diff --git a/app/MindWork AI Studio/Assistants/ERI/AssistantERI.razor b/app/MindWork AI Studio/Assistants/ERI/AssistantERI.razor index 19ad57c..0677cbc 100644 --- a/app/MindWork AI Studio/Assistants/ERI/AssistantERI.razor +++ b/app/MindWork AI Studio/Assistants/ERI/AssistantERI.razor @@ -3,12 +3,12 @@ @using MudExtensions @inherits AssistantBaseCore - + You can imagine it like this: Hypothetically, when Wikipedia implemented the ERI, it would vectorize all pages using an embedding method. All of Wikipedia’s data would remain with Wikipedia, including the vector database (decentralized approach). Then, any AI Studio user could add Wikipedia as a data source to significantly reduce the hallucination of the LLM in knowledge questions. - + Related links: @@ -25,10 +25,10 @@ ERI server presets - + Here you have the option to save different configurations for various ERI servers and switch between them. This is useful if you are responsible for multiple ERI servers. - + @if(this.SettingsManager.ConfigurationData.ERI.ERIServers.Count is 0) { @@ -59,26 +59,26 @@ else @if(this.AreServerPresetsBlocked) { - + Hint: to allow this assistant to manage multiple presets, you must enable the preselection of values in the settings. - + } Auto save - + The ERI specification will change over time. You probably want to keep your ERI server up to date. This means you might want to regenerate the code for your ERI server. To avoid having to make all inputs each time, all your inputs and decisions can be automatically saved. Would you like this? - + @if(this.AreServerPresetsBlocked) { - + Hint: to allow this assistant to automatically save your changes, you must enable the preselection of values in the settings. - + } @@ -204,18 +204,18 @@ else Embedding settings - + You will likely use one or more embedding methods to encode the meaning of your data into a typically high-dimensional vector space. In this case, you will use a vector database to store and search these vectors (called embeddings). However, you don't have to use embedding methods. When your retrieval method works without any embedding, you can ignore this section. An example: You store files on a file server, and your retrieval method works exclusively with file names in the file system, so you don't need embeddings. - + - + You can specify more than one embedding method. This can be useful when you want to use different embeddings for different queries or data types. For example, one embedding for texts, another for images, and a third for videos, etc. - + @if (!this.IsNoneERIServerSelected) { @@ -258,12 +258,12 @@ else Data retrieval settings - + For your ERI server, you need to retrieve data that matches a chat or prompt in some way. We call this the retrieval process. You must describe at least one such process. You may offer several retrieval processes from which users can choose. This allows you to test with beta users which process works better. Or you might generally want to give users the choice so they can select the process that best suits their circumstances. - + @if (!this.IsNoneERIServerSelected) { @@ -299,13 +299,13 @@ else Add Retrieval Process - + You can integrate additional libraries. Perhaps you want to evaluate the prompts in advance using a machine learning method or analyze them with a text mining approach? Or maybe you want to preprocess images in the prompts? For such advanced scenarios, you can specify which libraries you want to use here. It's best to describe which library you want to integrate for which purpose. This way, the LLM that writes the ERI server for you can try to use these libraries effectively. This should result in less rework being necessary. If you don't know the necessary libraries, you can instead attempt to describe the intended use. The LLM can then attempt to choose suitable libraries. However, hallucinations can occur, and fictional libraries might be selected. - + @@ -313,17 +313,17 @@ else Provider selection for generation - + The task of writing the ERI server for you is very complex. Therefore, a very powerful LLM is needed to successfully accomplish this task. Small local models will probably not be sufficient. Instead, try using a large cloud-based or a large self-hosted model. - + - + Important: The LLM may need to generate many files. This reaches the request limit of most providers. Typically, only a certain number of requests can be made per minute, and only a maximum number of tokens can be generated per minute. AI Studio automatically considers this. However, generating all the files takes a certain amount of time. Local or self-hosted models may work without these limitations and can generate responses faster. AI Studio dynamically adapts its behavior and always tries to achieve the fastest possible data processing. - + @@ -331,19 +331,19 @@ else Write code to file system - + AI Studio can save the generated code to the file system. You can select a base folder for this. AI Studio ensures that no files are created outside of this base folder. Furthermore, we recommend that you create a Git repository in this folder. This way, you can see what changes the AI has made in which files. - + - + When you rebuild / re-generate the ERI server code, AI Studio proceeds as follows: All files generated last time will be deleted. All other files you have created remain. Then, the AI generates the new files. But beware: It may happen that the AI generates a file this time that you manually created last time. In this case, your manually created file will then be overwritten. Therefore, you should always create a Git repository and commit or revert all changes before using this assistant. With a diff visualization, you can immediately see where the AI has made changes. It is best to use an IDE suitable for your selected language for this purpose. - + diff --git a/app/MindWork AI Studio/wwwroot/changelog/v0.9.29.md b/app/MindWork AI Studio/wwwroot/changelog/v0.9.29.md index 468beec..109387e 100644 --- a/app/MindWork AI Studio/wwwroot/changelog/v0.9.29.md +++ b/app/MindWork AI Studio/wwwroot/changelog/v0.9.29.md @@ -7,4 +7,5 @@ - Added a generic RAG process to integrate possibly any data in your chats. Although the generic RAG process is now implemented, the retrieval part is working only with external data sources using the ERI interface. That means that you could integrate your company's data from the corporate network by now. The retrieval process for your local data is still under development and will take several weeks to be released. In order to use data through ERI, you (or your company) have to develop an ERI server. You might use the ERI server assistant within AI Studio to do so. This preview feature is hidden behind the RAG feature flag. - Improved confidence card for small spaces. - Improved data security by enforcing provider filtering based on the chosen confidence level. To ensure this in the future, source code analyzers have been added to warn developers about insecure code. +- Improved the readability of descriptions inside the ERI server assistant. - Fixed a bug in which 'APP_SETTINGS' appeared as a valid destination in the "send to" menu. \ No newline at end of file