Feature Request: Support additional placeholder variables in the system prompt

[LexioJ]

Describe the feature you'd like to request

Summary

Currently, the admin-configurable "Chat User Instructions" (system prompt) for both chat completions and title generation only supports a single placeholder: {user}, which resolves to the current user's display name (see lib/Controller/ChattyLLMController.php, str_replace('{user}', $user->getDisplayName(), $userInstructions)).

It would be useful to support a small set of additional placeholders so admins can write more context-aware and personalized system prompts without needing code changes for every use case.

Motivation

Right now the system prompt is fairly static aside from the display name. Being able to inject a few more pieces of contextual information (date, language, login name, etc.) would let admins:

• Localize/adjust behavior based on the user's preferred language
• Reference the current date/time for time-sensitive tasks (e.g. "today is {date}")
• Distinguish between a user's display name and their login/username

Describe the solution you'd like

Proposed variables

Placeholder	Resolves to
{user}	(existing) User's display name
{username}	User's login name / UID (IUser::getUID())
{email}	User's email address (IUser::getEMailAddress())
{language}	User's preferred language (e.g. from IUser/IL10N)
{date}	Current date (server or user timezone)
{time}	Current time (server or user timezone)

Additional context

• Substitution currently happens via simple str_replace() calls in lib/Controller/ChattyLLMController.php (used in newSession(), generateTitle(), and checkTitleGenerationTask()), so adding more placeholders would follow the same pattern.
• The available placeholders are documented inline in the admin settings UI (src/components/AdminSettings.vue) via NcNoteCard hint text below each prompt field — this documentation should be updated to list any newly added variables.
• Open question: should placeholder resolution be strict (only replace exact known tokens) or should unknown {...} tokens be left untouched/escaped to avoid breaking existing prompts that happen to contain literal curly braces?

Describe alternatives you've considered

None

Assisted-by: ClaudeCode:claude-fable-5

[janepie]

Hello, was this issue generated with or assisted by AI? Please check out our AI contribution policy: https://github.com/nextcloud/.github/blob/master/AI_POLICY.md

[LexioJ]

AI helped me analyze the source code to understand if the required capabilities are present and supported me writing this issue.

[janepie]

Using AI is totally fine, but we expect you to

1. disclose this - thanks for the edit!
2. write all comments and issues in your own words:

Issues, PR descriptions, and review comments must be in the contributor's own words. Translation assistance and grammar/spelling help are acceptable exceptions and do not need to be disclosed - the intent of this rule is to ensure that the ideas, reasoning, and decisions in community communication come from the contributor.

Can you please give a concise summary of what you would like to achieve and give some example use cases? I'm trying to find out if this is something that should be part of the configurable chat instruction or if it should go into the Context Agent system prompt in general, or both.

[LexioJ]

Today there is only one placeholder available. The LLM may need more context within the conversation like current date for a proper response. I think admins should decide which context is being provided to an AI Service (eg email address) - I don’t like the idea sending data not govern/decided by the admin.