# Changelog ## [0.9.1] - 2026-07-17 ### Changed - Exports now save to `/tmp` (or the OS temp dir where `/tmp` doesn't exist) instead of the pi session's cwd, so they never litter a repo or home directory. The `✓ Saved` note shows the full path. - The export directory is configurable: `{ "usage-extension": { "exportDir": "~/Downloads" } }` in `~/.pi/agent/settings.json` (`~` expands; the directory is created if missing). ## [0.9.0] - 2026-07-17 ### Added - **Export (`e`) in every view** (#67): table → per-model CSV with TOTAL row; graphs → CSV of exactly the plotted slice (visible series, current metric/grouping/cumulative, ISO buckets); insights → structured JSON. Files land in the cwd with self-describing names; a ✓/✗ note confirms the write. - **Table cut/filter** (#68): `/` live type-to-filter over provider and model names — when only models match, the provider row is recomputed from just those models; `x` hides the selected row; `a` resets. The Total row recomputes over the visible slice, with an explicit status line, and table exports honor the slice. ### Changed - **Graphs is now the default view** (#66); `v` cycles Graphs → Table → Insights, and the title renders a view tab strip so switching is discoverable. - README refreshed with current screenshots of all three views (#69) and documentation for export and table filtering. ### Fixed - Table selection could desync from rendered rows (the unsliced provider map leaked aggregation insertion order instead of cost-sorted order). ## [0.8.1] - 2026-07-17 ### Changed - **Consistent units.** Every alarm in “Worth attention” now leads with dollars (the act-or-ignore number), with the period share in the dimmed parenthetical; “Where it went” keeps percentages (shares of the pie). Session concentration and upfront tax previously led with %. ## [0.8.0] - 2026-07-17 ### Changed - **Cache-miss taxonomy split into three behaviours**, each with its own alarm: resuming a conversation after a break (cache expired), switching models mid-conversation (the previous model's cache doesn't transfer), and true mid-session prefix changes (no break, compaction, or model switch to explain them). Validated against a second public corpus (badlogicgames/pi-mono, 628 sessions). - **Insights sections reordered facts-first**: “Where it went” (structure) now comes before “Worth attention” (alarms). - **pi's built-in test providers excluded**: `faux-provider` and `fake-provider` never call a real API and no longer appear in tables, graphs, totals, or insights. ## [0.7.2] - 2026-07-17 ### Changed - **Insights view polish.** Section headers with rules (“Worth attention” in warning colour, “Where it went” in accent), alarm markers and stats in warning colour, the trailing “(x% of this period)” de-emphasised, content capped at 100 columns for readability on wide terminals, and an explicit all-clear line (✓) when no alarms fire. The stat column fits 6-character values (e.g. `$58.79`). - **Plain-language copy.** Insight headlines now state what happened rather than the cache mechanism (“spent re-sending conversations after a break” instead of “re-warming caches that expired during idle gaps”), advice avoids jargon, and vendor-specific cache details are gone. - The period · “independent lenses” note is folded into the subtitle: “Approximate, based on local sessions on this machine (these are independent and don't sum to 100%).” ## [0.7.1] - 2026-07-17 ### Added - Descriptive loading messages with a live file counter while `/usage` opens: “Building your usage history for the first time…” on a fresh install, “Rebuilding your usage history — the cache format changed…” after an upgrade that bumps the cache version, and “Updating your usage history since \…” for routine incremental refreshes. Warm opens with nothing to parse keep the plain spinner. - `collectUsageData` accepts an `onProgress` callback reporting mode (`first-run` / `rebuild` / `update`), files to parse, files parsed, and the newest already-ingested session activity timestamp. ## [0.7.0] - 2026-07-17 ### Changed - **Insights view redesigned around materiality.