I completed an AI session yesterday but did not copy the conversation to an external file. Today, I looked at this session in the History tab of the AI Manager, but only 1/4 of the beginning conversation is displayed and stuck with the spinning icon at:
In the editor tab of this file in App Connect, the AI Assistant shows no history.
Where are the conversations saved locally?
PS: While the AI was doing its thing, there were 15 AI changes along the way to accept which I did not noticed right away until the very end where I applied them. Should this be more prominent/noticeable and okay to apply at the end of the session?
The conversations are stored in a local indexdb database. The chats per editor are stored with the editor state and will be lost after the editor was closed. We are investigating on how we could improve this and store chat history per editor (file) permanent.
If you have suggestions on how to improve the current AI flow, please let us know.
I think having some indication in the tab that it was edited by AI will help.
We have a prompt in our instructions.md file to save session prompts and responses to a chat_dateStamp.json file which works really well (with an additional instruction to prompt the User to save the file chatlog_dateStamp.txt, as a failsafe, upon the response generation).
Chat Start Logging Prompt:
When a new chat session begins, ask the user:
“Would you like to save a log of this chat? (Yes/No)”
If the user answers Yes:
Create a file named chat_<dateStamp>.json (e.g., chat_2025-08-18.json).
Saving Prompts and Responses
For each user prompt and LLM response, save the data in the JSON file using the following structure:
{
"chat": [
{
"timestamp": "2025-08-18T14:32:05Z",
"prompt": "User's message text",
"response": "LLM's response text"
},
{
"timestamp": "2025-08-18T14:34:10Z",
"prompt": "Next user message",
"response": "Next LLM response"
}
]
}
Append each new prompt/response pair sequentially.
Periodic Full Chat Log Save
At defined intervals (e.g., every 10 messages or after a set duration), ask the user:
“Would you like to save the entire chat as a text log? (Yes/No)”
If the user answers Yes:
Save the full chat transcript into a plain text file named chatlog_<dateStamp>.txt.
This file should contain the conversation in chronological order, including all prompts and responses.
File Naming Convention
chat_<dateStamp>.json → structured storage of prompts/responses.
chatlog_<dateStamp>.txt → plain text transcript of the full conversation.
I assume with Act that no user interaction is required since this is the case with Server Connect files, but with App Connect files you have to review/accept changes. Either automatically save with App Connect files or add an option to `Confirm before Act"?
Please consider adding an icon/option to save task history without having to jump through hoops with placing the cursor at the top of the conversation, scroll to the bottom, copy and paste to an external editor.
When the AI makes changes to a document it will fist create a snapshot. All changes will then apply directly and file with the new changes will be saved. If you choose Keep it only removes the snapshot that was created, when you Undo it will set the document back to the snapshot.
As for saving the conversation, do you want to have the full JSON with tool calls etc. or just a text-based version with only the prompt and AI responses?
Text-based version please. Non-relevant information is removed in the external saved version.
Mainly for notes/documentation and for organizational purposes to have a clean version with relevant information only that can be searched/categorized into folders.
But also serves as a backup just in case Wappler saved these conversations partially as reported recently.
If you are on a monthly plan, you won't have access to these conversations if your subscription is not active.
