Neuromancer
Back to home
Reports & Timesheets

Four dashboards. Zero extra instrumentation.

Four reports that live under /reports, all aggregated from data the platform already collects: AI logs, task events, time entries. Nothing is instrumented separately — if someone used the chat, generated a content or logged minutes, it's already in the report.

Token Usage

Where is my token budget going?

Every LLM call writes an AIUsageLog. The report pivots that log across four breakdowns.

Endpoint
GET /ai-logs/token-usage
Filters
  • provider — claude_code / anthropic / openai
  • model — claude-sonnet-4, gpt-4o, etc.
  • user_id
  • date_range
What it returns
  • Daily time series — input_tokens, output_tokens, cost_estimate.
  • Breakdown by model
  • Breakdown by user
  • Breakdown by function — chat.send, content.generate, learnings.extract, chat.title…

The AIUsageLog stores function, provider, model, tokens_in, tokens_out, duration_ms, user_id — the report is a pivot over that record, nothing instrumented on the side.

Leaderboard

Who's pulling the team forward?

Endpoint
GET /ai-logs/leaderboard
Engagement score (real formula)
score = messages_sent × 2
      + ai_calls × 1
      + content_generated × 5
      + chat_interactions × 1
Badges (actual rules in code)
  • champion — rank 1.
  • top3 — ranks 2 and 3.
  • power_user — score > 50 (independent of rank).

Each badge has its own color in light and dark themes (see BADGE_CONFIG), with a gold / silver / bronze podium for the top 3.

Timesheet

Who worked on what, and how much is billable?

Entries arrive in two ways: automatic timer on the Kanban task card, or manual CRUD. The report aggregates by user, client, project and date, with native split between billable and non-billable.

Model — TimeEntry
  • user_id, task_id, client_id, project_id
  • started_at, ended_at, minutes
  • billable — bool
  • description
  • source — timer / manual
Endpoint
GET /timesheets/report
Aggregations
  • Total minutes
  • Billable vs non-billable (split)
  • Daily chart
  • Entry list with drill-down
Two sources, one log
  • Automatic timer — Play/Stop on the Kanban task card creates the entry on stop.
  • Manual — CRUD directly via /timesheets/entries.
Productivity & Delays

Where does the team slip, and on which column?

Endpoint
GET /tasks/productivity-report
Real metrics
  • completion_rate — % of tasks closed in the period.
  • on_time_rate — % of completed tasks done before or on due_date.
  • variance_pct — average difference between estimated_minutes and actual_minutes.
  • daily_completed — time series of tasks completed per day.
Breakdowns
  • By priority — urgent / high / medium / low.
  • By user (assignee) — Who closes on time, who carries the delay.
  • By column — Kanban bottlenecks — which column holds things the longest. Not just knowing it's late, but knowing where it stalled.
Real stack

Where each number is collected.

Source — Token Usage
AIUsageLog written on every LLM call, across every route.
Source — Leaderboard
Same AIUsageLog + content.generate events; score computed on-the-fly, no parallel table.
Source — Timesheet
Model TimeEntry — Kanban timer + manual CRUD via /timesheets/entries.
Source — Productivity
Task events (creation, column moves, closure) + due_date + estimated/actual minutes.
Highlights

Why these four reports, and not twenty.

01

Zero configuration

The numbers already exist because the platform records everything. You don't 'turn on' the report — it exists because the data does.

02

Native billable split

Billable vs non-billable comes from TimeEntry itself. Not an add-on, not computed after — it's in the schema.

03

Bottleneck per column

Productivity broken down by column reveals the real Kanban bottleneck, not just ticket delay. Where work stalls shows up.

Want to see this running on your own pipeline?

We'll show you in a quick demo, using data you already work with.

Get a demoRead the docs