fix: default view falls back to totals for pure-active sessions (#132)

[microsasa]

Summary

Fixes #132 — the default view (copilot-usage) showed all 0s for the "Active Sessions (Since Last Shutdown)" section when a session had never been shutdown.

Root cause

_render_active_section and render_cost_view always used the active_* fields directly (active_model_calls, active_user_messages, active_output_tokens). For pure-active sessions where these fields are 0 (e.g. older parser output or direct SessionSummary construction), the display showed zeros while the total counters and model_metrics held the real data.

Meanwhile, render_live_sessions already had a has_active_stats fallback that correctly displayed totals in this scenario.

Fix

Added the same has_active_stats fallback to both _render_active_section and render_cost_view:

• When active_* fields are all zero and last_resume_time is None, fall back to model_calls, user_messages, and _estimated_output_tokens()
• When any active_* field is non-zero or last_resume_time is set (resumed sessions), use the active_* fields as before

This makes all three views (render_live_sessions, _render_active_section, render_cost_view) consistent.

Tests

• test_pure_active_never_shutdown_falls_back_to_totals — verifies the default full-summary view shows totals (58 model calls, 382 user msgs, 204.0K tokens) when active_* are 0
• test_pure_active_never_shutdown_cost_falls_back — verifies the cost view's "↳ Since last shutdown" row shows totals when active_* are 0

All 430 tests pass. Coverage: 98%.

Generated by Issue Implementer · ◷

[Copilot]

Pull request overview

Fixes the default copilot-usage view (and cost view) so “Active Sessions (Since Last Shutdown)” doesn’t show all zeros for pure-active sessions that have never been shut down, by falling back to total counters when active_* fields are unset/zero.

Changes:

• Add has_active_stats fallback logic to _render_active_section to display totals when active_* == 0 for never-shutdown active sessions.
• Apply the same fallback in render_cost_view for the “↳ Since last shutdown” row.
• Add regression tests covering both the full summary view and the cost view fallback behavior.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
src/copilot_usage/report.py	Adds fallback logic for pure-active never-shutdown sessions in active section + cost view.
tests/copilot_usage/test_report.py	Adds regression tests for the fallback behavior in full summary and cost views.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

In render_cost_view, when has_active is false you set cost_tokens = _estimated_output_tokens(s), which is derived from s.model_metrics. But grand_output has already been incremented by mm.usage.outputTokens earlier in the loop when s.model_metrics is present, so grand_output += cost_tokens will double-count tokens for the fallback case (issue #132 scenario). Consider only adding to grand_output the incremental active tokens that are not already represented in model_metrics (e.g., add s.active_output_tokens only for resumed sessions, or guard the addition when falling back to totals).

Suggested change

	grand_output += cost_tokens
	if has_active or not s.model_metrics:
	grand_output += cost_tokens

[microsasa]

Great catch on the double-counting. Applied the guard: grand_output is now only incremented by cost_tokens when has_active is true (active fields, not derived from model_metrics) or when s.model_metrics is absent (so line 943 never ran). This prevents the fallback path from adding tokens that were already counted via model_metrics.

Generated by Review Responder

Warning

⚠️ Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• astral.sh

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "astral.sh"

See Network Configuration for more information.

[Copilot]

has_active uses active_user_messages/active_output_tokens but not active_model_calls. If active_model_calls > 0 while the other two counters remain 0 (e.g., outputTokens missing/non-int), the cost view will incorrectly fall back to totals and estimate cost from model_calls instead of active_model_calls. Add s.active_model_calls > 0 to the has_active predicate.

Suggested change

	or s.active_output_tokens > 0
	or s.active_output_tokens > 0
	or s.active_model_calls > 0

[microsasa]

Agreed — applied the suggestion. Added s.active_model_calls > 0 to the has_active predicate in render_cost_view. This prevents the cost view from incorrectly falling back to totals when only active_model_calls is non-zero.

Generated by Review Responder

Warning

⚠️ Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• astral.sh

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "astral.sh"

See Network Configuration for more information.

[Copilot]

This regression test exercises the fallback display in the "Since last shutdown" row, but it doesn't assert the cost view's "Grand Total" output tokens. Given the new fallback uses _estimated_output_tokens(s) (derived from model_metrics), it's easy to accidentally double-count tokens in grand_output. Adding a Grand Total assertion here would prevent that regression.

Suggested change

	# Grand Total row must also reflect the correct token total without double-counting.
	grand_row = next(line for line in lines if "Grand Total" in line)
	grand_cols = [c.strip() for c in grand_row.split("│")]
	assert "50.0K" in grand_cols[6], (
	f"Output Tokens in Grand Total row should be 50.0K, got '{grand_cols[6]}'"
	)

[microsasa]

Good idea — added the Grand Total assertion to test_pure_active_never_shutdown_cost_falls_back. This verifies that the Grand Total output tokens are 50.0K (matching the single session's model_metrics), catching any double-counting regression in grand_output.

Generated by Review Responder

Warning

⚠️ Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• astral.sh

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "astral.sh"

See Network Configuration for more information.

[Copilot]

has_active_stats ignores active_model_calls. A resumed session can have post-shutdown assistant.turn_start events (so active_model_calls > 0) while active_user_messages/active_output_tokens remain 0 (e.g., missing/non-int outputTokens). In that case this code will incorrectly fall back to totals and display the wrong "since last shutdown" model-call count. Include s.active_model_calls > 0 in the has_active_stats predicate (matching what you later read/display).

Suggested change

	or s.active_output_tokens > 0
	or s.active_output_tokens > 0
	or s.active_model_calls > 0

[microsasa]

Good catch. Applied the suggested fix — added s.active_model_calls > 0 to the has_active_stats predicate in both render_live_sessions (line 178) and _render_active_section (line 840). This ensures that a resumed session with post-shutdown assistant.turn_start events but zero active_user_messages/active_output_tokens correctly uses the active fields instead of falling back to totals.

Generated by Review Responder

Warning

⚠️ Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• astral.sh

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "astral.sh"

See Network Configuration for more information.

[github-actions]

Quality Gate: APPROVE ✅

Low-impact bug fix with good test coverage. Auto-approving for merge.

Evaluated:

• Adds has_active_stats fallback to _render_active_section and render_cost_view, matching the existing pattern in render_live_sessions — all three instances use identical condition logic
• Two meaningful regression tests verify the fallback for both the default full-summary view and cost view
• Only touches display/rendering logic in report.py — no changes to data models, parsers, or API contracts
• CI passing (check, CodeQL, analyze, review all green)

Impact: LOW — display-only fix for an edge case where pure-active sessions showed zeros instead of totals.