m/paliad
1
0
Fork 0
Code Issues 146 Pull Requests 8 Actions 1 Packages Projects Releases Wiki Activity

AI assistant: save outputs to user folder + tags (mini-DMS for AI artefacts) #29

New Issue
Open
opened 2026-05-09 09:20:21 +00:00 by mAi · 0 comments
mAi commented 2026-05-09 09:20:21 +00:00
Collaborator
Copy Link

Idea (m, PWA 2026-05-09 11:19)

"Da, wo ich die KI-Assistenten einbaue, möchte ich auch deren Ergebnisse teilweise speichern können. Das heißt, wir brauchen ein zusätzlich zu dem KI-Chatbot die Möglichkeit, Output der KI in einem User-Folder oder so zu speichern. Und dann eben natürlich auch abrufbar zu machen, zu verschlagworten, sonst was. Also quasi nochmal ein Mikro-Datei-Management-System. Maybe too much, aber als Idee einmal parken."

Concept

Beyond the linear chat surface, give the user a place to save individual AI outputs as durable artefacts:

  • Save action on any AI message (button: "Speichern" / pin / star)
  • User folder per user — a personal collection of saved outputs
  • Tagging / Verschlagwortung for retrieval
  • Browse / search / filter — list view, full-text search, tag filter
  • Effectively a mini DMS scoped to AI-generated content

Why this is useful

Today: any insight from a chat session lives only in the chat history. Going back to find "that one analysis Lexie did about extraterritorial claims three weeks ago" means scrolling. With a Save Folder:

  • The user curates the keepers
  • The keepers are tagged, searchable, structured
  • The chat surface stays light (a conversation, not a permanent record)

This is a familiar pattern from products like Notion AI, ChatGPT canvases, Perplexity collections — but here scoped to the project (UPC research / Patentanwalts-Workflow).

Key design questions (for the inventor phase, not this issue)

  1. Granularity: save the entire turn, or let the user select a passage?
  2. Editing: is the saved output read-only (frozen at save time) or does the user keep editing/annotating?
  3. Folder structure: flat tag-based, hierarchical (folders + sub-folders), or hybrid?
  4. Cross-references: can a saved output reference the originating chat thread? Reference other saved outputs?
  5. Sharing: per-user only, or can saved outputs be shared with collaborators / a team?
  6. Storage: own table in Supabase (saved_ai_outputs with user_id + content + metadata + tags), or piggyback on existing artefact tables?
  7. Retention / quotas: limit per plan tier, or unlimited?
  8. Export: copy to clipboard, download as Markdown/PDF, send via email — what export formats?

Out of scope (this issue)

  • Implementation — this is a parked idea. Inventor designs first.
  • Mobile UX, keyboard shortcuts, offline-mode — all later.
  • "Maybe too much" — m's own qualifier. The decision whether to ship at all is part of the inventor phase.

Cross-link

This idea applies symmetrically to youpc.org (also has AI assistants where users would benefit from the same save-and-organise pattern). Cross-tracked at yoUPC/youpc.org (separate issue, same parked status). If both projects ship this, they could reuse a common storage model — or each implements its own scoped to its data shape. To be decided.

Workflow

Park-only. When m wants to pursue:

  1. Inventor designs UX + storage model + answers the design questions above
  2. Coder implements stage 1: save + list + tag-filter
  3. Coder implements stage 2: full-text search + export
  4. Coder implements stage 3: sharing / collaboration if needed

Refs

  • m's voice via PWA 2026-05-09 11:19 (parking request)
## Idea (m, PWA 2026-05-09 11:19) > "Da, wo ich die KI-Assistenten einbaue, möchte ich auch deren Ergebnisse teilweise speichern können. Das heißt, wir brauchen ein zusätzlich zu dem KI-Chatbot die Möglichkeit, Output der KI in einem User-Folder oder so zu speichern. Und dann eben natürlich auch abrufbar zu machen, zu verschlagworten, sonst was. Also quasi nochmal ein Mikro-Datei-Management-System. Maybe too much, aber als Idee einmal parken." ## Concept Beyond the linear chat surface, give the user a place to **save individual AI outputs** as durable artefacts: - **Save action** on any AI message (button: "Speichern" / pin / star) - **User folder** per user — a personal collection of saved outputs - **Tagging / Verschlagwortung** for retrieval - **Browse / search / filter** — list view, full-text search, tag filter - Effectively a **mini DMS** scoped to AI-generated content ## Why this is useful Today: any insight from a chat session lives only in the chat history. Going back to find "that one analysis Lexie did about extraterritorial claims three weeks ago" means scrolling. With a Save Folder: - The user curates the keepers - The keepers are tagged, searchable, structured - The chat surface stays light (a conversation, not a permanent record) This is a familiar pattern from products like Notion AI, ChatGPT canvases, Perplexity collections — but here scoped to the project (UPC research / Patentanwalts-Workflow). ## Key design questions (for the inventor phase, not this issue) 1. **Granularity**: save the entire turn, or let the user select a passage? 2. **Editing**: is the saved output read-only (frozen at save time) or does the user keep editing/annotating? 3. **Folder structure**: flat tag-based, hierarchical (folders + sub-folders), or hybrid? 4. **Cross-references**: can a saved output reference the originating chat thread? Reference other saved outputs? 5. **Sharing**: per-user only, or can saved outputs be shared with collaborators / a team? 6. **Storage**: own table in Supabase (`saved_ai_outputs` with user_id + content + metadata + tags), or piggyback on existing artefact tables? 7. **Retention / quotas**: limit per plan tier, or unlimited? 8. **Export**: copy to clipboard, download as Markdown/PDF, send via email — what export formats? ## Out of scope (this issue) - Implementation — this is a parked idea. Inventor designs first. - Mobile UX, keyboard shortcuts, offline-mode — all later. - "Maybe too much" — m's own qualifier. The decision whether to ship at all is part of the inventor phase. ## Cross-link This idea applies symmetrically to **youpc.org** (also has AI assistants where users would benefit from the same save-and-organise pattern). Cross-tracked at yoUPC/youpc.org (separate issue, same parked status). If both projects ship this, they could reuse a common storage model — or each implements its own scoped to its data shape. To be decided. ## Workflow Park-only. When m wants to pursue: 1. Inventor designs UX + storage model + answers the design questions above 2. Coder implements stage 1: save + list + tag-filter 3. Coder implements stage 2: full-text search + export 4. Coder implements stage 3: sharing / collaboration if needed ## Refs - m's voice via PWA 2026-05-09 11:19 (parking request)
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
deferred
Parked — not pursuing now, may un-park later
done
Work completed by an agent; awaiting m to close
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
mAi m
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: m/paliad#29
No description provided.
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?