DocsIntegrationsAI Suggestions

AI Suggestions

Nareli's AI-powered suggestion system analyzes your work patterns and generates intelligent recommendations for tasks, time entries, and workflow improvements.

Overview

The AI suggestion system observes your work across Slack conversations and desktop activity, then generates actionable recommendations. It runs entirely on your machine using Ollama - your data never leaves your computer.

Analyzes Slack messages and desktop activity for work patterns
Generates suggestions you can accept or decline at your own pace
Learns from your decisions to improve over time
Integrates into the day view alongside manual time entries

AI suggestions require the Pro plan (€7/mo) or higher. Ollama must be installed and running on your Mac (localhost:11434). See Subscription & Billing for plan details.

What AI Analyzes

The AI receives raw data along with your existing projects, tasks, and recent time entries to make informed suggestions.

Slack messages from monitored channels and their conversation threads
Desktop activity logs (application names and window titles)
Your current project and task list for matching context
Recent time entries to avoid suggesting duplicates
Channel-to-project mappings for accurate project assignment
Historical accept/decline patterns (fingerprints) for relevance tuning

Suggestion Types

Four distinct suggestion types cover different aspects of your workflow.

Type	Purpose	On Accept
task_create	New work identified from Slack or activity patterns	Creates a new task with title, description, and project
task_update	New info relevant to an existing task	Merges updated context into the existing task
timeentry_create	Untracked work period detected from activity or Slack	Creates a time entry with start time, duration, and task
info	Contextual awareness (decisions, status updates)	Informational only - no action taken

The suggestion type is displayed prominently on each card so you can quickly scan and prioritize your queue.

Suggestion Workflow

Every suggestion starts as pending and waits for your review. There is no penalty for letting suggestions sit in the queue.

AI generates a suggestion from Slack or activity data - status: pending

Suggestion appears in the Suggestions page and as a block in the day view

You review the suggestion's type, content, source, and AI reasoning

Accept - executes the proposed action (create task, log time entry, etc.)

Or decline - removes from queue and signals your preference to the AI

Modify any details before confirming an acceptance
Declined suggestions are retained as learning signals, not deleted
System notifications alert you when new suggestions arrive
Day view shows suggestion blocks with overlap detection against existing entries

Learning from Your Patterns

The AI uses "fingerprints" - patterns extracted from your accept/decline history - to improve future suggestions.

Fingerprints capture which types, projects, times, and channels you value
Learning is gradual - requires consistent patterns before adjusting behavior
View fingerprints and acceptance stats on the Learnings page
Clear learning data anytime to start fresh

Privacy & Local Processing

Every component runs locally. No data is sent to external servers for processing.

AI model runs through Ollama on your Mac's hardware
All suggestion data and fingerprints stored in your local database
No cloud sync, no telemetry, no remote analytics
Works offline with no usage limits or API costs
Choose your Ollama model - larger models produce better suggestions but need more RAM