Core Concepts - Hitler

Bot-First Architecture

Hitler is designed with the bot as the primary interface. Unlike traditional SaaS apps where users interact through a web dashboard, Hitler employees interact primarily through Slack.

The web app provides an ops-brain dashboard for admins and managers, plus a simplified employee dashboard with task filters, list/board views, and pipeline access.

Employee ←→ Slack Bot ←→ Hitler API ←→ Database
                              ↓
                         LLM Provider

AI-Driven Task Creation

Hitler uses AI to parse natural language into structured tasks, which are auto-confirmed immediately — no draft confirmation step required. Users can modify tasks after creation via text.

Task Flow

When an employee asks the bot to create a task:

Natural Language Input

Employee: “Remind me to review the budget proposal by Friday”

AI Parsing

LLM extracts: title, due date, priority, description

Auto-Confirmed

Task is created and confirmed automatically

Bot Reply

Bot replies with a summary of the created task(s)

Modify via Text

User can update or cancel tasks through natural language

Each task gets a display ID (e.g., #12) for easy reference in conversation — users can say “mark #12 done” instead of typing the full title.

Conversational AI

Hitler goes beyond simple command parsing - it’s a full conversational assistant that understands context, detects intents, and maintains conversation memory.

Intent Detection

The AI analyzes each message to determine intent:

Intent	Description	Example
`task_create`	User wants to create a task	”I need to finish the report”
`task_list`	User asks about tasks	”What’s on my plate?”
`mood`	User wants to log mood	”Log my mood”
`greeting`	Casual greeting	”Hey, how’s it going?”
`general`	General conversation	”Feeling overwhelmed today”

Casual Task Detection

Hitler detects tasks even in casual conversation:

User: "tomorrow I need to do the budget review and send invoices"

Hitler automatically extracts:
1. "budget review" (due: tomorrow)
2. "send invoices" (due: tomorrow)

This allows natural conversation without explicit task syntax.

Mood Inference

The AI can silently detect mood signals:

User: "Ugh, feeling really burned out today"
→ Inferred mood: 2 (logged silently)
→ Response: "Sounds rough, hang in there!"

This provides mood data without interrupting the conversation flow.

Memory System

Hitler maintains two types of memory:

Session Memory

Short-term (Redis) - Last 10 messages - 30-minute TTL - Enables context-aware replies

Persistent Memory

Long-term (PostgreSQL) - User preferences (language, style) - Conversation summaries - Pending plans

Language Support

Hitler supports 13 languages with automatic detection and native responses:

Language	Code	Auto-Detected
English	`english`	Yes
Hinglish (Hindi-English)	`hinglish`	Yes
Spanish	`spanish`	Yes
French	`french`	Yes
German	`german`	Yes
Portuguese	`portuguese`	Yes
Italian	`italian`	Yes
Dutch	`dutch`	Yes
Japanese	`japanese`	Yes
Korean	`korean`	Yes
Chinese (Simplified)	`chinese`	Yes
Arabic	`arabic`	Yes
Russian	`russian`	Yes

The bot automatically detects the language from your message and responds in the same language:

User: "kal mujhe report submit karna hai"
Hitler: "acha bhai, 'submit report' add kar raha"

User: "añade una tarea: revisar el informe"
Hitler: "vale, añadiendo 'revisar el informe'"

User: "タスクを追加: レポートを確認"
Hitler: "了解、「レポートを確認」を追加するね"

You can also manually set your preferred language in the Chat page settings. Your preference is remembered across sessions.

Multi-Tenancy

Hitler is built for multi-tenant deployment. Each organization’s data is completely isolated.

Resource	Isolation Level
Users	Per organization
Tasks	Per organization
Mood data	Per organization
Flags	Per organization
Platform connections	Per organization
Secrets (tokens)	Per organization (encrypted)

Roles & Permissions

Hitler uses role-based access control with three levels:

Employee

View/manage own tasks - Log moods - Create inquiries - View own flags

Manager

Everything Employee can do - View team tasks - View anonymized team mood stats - Resolve team flags - Respond to inquiries

Admin

Everything Manager can do - Manage users & roles - Configure integrations - Organization settings - Audit logs

Tasks

Tasks in Hitler go through a specific lifecycle: Tasks are auto-confirmed on creation — there is no draft confirmation step.

