Yellx - Pro Config Sharing

[tardy quartz]

After simplification, automata is the overall headline, state is the subheadline, and there are more items below.

• The simplest app may only need one state to complete an interaction flow from inputs (user input) > tasks (component execution tasks) > outputs (output results) > render (rendering results on the screen).
• Transitions can be used to jump between different states and are used in the creation of multi-state apps. (If the state to jump to is not specified, it will remain in the original state by default).

On the basis of understanding the above content, explain with specific examples.

[tardy quartz]

App 1 - Mail Assistant: 📧

• Initial version:

Simplify and understand the following process, you can refer to the comments in the json file, pay attention to the format and the connection of variables. The basic operation is to use double brackets {{}} to reference variables that have been input or output before.

• The home state only renders the introduction text and the jump button. Clicking the button jumps to the input state.
• Based on the user's input, the input state generates text using GPT, and the final email draft is rendered on the screen.

To clarify, there can also be transitions and other content under the "automata" (main title), which can be referenced in different states. Context can be used to store values that multiple states need to use, and can be used to add memory to GPT later.

Based on the initial version, I added the function of rewriting emails, which means there is now a "rewrite_email_state". Friends who want to try it can complete it on the basis of the initial version and use it as a reference. Also, please understand the "context" and "transitions" under the automata, which can be referenced in different states.

• Bot Link: https://app.myshell.ai/bot/MZvaui/1711848718
• New version of JSON file:

App 2 Text Tools Pro📄

This app is a simple text processing toolbox. The home page has six different buttons that correspond to six different states to complete different tasks. If you understand the content above, the production of this app is actually very simple.

• Bot Link: https://app.myshell.ai/bot/bQVnY3/1711694375
• Json:

In this app, I also used the MeloTTS widget to implement the function of text-to-speech. How do I reference the official widget?
Workshop > Widget Center > Find the widget you want to use > Three dots in the upper right corner > Copy ID/Copy pro config.

Please note that the official widget configurations provided as examples may not be directly usable. Some widgets have specific requirements for input values. For example:

• When selecting a language for MeloTTS, only formats such as en_us and en_br are recognized, and inputs such as "English" will result in an error.
• When selecting a model for the StableDiffusion widget, a string of numerical IDs rather than names is required, which can be obtained from a link on the Civitai website. This will not be expanded upon here, but can be modified based on error prompts or code shared by others.

App3 Talkshell🗨️

Botlink: https://app.myshell.ai/bot/YvYv2e/1711643317
This app is designed according to the process of Tinder and combines the abilities of LLM and SD.

The first two apps' LLMs have no memory, while in this app, under different tasks, LLMs have one-time memory, remember the last memory, or have long-term memory.

• I wrote in the system prompt for generating simple character information, Name and age are different from {{context.random_sample_info}} to make the GPT generate different results from the last time, achieving a similar effect of random generation, so as not to generate characters similar to the previous ones.

• For the LLM used in chatting, it relies on "context.memory": "{{[...context.memory, {'user': user_message}, {'assistant': reply}]}}" to achieve long-term memory.

• The app's voice is selectable, corresponding to preset male/female voices based "context.tts_widget_url": "{{context.target_gender == 'Male' ? 'https://app.myshell.ai/widget/UVNFVb' : 'https://app.myshell.ai/widget/e2E7Zj'}}",

App 4: Language Parner Zoe

• Bot Link：https://app.myshell.ai/bot/uYVR3i/1711694987
  This app uses the multilingual abilities of LLM and Melotts to create a multilingual language partner. Users can choose their native language and the language they want to learn to communicate and improve their expression abilities.

• The Chat page performs three main tasks: LLM 1 - generating Zoe's reply, MeloTTS voice outputting Zoe's reply, LLM 2 - generating sentence structure analysis, translation, and reply suggestions based on Zoe's reply.
• The three results are rendered uniformly.

• The app uses dictionary mapping to convert English (American) and other content into en_us format for MeloTTS language recognition.
• The output format for analyzing content uses html <detail> to make each item collapsed by default, without affecting the normal chatting experience.
  Please refer to the Json file below for specific content.

The following new app content will also be shared here. Thank you for reading.

[deep owl]

okyellx has been warned

Reason: Duplicated text

okyellx has been warned

Reason: Duplicated text

[tardy quartz]

MEET6 !!! It deleted the content I edited.