docs: one analytics story, and the images catch up with the consolidated surface (issue #181)

The last slice of #181: every doc and both READMEs now describe the
one analytics surface.

- docs/costs.md and docs/graph.md lose their view/command sections and
  survive as the semantics references behind what the insights page
  draws; docs/insights.md speaks the bare grammar, gains the budget
  and trend-line reading bullets, and points the what-if at
  insights whatif; demo.md's shot list gains the i beat and the exact
  new footer; workflows.md and launch-kit.md updated.
- README.md: the intro bullet, the when-to-use paragraph (bare
  /dispatch insights, budget dials named as the one real lever), the
  admin-panel section rewritten around the i key and the page, the
  command list, and the usage-ledger sentence. admin/README.md: the
  three analytics bullets merge into one insights story. Dash rule
  holds in both.
- Images: costs-view.svg/png DELETED (the view is gone);
  dispatch-dashboard.svg's footer line rewritten to the shipped
  footer (i insights; row width preserved at 80 chars inside borders)
  and its PNG re-rendered at the exact prior 1803x1811;
  insights-view.png regenerated from the page WITH the budget panel,
  cumulative line and flow multiples; graph-view.png kept byte-
  unchanged as the topology close-up. Changed references bump to
  ?v=0.12.0.

Specs UNCHANGED, checked: this slice states what the code slices
shipped and changes no behaviour.

Suite green in the CI posture: 2192 tests, 0 skipped, live Valkey.

Signed-off-by: Rob Boerman <robboerman@live.nl>

Author: edgehero

README.md

@@ -13,9 +13,7 @@ did and what it spent, and shuts the container down. A durable queue absorbs bur
 checked before a single token is spent, and a live admin panel shows everything and can turn the whole
 thing off.
 
-![The /dispatch dashboard overlay, theme-colored: live queue state, day/week/month spend meters plus a daily token counter, the unified triggers pane (cron, label, comment, pull_request; selectable and editable), scheduled pause windows, and the interactive runs list, in one framed TUI](docs/images/dispatch-dashboard.svg?v=0.5.0)
-
-![The COSTS view, verdict-first cost analytics: per-plan SAVING/LOSING verdicts against API rates, a daily spend sparkline, per-flow, per-trigger, per-model and per-repo spend, subscription amortization with peak-window facts, and a what-if that re-prices a flow under another model, every estimate visibly marked](docs/images/costs-view.svg?v=0.11.0)
+![The /dispatch dashboard overlay, theme-colored: live queue state, day/week/month spend meters plus a daily token counter, the unified triggers pane (cron, label, comment, pull_request; selectable and editable), scheduled pause windows, and the interactive runs list, in one framed TUI](docs/images/dispatch-dashboard.svg?v=0.12.0)
 
 ![Transcript of /dispatch status, runs, and triggers: queue counts, the run-history table with per-job token and cost accounting, and the unified {on,run} triggers list](docs/images/dispatch-commands.svg?v=0.5.0)
 
@@ -25,9 +23,9 @@ README, no permission system. pi-dispatch is exactly that missing operational la
 - **The container is the boundary.** Every job runs `--cap-drop=ALL`, non-root, ephemeral, with its
   instructions mounted read-only. That is pi's missing permission system, enforced by Docker.
 - **Spend is bounded before a container starts**: a per-job turn budget plus daily, weekly and monthly
-  caps, checked before a single token is spent. And analyzed after: the panel's COSTS view shows spend
-  per flow, trigger, model, day and repo, what a subscription actually saves, and what a flow would
-  cost on another model ([`docs/costs.md`](docs/costs.md)).
+  caps, checked before a single token is spent. And analyzed after: the insights page shows spend
+  per flow, trigger, model, day and repo, what a subscription actually saves, what a flow would
+  cost on another model, and the budget dials themselves ([`docs/costs.md`](docs/costs.md)).
 - **The image is yours to shape.** Bake a project's toolchain into [`image/Dockerfile`](image/Dockerfile);
   it ships Playwright and Chromium, so a flow can build a frontend, screenshot it, and iterate on the
   rendered result. Any trigger can name its own image with `run.image`
@@ -55,10 +53,11 @@ here it is just how the pieces fit). You **design** loops as triggers plus commi
 **graph** shows the loops you actually built (what triggers what, what chained to what, where a skill's
 own text says it might loop, [`docs/graph.md`](docs/graph.md)); and **insights** prices them (what each
 trigger and flow costs, whether a subscription pays off, drawn as charts beside that same topology,
-[`docs/insights.md`](docs/insights.md)). One `/dispatch insights html` gives you the whole picture as a
-single file your browser opens from disk:
+[`docs/insights.md`](docs/insights.md)). One `/dispatch insights` gives you the whole picture as a
+single file your browser opens from disk, budget dials included, because the caps are the one lever
+that actually changes what all of this costs:
 
-![The insights page: KPI tiles, a plan verdict card, the daily spend chart, per-flow/per-trigger/per-model/per-repo breakdowns with plan-covered buckets drawn as chips instead of dollar bars, and the trigger/flow topology with spend badged onto the triggers that earned it](docs/images/insights-view.png?v=0.11.0)
+![The insights page: KPI tiles, the budget dials, a plan verdict card, the daily and cumulative spend charts, per-flow trend panels, the four breakdowns with plan-covered buckets drawn as chips instead of dollar bars, and the trigger/flow topology with spend badged onto the triggers that earned it](docs/images/insights-view.png?v=0.12.0)
 