Task Properties

Property	Description
`title`	Short description (max 200 chars)
`displayId`	Sequential per-org number (e.g., `#12`)
`description`	Detailed description (optional)
`priority`	1 (highest) to 5 (lowest)
`status`	open, in_progress, completed, cancelled
`dueDate`	Optional deadline
`stageId`	Pipeline stage assignment (optional)
`size`	Task size estimate (optional)
`clientId`	Associated client (optional)
`carryForwardCount`	Times task was carried forward from EOD
`rawInput`	Original natural language (for AI learning)

Pipeline Stages

Tasks can be organized into pipeline stages — customizable workflow columns (e.g., “Backlog”, “In Progress”, “Review”, “Done”). Pipeline stages are defined per organization and provide a Kanban-style view of work.

Task Logs

Every task change is recorded in an event log:

{
  "action": "status_changed",
  "changes": {
    "status": { "from": "open", "to": "in_progress" }
  },
  "actorId": "user-uuid",
  "timestamp": "2024-01-15T10:30:00Z"
}

This provides full auditability and enables analytics.

Moods

Mood tracking helps organizations understand team wellbeing over time.

Mood Values

Value	Meaning
1	Very unhappy
2	Unhappy
3	Neutral
4	Happy
5	Very happy

Privacy Model

Individual mood data is private by default. Managers only see aggregated, anonymized statistics.

Employees see their own mood history and trends
Managers see team averages and participation rates
No manager can see an individual employee’s mood entries

Journals

Employees can also keep private journal entries - free-form text reflections that are never shared with anyone.

Flags

Flags are system-detected concerns that require attention. They’re raised based on patterns in mood data, task completion, and other signals.

Severity Levels

Level	Description	Example Trigger
Low	Minor concern	Slight mood decline
Medium	Moderate concern	Multiple low moods in a week
High	Significant concern	Consistently low mood + missed tasks
Critical	Urgent attention needed	Crisis indicators detected

Escalation Actions

Each severity level has recommended actions:

{
  "severity": "high",
  "actions": [
    { "action": "Schedule 1:1", "priority": 1 },
    { "action": "Review workload", "priority": 2 },
    { "action": "Offer EAP resources", "priority": 3 }
  ]
}

Email Notifications

When a flag is created, managers automatically receive email notifications:

Immediate notification - Sent when flag is created (not waiting for escalation)
Includes severity - Subject line shows severity level (LOW, MEDIUM, HIGH, CRITICAL)
Employee context - Shows employee name and flag reason
All managers notified - Emails sent to all managers and admins in the organization

Email notifications require the RESEND_API_KEY environment variable to be configured. See Environment Variables.

Inquiries

Inquiries provide a private communication channel between employees and HR/managers.

Inquiries can be linked to flags for context, or created independently for any HR-related question.

Inquiry Flow

Employee creates inquiry (optionally linked to a flag)
HR/Manager receives notification
Back-and-forth messaging in private thread
Either party can close the inquiry when resolved

Morning Threads

The morning thread system automates daily task tracking rituals in Slack:

9 AM (configurable) — Bot posts a thread to #today-in-progress: “Good morning team! Share your tasks for today.”
Team replies — Each user replies to the thread with what they’re working on. The bot records each reply as a thread_submission.
11 AM (configurable) — Bot checks who hasn’t replied and sends a DM reminder to missing users.

This creates a daily accountability ritual where the team shares their plans publicly and the bot tracks participation automatically.

Follow-Up Matching

Morning thread follow-ups (submission deadline reminders) use LLM-based matching with scored matching to intelligently determine which users have already submitted and which are missing — rather than relying on exact string matching.

Org Preferences

Preference	Default	Description
`enableMorningThread`	`false`	Enable/disable morning thread system
`morningThreadTime`	`09:00`	When to post the morning thread
`submissionDeadlineTime`	`11:00`	When to check for missing submissions
`progressChannelId`	—	Slack channel ID for the thread

End-of-Day Collection

The EOD collection system closes the daily accountability loop started by the morning thread:

5 PM (configurable) — Bot posts a nudge as a reply to the morning thread: “Time to wrap up! Update your tasks with emoji statuses.”
7 PM (configurable) — Bot collects EOD replies and uses LLM per-task matching to determine the status of each task. Instead of simple emoji parsing, the LLM analyzes each user’s reply against their task list and scores how well each status update matches each task:
- Done — task marked completed
- Carry forward — task rolls to next day, carryForwardCount incremented
- Dropped — task cancelled
- No update — user did not reply, flagged for follow-up
Summary — Aggregated results posted to the progress channel showing team completion rates and carry-forward counts.

