LLM Controller CE v1.2.0

LLM Controller CE v1.2 is a focused capability release built on the usability and stability work completed in v1.1.

Rather than another broad interface pass, this release concentrates on four practical additions: a unified attachment workflow, multimodal image understanding, OpenAI-compatible API access, and stronger model and title-generation controls. The existing chat, model-loading, benchmarking, monitoring, analytics, and administration experience remains familiar.

Highlights

• Expanded the existing file-aware conversation system with a unified attachment workflow supporting file selection, drag-and-drop, and clipboard paste.
• Added image attachments with thumbnail previews and a larger image viewer.
• Added image understanding for compatible multimodal GGUF models through llama-server MMPROJ support.
• Added projector selection and association within Model Management.
• Added clear multimodal capability visibility in model selection and loaded-model information.
• Added an OpenAI-compatible /v1 API for using the currently loaded model with compatible applications and development tools.
• Added API-key authentication and clear validation for invalid credentials, unavailable models, and requests targeting a model other than the one currently loaded.
• Made the existing automatic title-generation system more controllable through an administrator-selected title model.
• Integrated the title model into the main model lifecycle so it starts and stops as part of the same operating flow rather than functioning as a separate runtime.
• Expanded Model Management with Friendly Names, average tokens-per-second visibility, administrator notes, and capability classifications.
• Added model capability labels for General, Coding, Writing, Reasoning, Math, Agents, and Images.
• Improved model and projector display names while preserving exact source filenames and removing the previous lossy filename-prettifying behavior.
• Completed a focused stabilization pass across attachment handling, multimodal model launching, API validation, title-model runtime ordering, and the updated model registry paths.

Upgrade Notes

This release includes database schema changes.

• New installations use the updated install/schema.sql.
• Existing installations upgrading from v1.1 must apply install/upgrade_v1_2.sql.
• After updating the application files and database schema, restart the application service normally.
• Existing chats, users, benchmark history, analytics data, registered models, and runtime configuration are intended to remain in place.

Validation

The v1.2 release was validated through existing Python and JavaScript syntax checks, browser testing across the changed application surfaces, API authentication and error-path testing, multimodal model and projector launch testing, title-generation lifecycle testing, and a focused pre-release regression review.

Key areas reviewed included file selection, drag-and-drop, clipboard paste, image previews, MMPROJ association, loaded-model visibility, title-model startup and shutdown, API model enforcement, Model Management, and the existing chat and runtime workflows.

LLM Controller CE v1.1.0

LLM Controller CE v1.1 is a post-launch refinement release focused on usability, runtime visibility, drawer polish, mobile behavior, and release stability.

This update improves the core Community Edition experience without changing the database schema, model loading behavior, benchmark scoring, analytics calculations, authentication behavior, or core chat storage behavior.

Highlights

• Added chat sidebar search so large chat histories can be filtered quickly by title.
• Added direct chat URLs so the active chat can be opened or refreshed without losing place.
• Preserved active chat state across page refreshes.
• Improved chat sidebar row behavior so rename/delete controls are cleaner and chat titles use available space better.
• Fixed mobile model selector layout so the final models are reachable instead of being hidden behind the bottom UI.
• Added automatic Logs drawer lifecycle behavior so the log stream starts when opened and stops when closed.
• Added automatic GPU Monitor drawer lifecycle behavior so monitoring starts when opened and stops when closed.
• Preserved manual Start, Stop, Clear, and Copy controls for Logs and GPU Monitor.
• Added system RAM used/total visibility to runtime monitoring.
• Improved Analytics drawer readability, spacing, sorting indicators, table layout, and summary presentation.
• Moved Total Tokens to the top of the Analytics drawer for faster visibility.
• Fixed Analytics drawer horizontal scrolling/layout issues.
• Improved Chat History drawer organization and made export/import/clear/admin sections clearer.
• Added clearer Chat History wording around backup/import/export behavior.
• Redesigned the Benchmarks drawer layout to better support many tested models.
• Made benchmark prompts collapsible/less dominant so results are easier to review.
• Improved Best Run per Model table readability and large-model-list handling.
• Added a cleaner benchmark detail view with a Back to Results flow.
• Improved Admin drawer layout, spacing, registry toolbar, and overall scanability.
• Improved Admin model registry presentation while preserving existing model controls.
• Improved Admin Users password-policy page styling and clarity.
• Added explanations for preset password policies while preserving existing Custom policy controls.
• Updated README feature wording and screenshots for the refined v1.1 experience.
• Cleaned up stale/noisy comments across Python, HTML, CSS, and JavaScript while preserving safety-relevant comments.
• Performed a stabilization/regression pass across v1.1 surfaces.
• Fixed one regression found during stabilization.

Upgrade Notes

No database migration is required for this release.

Existing installs should preserve their current configuration, chats, benchmark data, analytics data, model registry, authentication behavior, and runtime settings.

After updating, restart the application service normally.

Validation

The v1.1 work was validated with syntax checks for touched Python and JavaScript files, browser testing across the changed app surfaces, and a focused pre-release regression review.

Key areas reviewed included chat navigation, direct chat URLs, mobile model selection, Logs and GPU Monitor lifecycle behavior, Analytics, Chat History, Benchmarks, Admin settings, Admin Users, and shared drawer/layout behavior.

LLM Controller CE v1.0.0

Initial release of LLM Controller Community Edition (CE).

LLM Controller CE is a self-hosted, local-first browser application for operating local GGUF models through a configured llama-server runtime.

Features

• Persistent chat workspace
• Reasoning-aware responses
• Managed GGUF model library
• File-aware conversations
• Administrator-controlled benchmarks
• Runtime visibility and live logs
• GPU telemetry support for compatible NVIDIA and AMD environments
• Installation workflow
• User administration and authentication controls

Supported Environments

Validated on:

• Windows
• Ubuntu Linux

Requires:

• Python 3
• MySQL
• Compatible llama-server runtime
• Local GGUF model files

Documentation

See:

• README.md
• INSTALL.md

Notes

LLM Controller CE v1.0 represents the Community Edition foundation of LLM Controller.

This release focuses on providing a stable local-first platform for managing and operating local AI models on your own hardware.

Feedback and issue reports are welcome.