** Insights are now split into two sections: **Worth attention** (alarms that only fire when a wasteful pattern is material for the period) and **Where it went** (always-on structural lenses). Each insight leads with a pre-formatted stat ($, %, or ×) instead of a raw percentage. - Alarms: TTL re-warm tax (large context re-written after a > 5 min idle gap; provider caches expire after a few minutes), prefix-change cache misses (no idle gap; messages right after a pi compaction are excluded), top-5 session concentration (≥ 35%), upfront session-start tax (≥ 8%), and a cache-leverage floor (< 5× cached-per-fresh tokens). - Structure: context tax (share of cost at ≥ 150k context, with $/msg vs small-context messages), project mix (top 3 project directories from each session's `cwd`; home prefixes from other machines collapse to `~`, worktrees collapse into their repository), reasoning share of output tokens, and a burn trend (last 7 days vs prior 4-week weekly pace). - Removed the old parallel-sessions, large-uncached-prompt, long-running-sessions, and unconditional top-5-concentration insights — measured against real usage they were either immaterial or misattributed (e.g. "parallel cost" was mostly just cost that happened while other sessions existed). ### Internal - Cache format bumped to **v3**: per-message `afterCompaction` flag and per-file session `cwd`. First open after upgrading does a one-off full rebuild (`compaction` entries and session headers are now parsed), then warm opens are fast again. - Insight aggregation is accumulator-based — raw per-period message arrays are no longer kept, so the insights redesign adds no steady-state memory or time cost. ## [0.6.1] - 2026-07-17 ### Changed - Graph lines are clipped to each series' active range (first to last bucket with usage). Applies to every grouping — provider, model, thinking level, `other`, and Total — so a model that starts late (e.g. reasoning tokens, only recorded since pi 0.80.3) or gets retired no longer draws a flat zero/flat cumulative tail across the whole period. ## [0.6.0] - 2026-07-17 ### Added - **Interactive graph explorer.** New third view mode (cycle with `v`): a braille line chart of usage over time for the active period, with a legend showing per-series totals and shares. - Metrics (`m`): cost, tokens, messages, reasoning tokens. - Grouping (`g`): by provider, by model, by thinking level, or total only — top 6 series plus an `other` rollup, with a bold Total line always drawn. - Cumulative running totals or per-bucket rates (`c`), hourly buckets for day/week periods and daily buckets for Last 30 Days / All Time. - Legend filtering: `↑`/`↓` moves the cursor, `Enter`/`Space` hides or shows a series, `a` shows all. The y-axis rescales to the visible series so small series can be inspected by hiding large ones. - **Thinking level and reasoning tokens.** Session parsing now replays `thinking_level_change` entries (compact and spaced JSON styles) to attribute a thinking level to each assistant message, and records `usage.reasoning` token counts. Messages before the first recorded change appear as `unknown`. ### Changed - On-disk cache format bumped to v2 (per-message thinking level and reasoning tokens). The first open after upgrading does a one-off full rebuild; older caches are ignored safely. ## [0.5.0] - 2026-07-17 ### Added - **Last 30 Days period.** New tab between Last Week and All Time covering the last 30 calendar days including today (from midnight 29 days back, DST-safe). Available in both the table and insights views. ## [0.4.0] - 2026-07-17 ### Performance - `/usage` no longer re-reads and JSON-parses every session JSONL file on every open. On a real 5.2 GB / 3,310-file corpus, warm opens went from ~17 s to ~0.3 s (~50×) and the one-off cold build from ~17 s to ~7 s, with peak memory roughly halved. - **On-disk cache.** Per-file extraction results are persisted to `/usage-extension-cache.json` (respects `PI_CODING_AGENT_DIR`), keyed by file size + mtime. Only new or changed session files are re-parsed; entries for deleted files are evicted. Deleting the cache file forces a full rebuild — it is recreated automatically. - **Buffer-level pre-filter.** Session files are scanned as raw bytes and only lines that can be a session header or an assistant message are UTF-8-decoded and JSON-parsed; the multi-megabyte tool-result lines that dominate session files are skipped with a cheap byte search. Both compact (`"role":"assistant"`) and spaced (`"role": "assistant"`, seen in imported third-party session files) JSON styles are matched. - Changed files are parsed with bounded concurrency, and a cancelled cold build persists partial progress so the next open resumes instead of restarting. ### Changed - Data collection, caching, and insights moved from `index.ts` into `usage-extension/data.ts`, with tests in `tests/usage-data.test.mjs`. Verified equivalent to the previous parser across all 3,310 real session files (zero extraction mismatches). - Files without a session header (which are ignored, as before) no longer register their messages in the branch-dedupe hash set, so an identical message in a later, valid session file is now counted instead of silently dropped. ## [0.3.2] - 2026-05-07 ### Changed - Declare the `@earendil-works` Pi peer and development dependencies used by runtime imports. - Update Pi extension imports to the new `@earendil-works` namespace. ## 0.3.1 - 2026-04-19 - **Cost-only insights.