# Changelog

## Score

- CAI 41 → 42 (+1.8)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.

## Lenses

- Code Health 59 → 54 (-5.0)
- Architecture 100 → 100 (+0.0)
- Maturity 47 → 51 (+4.8)
- Readiness 21 → 23 (+2.6)
- Security 76 → 76 (+0.0)

## Resolved (4)

- Form1.SiticoneButton_set_num_ctx_Click (cognitive 18) (Ollama_Desktop/Form1.vb)
- Form1.render_response (cognitive 20) (Ollama_Desktop/Form1.vb)
- Members sharing a duplicated core (4 members, 50+ identical tokens) (Ollama_Desktop/Form1.vb)
- Vulnerable: SSH.NET

## New (16)

- Duplicated block (8 lines × 2) (Ollama_Desktop/Form1.vb)
- EmptyCatchBlock (Ollama_Desktop/Form1.vb)
- EmptyCatchBlock (Ollama_Desktop/Form1.vb)
- EmptyCatchBlock (Ollama_Desktop/Form1.vb)
- Form1.CheckActiveSwitchesAgainstModelCapabilities (cognitive 43) (Ollama_Desktop/Form1.vb)
- Form1.CheckActiveSwitchesAgainstModelCapabilities (cyclomatic 37) (Ollama_Desktop/Form1.vb)
- Form1.ExclusiveToggleSwitches_CheckedChanged (cognitive 25) (Ollama_Desktop/Form1.vb)
- Form1.ExclusiveToggleSwitches_CheckedChanged (cyclomatic 22) (Ollama_Desktop/Form1.vb)
- Form1.GetResponseMarkdown (cognitive 50) (Ollama_Desktop/Form1.vb)
- Form1.GetResponseMarkdown (cyclomatic 32) (Ollama_Desktop/Form1.vb)
- Form1.RenderChatHistoryUpTo (cognitive 24) (Ollama_Desktop/Form1.vb)
- Form1.RenderChatHistoryUpTo (cyclomatic 20) (Ollama_Desktop/Form1.vb)
- Form1.SiticoneDropdown_calc_ctx_SelectedIndexChanged (cognitive 17) (Ollama_Desktop/Form1.vb)
- Form1.SiticoneDropdown_model_SelectedIndexChanged (cyclomatic 22) (Ollama_Desktop/Form1.vb)
- Form1.SiticoneToggleSwitch_use_num_ctx_CheckedChanged (cognitive 16) (Ollama_Desktop/Form1.vb)
- Members sharing a duplicated core (4 members, 50+ identical tokens) (Ollama_Desktop/Form1.vb)

## Changes since last survey

- 5 commits — 5 feature/other, 0 fixes

## By area

- Ollama_Desktop/Form1.Designer.vb — 3 commits
- Ollama_Desktop/Form1.vb — 1 commit
- screenshots/code_block.png — 1 commit

## Notable commits

- change: V 1.6.4.35 - Unified Multimodal Payload: Combined audio and image inputs into the standard "images" payload array for /api/generate requests, ensuring seamless processing by multimodal models. - Optimized Request Logging: Updated the JSON request preview parser to cleanly truncate large base64 media and audio strings, preventing UI lag in the Request JSON view. - Model Capability Flag Validation: Introduced intelligent evaluation and validation of model flags (completion, vision, audio, diffusion, tools, thinking, rerank, decision, insert) to verify active toggles and media attachments against model capabilities, warning users when features are unsupported. - add nemotron-3.5-bolt to models_config.json - add flint to models_config.json
- change: V 1.6.5.36 - Simplified models_config.json: Removed redundant capability flags; model features are now dynamically inferred from Ollama API metadata, keeping configuration limited to custom descriptions. - Native Text-to-Speech (TTS) Integration: Added full support for speech-capable models (e.g., lfm3-audio, lfm3-omni, phonon) via Ollama's /v1/audio/speech endpoint. Prompts sent to models with the speech capability automatically generate WAV audio instead of text, with available voice profiles dynamically fetched from /api/show into a new voice selector dropdown. Generated speech is rendered with an embedded HTML5 audio player directly inside the WebView2 response panel. - add lfm3-audio to models_config.json - add lfm3-omni to models_config.json - add phonon to models_config.json
- change: V 1.6.6.37 - Save & Load Chats (*.chat.json): Export complete conversations to *.chat.json files and reload them at any time with full session recovery. - Instant Branching: Navigate back to any previous response and continue the conversation from that exact point. - Smart History Browsing: Step back and forth through your chat while the conversation view, token counts, and metrics update dynamically for each turn. - Step-Specific Code Extraction: Code blocks, syntax highlighting, and execution tools now automatically update to match the currently selected turn. - Direct Context Size Toggle (num_ctx): Easily enable or disable the custom context window via a dedicated toggle switch, immediately applying and syncing the selected token size without requiring a confirmation button.
- change: V 1.6.6.37 - Save & Load Chats (*.chat.json): Export complete conversations to *.chat.json files and reload them at any time with full session recovery. - Instant Branching: Navigate back to any previous response and continue the conversation from that exact point. - Smart History Browsing: Step back and forth through your chat while the conversation view, token counts, and metrics update dynamically for each turn. - Step-Specific Code Extraction: Code blocks, syntax highlighting, and execution tools now automatically update to match the currently selected turn. - Direct Context Size Toggle (num_ctx): Easily enable or disable the custom context window via a dedicated toggle switch, immediately applying and syncing the selected token size without requiring a confirmation button.
- change: V 1.6.7.38 - Model-Aware Context Preservation: Previous context tokens are now strictly bound to the model that generated them. Context is no longer discarded when switching models, allowing you to return to the original model later without losing conversation state. - Visual Context Status Indicator: The navigation clear button dynamically reflects the current context validity via its background color, or mismatched with the current model.
