feat: implement stats command

[kaanmertkoc]

Description

Closes #1634

Implements the /stats command to aggregate token usage and costs across all sessions.

Usage

opencode stats --days 7

What it shows

• Overview: Session count, total messages, time range
• Costs & Tokens: Total cost, cost per day, input/output tokens, cache reads/writes
• Tool Usage: Breakdown of which tools were used most frequently

Implementation

• Fetches all sessions and projects
• Processes in batches of 20 to avoid memory overhead
• Filters by date range with --days flag (default: 7 days)
• Works with existing storage, no schema changes needed
• Shows warning for >100 sessions

Performance

Tested on M4 Pro, 24GB RAM:

• 14 sessions, ~1.5k messages = ~1.3s

Unsure how it scales with 100+ or 1000+ sessions.

Feedback welcome if anyone's willing to test at scale.

Screenshots

[rekram1-node]

can we default to all time, also can we have some breakdown per project? Or have a global or some other flag

[kaanmertkoc]

Hey! Removed the default days to all time.

Quick question about the project breakdown. I've implemented an --all flag that toggles between current project and all projects. But I'm wondering if you wanted a more detailed breakdown showing stats for each project separately?

Current implementation:

• opencode stats → current project only
• opencode stats --all → all projects combined

Let me know if you'd prefer a different approach!

[rekram1-node]

May need some tool name aligning

[rekram1-node]

there have definitely been context7 calls in the past year I think I had some yesterday are we hiding some number of tools?

[kaanmertkoc]

there have definitely been context7 calls in the past year I think I had some yesterday are we hiding some number of tools?

Yes! The original code was limiting this to top 10 tools with .slice(0, 10), which was cutting off context7. I bumped it to 20 temporarily, but that's not really a solution.

How should we approach to this?

[kaanmertkoc]

May need some tool name aligning

Thanks for catching that. I fixed the alignment by truncating long tool names with ".." so the table lines up properly now.

Please let me know what do you think about this approach.

[rekram1-node]

we could add some flag for # of tools and default to all

[rekram1-node]

let's make this a project flag, that is a string,

if not added:

• all

if added but empty:

• current project

if added and has a value:

• use that project id

[kaanmertkoc]

Perfect, changed it to --project flag with the behavior you described:

• Not provided → all projects
• Empty string → current project (opencode stats --project "")
• Specific value → that project ID (opencode stats --project <id>)

Let me know if this works!

[kaanmertkoc]

we could add some flag for # of tools and default to all

Done! Added --tools flag that accepts a number. Defaults to showing all tools if not specified.

Usage:

• opencode stats → shows all tools
• opencode stats --tools 10 → shows top 10 tools

[rekram1-node]

looking good! should be good to merge when i get back from grocery store