openAIApps

2026-10-02 — Developer Tools for local .NET workspaces

The Responses experience includes optional Developer Tools that let an OpenAI model work with a local C#/.NET workspace through constrained, auditable tools. The user chooses the workspace and explicitly enables each capability; the model is not given an unrestricted shell or arbitrary filesystem access.

Workspace workflow

• For an existing repository or solution, select its folder as the Workspace root.
• To create a project, initially select the parent folder where the new solution should be created.
• When creation succeeds, the app automatically changes the Workspace root to the newly created solution folder. Later reads, edits, builds, cleans, rebuilds, and runs use that new root.
• Tool-call logs record the workspace root used for each operation, making the creation-parent and created-project transition clear in saved sessions.

Available capabilities

Capability	What it does
Inspect workspace files	Search project text, read selected file ranges, and list files within the configured workspace root.
Change files	Create/write project files or make exact text replacements when the corresponding write capability is enabled.
Create .NET projects	Create a constrained solution and one project in a new direct child folder using approved dotnet new templates.
Build, clean, and rebuild	Run controlled dotnet build, dotnet clean, or clean-then-build operations for a .csproj, .sln, or .slnx target.
Run in Debug	Start an already-built .NET project in the Debug configuration and track the launched process.
Manage launched processes	List Developer Tools-launched processes, read their bounded output, and stop an individual tracked process.

Safety and permissions

Developer Tools remain disabled until enabled by the user. Operations are constrained to the selected Workspace root and supported file/solution types.

• File writes and replacements can require per-operation approval. Users may disable this approval when they want enabled write tools to apply changes automatically.
• Creating projects, building, cleaning, rebuilding, running, and stopping launched processes are separate execution capabilities. They can require their own approval, independently of file-change approval.
• Project creation is limited to a new direct child directory of the initial workspace root and an application-maintained allowlist of .NET templates.
• Build, clean, and rebuild accept only supported .csproj, .sln, and .slnx targets and controlled Debug or Release configurations.
• Run operations launch only tracked .NET projects through a constrained dotnet run --no-build flow; there is no generic shell, PowerShell, batch-file, or arbitrary executable tool.
• Tool-call logging continues even when the log panel is hidden. Logs are saved with the conversation and include the operation, arguments, result, and workspace root.

Note: The project creation and build/run capabilities use the .NET CLI to generate and operate on standard solutions/projects. They do not automate the Visual Studio UI, but generated .sln/.slnx solutions can be opened normally in Visual Studio.

2026-03-15 Refactored WPF UI and session state management

• Continued the WPF refactor using an MVVM-lite approach: introduced bindable panel-state objects instead of doing a full MVVM rewrite.
• Logs tab now uses LogsPanelState for filter/search state, with XAML bindings and centralized filtering.
• Responses tab now uses ResponsesPanelState for prompt/response/selected turn plus options state (model, reasoning, tools, search/image settings).
• Responses session restore was improved so logs reload conversation content and latest saved settings consistently.
• Responses history list now shows per-turn metadata (model/reasoning/tools) for better traceability.
• Video tab was reworked toward job-based sessions instead of pseudo-conversation sessions.
• Added VideoPanelState and CurrentVideoMessages to separate job/session state from the global video library list.
• Video generation/remix now always creates a new session; remix lineage is stored via SourceRemoteId.
• Database/history layer was updated to support SourceRemoteId, including SQLite schema patching for existing databases.
• General cleanup: reduced direct control manipulation, improved tab/session restore behavior, and kept code-behind focused on orchestration rather than raw UI state.

2025-11-12 1745 Added Video

• Haven't worked on this project for a while so I made use of OpenAi. Very efficient...
• Videocreation tool by OpenAI. Using the basic calls, ++ reference image. I does the job
• Did not implement the remix-module. Maybe later
• Dalle-E is not really Dall-E anymore since I hardcoded in the model-name (gpt-image-1). There are restrictions, but there are restrictions on every image-model (they are different)

2024-05-20 1651 Added Vision

• Added OpenAI Vision using Azure OpenAI API
• Reason for this was that I had problems-/or it was confusing using JSON-format/structure. Couldn't make it work, so I shifted to Nuget->Azure.AI.OpenAI v1.0.0-beta.17

2023-11-10 0810 Did minor changes/tidying up a little

• All references to GPT3.5 Turbo is removed. All objects/classes have been reduces to just GPT. The sourcefile too.
• The tab that used to have the header GPT3.5Turbo has now the name of the model used in the chat.

2023-11-07 First commit of this hobby-project. Using WPF, OpenAI API. On this date there was a couple of updates from the OpenAI community regarding GPT 4 Turbo. And a bunch of other stuff. New to me so I thought I'd try it out and therefore I commited this stuff to Github. Two models available in the GPT4 Turbo family at this time:

1. gpt-4-1106-preview - works "out of the box" in my code The latest GPT-4 model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more. Returns a maximum of 4,096 output tokens. This preview model is not yet suited for production traffic_
2. gpt-4-vision-preview - this can be interesting to try out. Ability to understand images, in addition to all other GPT-4 Turbo capabilties. Returns a maximum of 4,096 output tokens. This is a preview model version and not suited yet for production traffic

DALL-E DALL-E v3 - I will see if I can update to this version. Updated to version 3, But now I can't use EDIT or VARIATIONS..but I can see that the pics are "good".

Assistant API Haven't tried yet. Can be interesting.

TTS Will not update or do more here for the time being. I've allready tried using TTS via "Microsoft.CognitiveServices.Speech". I think this one is cooler. More voices/dialects from all over the world. Narrowed it down to English, Norwegian and Tagalog. But if I'm right OpenAI offers a way so that you do not have to use this. Which is good. See source SpeechSynthesis.cs.

GPTs Somehing to look into in the future? See their quotes below: We launched a new feature called GPTs. GPTs combine instructions, data, and capabilities into a customized version of ChatGPT We launched a new feature called GPTs. GPTs combine instructions, data, and capabilities into a customized version of ChatGPT