** The Insights view now weights every insight by recorded USD cost, with no tokens fallback. The headline question is now "What's contributing to your cost?" and every bullet reads "X% of your cost …". Periods with no recorded cost show an explicit empty state instead of silently switching unit. - **Long-running sessions use true lifetime.** The 8h+ insight now looks at each session's global lifetime across all session files, not just the span visible inside the selected period slice. - **Exact ±2 min parallel window.** The "4+ sessions in parallel" insight now uses a precise ±120000 ms two-pointer sweep instead of rounded minute buckets. A message at second 1 of minute M and another at second 59 of minute M+2 are correctly treated as ~178 s apart (outside the window). - **Empty states.** Insights view now distinguishes three cases: no usage recorded in the period, usage but no cost data, and usage with cost but no insights clearing the 1% threshold. - **Narrow-terminal compact hint is hidden in Insights mode** (it only applied to the table layout). - **"Cache miss" bullet relabelled** to "of your cost came from >100k-token uncached prompts" — same math, more accurate wording. - **Messages with missing/invalid timestamps are excluded** from the parallel-sessions sweep so that older/incomplete logs don't inflate the insight by collapsing into a single synthetic instant. ## 0.3.0 - 2026-04-19 - Add an **Insights** view to `/usage` (press `v` to toggle). Surfaces Claude-style narrative characteristics of your usage for the active time period: - `X% of your usage was while 4+ sessions ran in parallel` - `X% of your usage was at >150k context` - `X% of your usage hit a >100k-token cache miss` - `X% of your usage came from sessions active for 8+ hours` - `X% of your usage came from your top 5 sessions` - Insights are weighted by cost when cost data is recorded, otherwise by tokens, with a small footer noting which basis is in use. - Insights are independent characteristics of usage (they overlap), not a breakdown — the view makes this explicit. ## 0.2.1 - 2026-04-17 - Add a one-line formula footer to the `/usage` dashboard (`Tokens = Input + Output + CacheWrite · ↑In = Input + CacheWrite`) - README now calls out the 0.2.0 formula change explicitly under the columns table ## 0.2.0 - 2026-04-17 - Include `cacheWrite` in the main `Tokens` total and in the `↑In` column so providers like Anthropic that report fresh prompt work under `cacheWrite` are no longer undercounted - Keep `cacheRead` out of `Tokens` so repeated cache hits do not swamp the dashboard - Keep the `Cache` column as combined cache read + write tokens for reference - Minor semver bump: the numbers shown under `Tokens` and `↑In` are now higher for Anthropic usage. Cost, `↓Out`, and `Cache` are unchanged. ## 0.1.7 - 2026-04-09 - Prevent `/usage` from crashing in narrow terminals by switching to a compact responsive table and truncating every rendered line to the terminal width - Thanks @markokocic ## 0.1.6 - 2026-04-09 - Add a "Last Week" time period tab - Thanks @ttttmr ## 0.1.5 - 2026-04-09 - Keep recursive subagent session scanning in `/usage` - Remove the deduped/raw mode toggle and keep the deduped view as the default behavior ## 0.1.4 - 2026-04-09 - Scan session files recursively so nested subagent runs are included in `/usage` - Add deduped vs raw counting modes to compare copied branch history against raw file totals ## 0.1.3 - 2026-02-03 - Add preview image metadata for the extension listing. ## 0.1.2 - 2026-01-13 - Add loading spinner while parsing session files (Esc to cancel) - Make data loading async to keep UI responsive - Thanks @nicobailon ## 0.1.1 - 2026-01-12 - Deduplicate assistant messages across branched sessions to avoid double-counting - Tokens total now excludes cache read/write tokens (cache remains in Cache column) - Thanks @nicobailon ## 0.1.0 - 2026-01-10 - Initial release - Collapsible provider/model view - Three time periods: Today, This Week, All Time - Token breakdown columns (dimmed for de-emphasis)