EOD Org Preferences

Preference	Default	Description
`eodNudgeTime`	`17:00`	When to post the EOD nudge to the thread
`eodCollectionTime`	`19:00`	When to scan and parse task statuses

Daily Cycle

AM  — Morning thread posted (tasks for today)
AM — Submission deadline check (DM missing users)
PM  — EOD nudge (reply to morning thread)
PM  — EOD collection (parse statuses, update tasks, post summary)

Notification Deduplication & Anti-Spam

Hitler uses a two-layer dedup system to prevent notification flooding while ensuring urgent deadlines are never missed:

Layer 1: Task-Level Dedup (24h)

Each task can only generate one notification per user per day. Redis key task-notif:{userId}:{taskId}:{date} tracks whether a task was already mentioned in a morning summary, nudge, or overdue reminder. Prevents the same task from appearing in multiple messages.

Layer 2: User-Level Cooldown (2h)

After any notification DM is sent, subsequent non-urgent nudges are suppressed for 2 hours. This prevents the 15-min proactive scanner from spamming users who just received a morning summary or batch nudge.

Urgent Bypass

Tasks with priority P1/P2 or due within 2 hours bypass the cooldown entirely. Critical deadlines always get through regardless of recent notifications.

Example Flow

00 AM  — Morning summary sent (2 overdue tasks) → cooldown starts
30 AM  — User adds 3 new tasks
45 AM  — Scanner fires → cooldown active, tasks are normal priority → skipped
15 AM — Cooldown expired → scanner sends ONE consolidated LLM nudge for 3 tasks
00 PM  — P1 task due in 1 hour → bypasses cooldown → immediate DM

Batching

The 15-min proactive scanner groups all upcoming + stale tasks by user and queues one batch nudge per user. The batch handler generates a single natural-language message via Claude Haiku covering all tasks, instead of sending separate DMs per task.

Smart Reminders

Users can set reminders via natural language (e.g., “remind me in 1 hour about the standup”). The bot uses LLM-based parsing with chrono-node for time extraction. Reminders are stored in the database and queued as BullMQ delayed jobs, making them restart-safe. Users can snooze reminders when they fire.

Platform Identities

Users can be linked to multiple platform identities:

User (john@company.com)
├── Slack: U1234567890
├── Teams: (future)
└── WhatsApp: (future)

This allows:

Single sign-on via Slack/Teams
Bot interactions linked to correct user
Cross-platform identity resolution

Secrets Management

Platform tokens (Slack bot tokens, etc.) are stored securely:

Encrypted at rest using AES-256-GCM
Stored in Cloudflare KV (not in the database)
Scoped per organization
Short-lived in memory - fetched, used, discarded

Key format: org:{orgId}:{platform}:{secretType} Example: org:abc123:slack:bot_token

​Bot-First Architecture

​AI-Driven Task Creation

​Task Flow

​Conversational AI

​Intent Detection

​Casual Task Detection

​Mood Inference

​Memory System

Session Memory

Persistent Memory

​Language Support

​Multi-Tenancy

​Roles & Permissions

Employee

Manager

Admin

​Tasks

​Task Properties

​Pipeline Stages

​Task Logs

​Moods

​Mood Values

​Privacy Model

​Journals

​Flags

​Severity Levels

​Escalation Actions

​Email Notifications

​Inquiries

​Inquiry Flow

​Morning Threads

​Follow-Up Matching

​Org Preferences

​End-of-Day Collection

​EOD Org Preferences

​Daily Cycle

​Notification Deduplication & Anti-Spam

​Layer 1: Task-Level Dedup (24h)

​Layer 2: User-Level Cooldown (2h)

​Urgent Bypass

​Example Flow

​Batching

​Smart Reminders

​Platform Identities

​Secrets Management

Bot-First Architecture

AI-Driven Task Creation

Task Flow

Conversational AI

Intent Detection

Casual Task Detection

Mood Inference

Memory System

Language Support

Multi-Tenancy

Roles & Permissions

Tasks

Task Properties

Pipeline Stages

Task Logs

Moods

Mood Values

Privacy Model

Journals

Flags

Severity Levels

Escalation Actions

Email Notifications

Inquiries

Inquiry Flow

Morning Threads

Follow-Up Matching

Org Preferences

End-of-Day Collection

EOD Org Preferences

Daily Cycle

Notification Deduplication & Anti-Spam

Layer 1: Task-Level Dedup (24h)

Layer 2: User-Level Cooldown (2h)

Urgent Bypass

Example Flow

Batching

Smart Reminders

Platform Identities

Secrets Management