# Changelog

## Score

- CAI 44 → 41 (-3.3)
- Rubric changed (rubric-2026.10.1 → rubric-2026.10.5) — scores are not directly comparable.

## Lenses

- Code Health 65 → 59 (-5.6)
- Architecture 100 → 100 (-0.0)
- Maturity 49 → 47 (-2.5)
- Readiness 21 → 21 (+0.0)
- Security 100 → 76 (-23.3)

## Resolved (3)

- Duplicated block (10 lines × 2) (Ollama_Desktop/Form1.vb)
- Form1.MainAsync_alt (cognitive 21) (Ollama_Desktop/Form1.vb)
- Form1.MainAsync_alt (cyclomatic 17) (Ollama_Desktop/Form1.vb)

## New (32)

- Duplicated block (8 lines × 2) (Ollama_Desktop/Form1.vb)
- EmptyCatchBlock (Ollama_Desktop/Form1.vb)
- EmptyCatchBlock (Ollama_Desktop/FrmOllamaPs.vb)
- EmptyCatchBlock (Ollama_Desktop/FrmOllamaPs.vb)
- Form1.PictureBox_vram_bar_Paint (cognitive 18) (Ollama_Desktop/Form1.vb)
- Form1.SiticoneButton_set_num_ctx_Click (cognitive 18) (Ollama_Desktop/Form1.vb)
- Form1.SiticoneDropdown_model_SelectedIndexChanged (cognitive 18) (Ollama_Desktop/Form1.vb)
- Form1.SyncCalcCtxDropdownFromDgv (cognitive 18) (Ollama_Desktop/Form1.vb)
- Form1.TryApplyContextLengthFromJson (cognitive 17) (Ollama_Desktop/Form1.vb)
- Form1.render_response (cognitive 20) (Ollama_Desktop/Form1.vb)
- FrmOllamaPs.LoadOllamaPsDataAsync (cognitive 77) (Ollama_Desktop/FrmOllamaPs.vb)
- FrmOllamaPs.LoadOllamaPsDataAsync (cyclomatic 35) (Ollama_Desktop/FrmOllamaPs.vb)
- High secret: WD-SECRET-0005 (Ollama_Desktop/App.config)
- High secret: WD-SECRET-0005 (Ollama_Desktop/App.config)
- High secret: WD-SECRET-0005 (Ollama_Desktop/settings/Tools-Parameter.json)
- High secret: WD-SECRET-0005 (Ollama_Desktop/settings/default_DE.tools.json)
- High secret: WD-SECRET-0005 (Ollama_Desktop/settings/default_GB.tools.json)
- High secret: WD-SECRET-0005 (Ollama_Desktop/settings/default_tools_parameter.json)
- High secret: WD-SECRET-0005 (Ollama_Desktop/settings/default_tools_parameter_neu.json)
- Hotspot: Ollama_Desktop/Form1.vb (Ollama_Desktop/Form1.vb)
- …and 12 more

## Changes since last survey

- 15 commits — 14 feature/other, 1 fixes

## By area

- (root) — 8 commits
- (repo) — 2 commits
- Ollama_Desktop/My Project — 2 commits
- Ollama_Desktop/Form1.Designer.vb — 1 commit
- Ollama_Desktop/FrmOllamaPs.Designer.vb — 1 commit
- Ollama_Desktop/settings — 1 commit

