Implement slash (/) commands for Assistant

[vanguard-hack]

Feature Ticket: Slash Commands for Crime Analytics Assistant Chatbot

Summary

Add slash command support to the assistant chatbot's input bar. Users should be able to type / to trigger an autocomplete dropdown of predefined commands, select or type a command with arguments, and get a fast, structured response instead of relying purely on free-text natural language queries.

---

Background / Motivation

Officers repeatedly run the same categories of queries (FIR lookup, case status, suspect check, crime stats, etc.). Free-text queries work but are slower and less predictable. Slash commands give a fast, discoverable, keyboard-driven shortcut for the most common actions, similar to Slack/Notion/Linear command patterns.

---

Scope

In scope

• Slash command autocomplete UI in the chat input bar
• Command parsing/routing on submit
• Backend handlers for the 10 approved commands (list below)
• Role-based visibility/filtering of commands and results
• Audit logging for data-sensitive commands

Out of scope (for this ticket)

• Any command not in the approved list below
• Voice input handling
• Multi-language command aliases (future ticket)

---

Input Bar Placeholder Text

Default (empty state):

Ask a question or type / for commands...

When / is typed, placeholder is replaced by the live command dropdown (see UI behavior below).

---

Approved Slash Commands

Keep this list locked to these 10. Do not add more without a follow-up scoping discussion — the goal is a focused, low-friction set, not a full command palette.

Command	Arguments	Description	Category
/fir	[FIR number]	Get FIR details and current status	Lookup
/case	[case ID]	Case summary, IO assigned, current stage	Lookup
/suspect	[name or ID]	Criminal record / antecedents check	Lookup
/vehicle	[registration no]	Vehicle ownership & crime linkage check	Lookup
/person	[name or phone]	Person search across connected records	Lookup
/crime-stats	[district/PS] [date range]	Crime count summary by type	Analytics
/hotspot	[area]	Crime hotspot data for a location	Analytics
/wanted	[name or area]	Search wanted/absconding offenders list	Alerts
/missing	[name or ID]	Missing person case lookup	Alerts
/help	—	List all available commands	System
/clear	—	Clear current chat context	System

Note: /suspect and /person may return overlapping results depending on backend data sources. Confirm with data team whether these hit the same index before building separate handlers — may be mergeable into one command with a type filter.

---

UI Behavior

1. Trigger: User types / as the first character in the input bar (or after clearing input).
2. Dropdown appears: Shows all 10 commands, each with:
   • Command name (e.g., /fir)
   • Short description (e.g., "Get FIR details and status")
   • Expected argument hint, greyed out (e.g., [FIR number])
3. Filtering: As the user continues typing after /, filter the list by matching command name (e.g., /fi → shows /fir).
4. Selection:
   • Click/tap or arrow-key + Enter selects a command.
   • On selection, input bar auto-fills /command with cursor positioned after the space, ready for argument entry.
5. Argument hint: While typing arguments, show the expected format as placeholder-style greyed text (e.g., /fir [FIR number] fades as user types real value).
6. Submit: Enter or send button triggers parsing and routes to the relevant backend handler.
7. Invalid command: If typed command doesn't match any of the 10, show inline suggestion for closest match (e.g., "Did you mean /fir?") rather than a hard error or falling silently into free-text mode. If no close match, fall back to treating the input as a normal natural-language query.
8. Empty argument: If user submits a command with no argument (e.g., just /fir), prompt inline for the missing value rather than sending an invalid request to backend.

---

Command Parsing Logic (backend)

• Detect leading / in submitted message.
• Extract command token (up to first space) and match against the approved command list (case-insensitive).
• Remaining text after the command = argument string.
• Route to corresponding handler/service based on command.
• For /crime-stats, argument string may contain two parts (location + date range) — parse via simple delimiter (comma or keyword like "from...to") rather than complex NLP; confirm exact input format with design before building parser.
• If command not recognized, do not attempt backend routing — pass through to standard chatbot NLU pipeline.

---

---

Audit Logging

Every execution of /fir, /case, /suspect, /vehicle, /person, /wanted, and /missing must log:

• User ID / badge number
• Command + arguments used
• Timestamp
• Result returned (or reference ID to result)

This is required for compliance and internal review — same standard as any other sensitive data access on the platform.

---

Acceptance Criteria

• Typing / in the input bar opens the command dropdown with all 10 commands visible
• Dropdown filters correctly as user types after /
• Selecting a command auto-fills the input with correct syntax and argument placeholder
• All 10 commands route to correct backend handlers and return expected data
• Invalid/unmatched commands fall back gracefully to normal chat (no hard errors)
• Commands respect existing role-based access control
• Sensitive commands are logged per audit requirements
• /help returns an accurate, up-to-date list of all commands
• /clear resets chat context without requiring page reload
• Placeholder text displays correctly in default (non-/) state

---

Also for the UI designs, take inspiration from the below designs:

Also the Assistant must be required to give out the latest info based on the slash commands,