timeouts up to ten minutes while preserving existing profile settings and cloud timeout limits.
- Clears previous attempt errors when durable AI jobs resume and when profile validation succeeds.
- Lets administrators delete AI profiles with version checks, removes stored credentials, prevents configured profiles from reappearing and preserves historical results.
- Adds a permission-scoped data catalog and generated SQL/MongoDB analysis across internal MCon data and approved SQL Server, PostgreSQL, MySQL/MariaDB and MongoDB sources.
- Defaults conversations to read-only; explicit per-chat grants enable reviewed changes or automatic authorized row changes without granting machine control.
- Binds write previews to unique keys, source revisions, data snapshots and the current sign-in; prevents blind replay after uncertain execution.
- Adds configurable machine semantics, exact downtime/fault counting, source evidence, bounded analysis runs, cancellation and authenticated continuation.
- Reads additional module entitlements from a versioned, signed extension while retaining the original payload for online approval.
- Ignores unknown future module flags without granting unsupported permissions.
- Preserves strict signature verification, machine binding, core entitlements and validity checks.
- Centralizes ERP credentials, reviewed imports, machine assignments and durable production feedback in MES while preserving existing direct integrations until an explicit, validated cutover.
- Adds machine-scoped operation selection with transactional start, pause, resume and finish actions, revision checks and repeatable execution receipts.
- Tracks quantities per operation and run, handles physical counter resets and manual corrections, and avoids counting the same finished parts at multiple machines.
- Preserves ERP destination and tenant identity for queued feedback, including idle-machine events, and reconciles uncertain delivery results without repeating production actions.
- Supports encrypted Bearer, Basic and API-key credentials for generic ERP connections alongside SelectLine authentication.
- Adds revision-checked, idempotent deletion of unused editable records, with permission checks, audit history and protection for booked or referenced data.
- Prevents new assignments and attendance bookings for inactive employees and unavailable records while preserving historical corrections and cancellations.
- Serializes deletion, deactivation and assignment transactions to prevent dangling references.
- Adds personal ChatGPT OAuth connections with PKCE, verified OpenID Connect identity, encrypted credentials, rotating refresh and remote revocation.
- Supports eligible ChatGPT plan requests with user-granted models and existing cloud consent, permissions and reviewed actions.
- Requires an OpenAI-approved client registration and an exact HTTPS callback before live ChatGPT sign-in can be enabled.
- Shares the verified online approval client with Ingestion while retaining the API's own version-bound approval and cache.
- Observes newer license status revisions recorded by the Manager and Ingestion so older approvals cannot override known license decisions.
- Preserves signature, nonce, installation, version, offline-period and replay checks across independently updated components.
- Stores sidebar group order and visibility with the user's existing navigation preferences.
- Validates group identifiers and rejects duplicate entries before saving preferences.
- Includes group preferences in authorized session responses and live user updates while preserving restricted-role projections.
- Preserves repeatable saves and native execution through idempotent requests, field validation, revisions and transaction history.
- Adds reviewed ERP record adoption, tenant-scoped stable mappings, SelectLine REST capability discovery and a durable production feedback queue with uncertain-result reconciliation.
- Exposes authorized production fields to graphical linkings and connects native production orders to machine terminals.
- Adds the general /ai/v1 assistant with permission-filtered native actions, stored previews, explicit confirmation, revision checks, protected credentials and partial-execution results.
- Supports an independent ai entitlement while retaining existing mes_ai licenses and legacy operations AI endpoints.
- Adds an authorized MES overview with exact server-side totals and bounded lists for orders, personal tasks, approvals, quality and maintenance.
- Applies optional machine and resource context before workspace list limits while preserving project, connection and module access checks.
- Validates additive product-catalog and advisory pricing metadata in license inquiries; existing signed license formats remain unchanged.
- Enforces separate MES module entitlements and adds workforce, shift-template and finite-capacity planning services.
- Adds attendance accounts, personnel privacy boundaries, production metrics and master-data transfer validation.
- Handles missing revisions consistently when hashing new master-data imports for repeatable, atomic submissions.
- Keeps MES specialist roles separate from administrative access and preserves fixed work already in progress during planning.
- Serves MES prices from the existing CDN pricing configuration and accepts older signed licenses without rewriting their payloads.
- Passes real-replica-set MES and operations integration tests. Extended endurance testing and customer process/model acceptance remain pending.
- Adds discrete and batch production orders, recipe versions, material movements, lot genealogy, quality inspections, maintenance and resource planning.
- Connects projects, tasks, dependencies, milestones, approvals and department templates in a permission-aware operations workspace.
- Provides centralized shift metrics, OEE, downtime analysis, production comparisons, energy reports and robust signal statistics.
- Adds a durable automation worker with restart recovery, deduplicated actions, scheduled triggers and a downtime workflow spanning production, quality and maintenance.
- Integrates OpenAI Responses, Anthropic tool use and local Ollama models through approved profiles, explicit processing policies, source references and editable proposals.
- Adds permission-aware knowledge search, model evaluation and release controls, encrypted provider credentials and audited AI execution history.
- Extends ERP integration with idempotent draft imports and an acknowledged event outbox; offers optional read-only MCP tools.
- Corrects linking validation and preserves personal and shared-team task visibility.
- Beta validation includes automated reference workflows and simulated AI providers. The 24-hour endurance run requires a rerun after the isolated test database aborted with an out-of-memory error; customer process/model acceptance also remains pending.