## Notable commits

- fix: V 1.6.3.34 - Bugfix: Resolved an issue where model expiration timestamps were displayed in raw UTC instead of the user's local timezone. - Displays active models loaded in memory, including dynamically allocated memory usage and GPU/VRAM offload percentage. - Fetches actual model file sizes independently via /api/tags to ensure visibility across third-party clients and remote sessions. - Shows active context length (context_length) and local expiration timestamps.
- change: Delete README.de.md
- change: Delete README.md
- change: Merge branch 'master' of https://github.com/barni007-pro/ollama_desktop_client
- change: Merge branch 'master' of https://github.com/barni007-pro/ollama_desktop_client
- change: Update README with new features and enhancements
- change: Update README with new features and tools
- change: Update gitignore
- change: V 1.6.0.31 - Removal of Chat Mode from the application The chat functionality has been fully integrated into Generate Mode. This opens up many more possibilities (e.g. vision and audio are now supported while chatting in Generate Mode) - The application now supports Decesion Models. These models can be used to make decisions based on the given question JSON and user prompt. - User prompts are now displayed as speech bubbles in the HTML output. - When context memory is enabled ("Use Last Context"), the HTML output now contains the full conversation history. - add tev1 to models_config.json - add embeddinggemma-2 to models_config.json - add nimble to models_config.json - add clef-flash to models_config.json - add clef to models_config.json - add laya to models_config.json - add qwen3.8 to models_config.json - add ornith-1.5 to models_config.json - add granite4.2 to models_config.json - add medgemma to models_config.json - add medgemma1.5 to models_config.json
- change: V 1.6.1.32 - Implemented a universal and highly accurate VRAM Calculator based on deep GGUF tensor metadata (`/api/show`): Architecture-aware KV cache calculation supporting hybrid models (Gemma 3/4 sliding windows and Qwen 3.5 SSM/recurrent layers). Dynamic compute-graph and attention-workspace memory allocation. - Added a configurable and persistent VRAM Base system overhead setting (`vram_base`) supporting headless systems, standard desktops, and multi-monitor setups. - Completely overhauled the VRAM usage visualization: - Strictly separates constant model weights and base load from dynamic context usage. - Provides precise text overlays and visual overflow warnings for system RAM offload. - Added bidirectional synchronization between the context size dropdown (`calc_ctx`) and the `num_ctx` parameter table (`dgv_option`). - Expanded drop-down lists for VRAM capacities (up to 1024 GB) and fine-grained context window sizes (up to 2M tokens). - add codestral to models_config.json - add devstral to models_config.json - add nemotron-3.5-lightning to models_config.json - add qwen3.6 to models_config.json - add bge-m3 to models_config.json - add bonsai-2 to models_config.json - add d1 to models_config.json - add pplx-decider to models_config.json - add d1-omni to models_config.json
- change: V 1.6.1.32 - Implemented a universal and highly accurate VRAM Calculator based on deep GGUF tensor metadata (`/api/show`): Architecture-aware KV cache calculation supporting hybrid models (Gemma 3/4 sliding windows and Qwen 3.5 SSM/recurrent layers). Dynamic compute-graph and attention-workspace memory allocation. - Added a configurable and persistent VRAM Base system overhead setting (`vram_base`) supporting headless systems, standard desktops, and multi-monitor setups. - Completely overhauled the VRAM usage visualization: - Strictly separates constant model weights and base load from dynamic context usage. - Provides precise text overlays and visual overflow warnings for system RAM offload. - Added bidirectional synchronization between the context size dropdown (`calc_ctx`) and the `num_ctx` parameter table (`dgv_option`). - Expanded drop-down lists for VRAM capacities (up to 1024 GB) and fine-grained context window sizes (up to 2M tokens). - add codestral to models_config.json - add devstral to models_config.json - add nemotron-3.5-lightning to models_config.json - add qwen3.6 to models_config.json - add bge-m3 to models_config.json - add bonsai-2 to models_config.json - add d1 to models_config.json - add pplx-decider to models_config.json - add d1-omni to models_config.json
- change: V 1.6.1.32 - Implemented a universal and highly accurate VRAM Calculator based on deep GGUF tensor metadata (`/api/show`): Architecture-aware KV cache calculation supporting hybrid models (Gemma 3/4 sliding windows and Qwen 3.5 SSM/recurrent layers). Dynamic compute-graph and attention-workspace memory allocation. - Added a configurable and persistent VRAM Base system overhead setting (`vram_base`) supporting headless systems, standard desktops, and multi-monitor setups. - Completely overhauled the VRAM usage visualization: - Strictly separates constant model weights and base load from dynamic context usage. - Provides precise text overlays and visual overflow warnings for system RAM offload. - Added bidirectional synchronization between the context size dropdown (`calc_ctx`) and the `num_ctx` parameter table (`dgv_option`). - Expanded drop-down lists for VRAM capacities (up to 1024 GB) and fine-grained context window sizes (up to 2M tokens). - add codestral to models_config.json - add devstral to models_config.json - add nemotron-3.5-lightning to models_config.json - add qwen3.6 to models_config.json - add bge-m3 to models_config.json - add bonsai-2 to models_config.json - add d1 to models_config.json - add pplx-decider to models_config.json - add d1-omni to models_config.json
- change: V 1.6.1.32 - Implemented a universal and highly accurate VRAM Calculator based on deep GGUF tensor metadata (`/api/show`): Architecture-aware KV cache calculation supporting hybrid models (Gemma 3/4 sliding windows and Qwen 3.5 SSM/recurrent layers). Dynamic compute-graph and attention-workspace memory allocation. - Added a configurable and persistent VRAM Base system overhead setting (`vram_base`) supporting headless systems, standard desktops, and multi-monitor setups. - Completely overhauled the VRAM usage visualization: - Strictly separates constant model weights and base load from dynamic context usage. - Provides precise text overlays and visual overflow warnings for system RAM offload. - Added bidirectional synchronization between the context size dropdown (`calc_ctx`) and the `num_ctx` parameter table (`dgv_option`). - Expanded drop-down lists for VRAM capacities (up to 1024 GB) and fine-grained context window sizes (up to 2M tokens). - add codestral to models_config.json - add devstral to models_config.json - add nemotron-3.5-lightning to models_config.json - add qwen3.6 to models_config.json - add bge-m3 to models_config.json - add bonsai-2 to models_config.json - add d1 to models_config.json - add pplx-decider to models_config.json - add d1-omni to models_config.json
- change: V 1.6.1.32 - Implemented a universal and highly accurate VRAM Calculator based on deep GGUF tensor metadata (`/api/show`): Architecture-aware KV cache calculation supporting hybrid models (Gemma 3/4 sliding windows and Qwen 3.5 SSM/recurrent layers). Dynamic compute-graph and attention-workspace memory allocation. - Added a configurable and persistent VRAM Base system overhead setting (`vram_base`) supporting headless systems, standard desktops, and multi-monitor setups. - Completely overhauled the VRAM usage visualization: - Strictly separates constant model weights and base load from dynamic context usage. - Provides precise text overlays and visual overflow warnings for system RAM offload. - Added bidirectional synchronization between the context size dropdown (`calc_ctx`) and the `num_ctx` parameter table (`dgv_option`). - Expanded drop-down lists for VRAM capacities (up to 1024 GB) and fine-grained context window sizes (up to 2M tokens). - add codestral to models_config.json - add devstral to models_config.json - add nemotron-3.5-lightning to models_config.json - add qwen3.6 to models_config.json - add bge-m3 to models_config.json - add bonsai-2 to models_config.json - add d1 to models_config.json - add pplx-decider to models_config.json - add d1-omni to models_config.json
- change: V 1.6.2.33 - Automatically detects and displays model capabilities (such as vision, tools, thinking, embedding, and more) directly from Ollama's model metadata and API. - Dynamic context window size (context_length) parsing and automatic chip status updates for selected models. - Added a new non-modal Ollama Processes window (/api/ps) accessible from the Ollama-LLM panel. - Real-time tabular monitoring of active models loaded in memory, including VRAM allocation status and expiration timers. - edit d1-omni in models_config.json - edit bonsai-2 in models_config.json