 ## Quickstart
 
@@ -407,16 +406,15 @@ Steer the running worker without stopping it, from any terminal:
 ## The admin panel
 
 The dashboard shown at the top of this README is a **pi extension**: it loads into your own interactive
-pi session. No daemon, no web app, no network port. Beside the dashboard, `/dispatch graph` renders the
-whole trigger and flow topology: what triggers what, what chained to what in the recorded runs, what a
-skill's own text says it might chain to, plus orphan skills and dangling triggers. It comes as a
-dashboard view (`g`), as plain text, and as `/dispatch graph html`, a self contained page your browser
-opens from disk (still no server and no port) with the topology drawn Node-RED style
-([`docs/graph.md`](docs/graph.md)). Trigger rows carry their schedule (next fire, or overdue) and
-their window spend, and `/dispatch insights html` combines this topology with the cost analytics on
-one page ([`docs/insights.md`](docs/insights.md), shown [above](#when-to-use-it)).
+pi session. No daemon, no web app, no network port. Analytics live on the insights page: pressing `i`
+(or typing `/dispatch insights`) writes one self contained file and opens your browser (still no
+server and no port), with the whole trigger and flow topology drawn Node-RED style, what triggers
+what, what chained to what in the recorded runs, what a skill's own text says it might chain to,
+orphan skills and dangling triggers flagged, each cron's next fire or overdue state, and spend badged
+onto the triggers that earned it ([`docs/graph.md`](docs/graph.md) explains every edge;
+[`docs/insights.md`](docs/insights.md) the page, shown [above](#when-to-use-it)).
 
-![The trigger and flow graph as /dispatch graph html draws it: cron and forge triggers wired to their flows, an observed chain edge carrying its run count and recency, a potential mention, a skill with its prose loop grouped inside it, cron re-arm loops with their schedules, an orphan skill dimmed, the forge group naming the repos its runs hit, and the legend stating the chain caps and honesty counters](docs/images/graph-view.png?v=0.11.0)
+![The topology pane of the insights page: cron and forge triggers wired to their flows, an observed chain edge carrying its run count and recency, a potential mention, a skill with its prose loop grouped inside it, cron re-arm loops with their schedules, an orphan skill dimmed, the forge group naming the repos its runs hit, and the legend stating the chain caps and honesty counters](docs/images/graph-view.png?v=0.12.0)
 
 ```bash
 pi install npm:@edgehero/pi-dispatch-admin   # then, in pi:  /dispatch
@@ -436,13 +434,13 @@ banner: setup is offered when there is nothing, never over an outage.
 
 Inside the panel: `p`/`r` pause and resume the queue, arrows and `Enter` drill into triggers and runs,
 `a`/`e`/`x` add, edit and delete triggers (validated, atomic, reloaded live by both services), `s` edits
-a limit, `w` manages quiet hours, `c` opens the COSTS view (its `f` key cycles spend by flow, model,
-trigger and repo). `Enter` on a run opens its full record:
+a limit, `w` manages quiet hours, `i` opens the insights page. `Enter` on a run opens its full
+record:
 
 ![The RUN_DETAIL drill-in, a colored post-mortem of one run's PII-free record: outcome, target, timing with duration, turns/exit/budget slot, tokens and cost, and a chain line naming spawned children](docs/images/dispatch-run-detail.svg)
 
-The same surface exists as plain commands (`/dispatch status | runs | logs | budget | triggers | costs |
-graph | insights | run | pause | resume | set | unset`), all local, no model involvement.
+The same surface exists as plain commands (`/dispatch status | runs | logs | budget | triggers |
+insights | run | pause | resume | set | unset`), all local, no model involvement.
 
 ### Operating pi-dispatch from your AI
 
@@ -525,7 +523,7 @@ Same machinery, per-forge correctness differences, each with a full setup doc:
 
 Every job writes a durable, id-only record under `PI_LOGS_DIR` (never issue or comment text; raw logs
 are opt-in via `PI_CAPTURE_JOB_LOGS=1` and stay host-side). Each record carries a per-model **usage
-ledger**, which is what the COSTS view, `/dispatch costs`, and the what-if re-pricing fold over. Declare
+ledger**, which is what the insights page, `dispatch_costs`, and the what-if re-pricing fold over. Declare
 what your subscriptions cost in `subscriptions.json` and the screen shows whether they actually save
 money; without it, zero-rate runs show `$0 (unrated)`, never "free" ([`docs/costs.md`](docs/costs.md)).
 

admin/README.md

@@ -55,26 +55,14 @@ Four forges: GitHub, GitLab, Forgejo (and Gitea), Azure DevOps. **Who may fire a
 One command puts a live TUI over the whole deployment:
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/edgehero/pi-dispatch/main/docs/images/dispatch-dashboard.png?v=0.5.0" alt="The /dispatch panel: status, spend meters, triggers, runs, and settings" width="820">
+  <img src="https://raw.githubusercontent.com/edgehero/pi-dispatch/main/docs/images/dispatch-dashboard.png?v=0.12.0" alt="The /dispatch panel: status, spend meters, triggers, runs, and settings" width="820">
 </p>
 
 - **Status and spend.** Queue and worker state, day/week/month spend meters, a daily token counter, and a run-history table with per-job tokens and cost.
-- **Costs, analyzed honestly.** The COSTS view (`c`) shows spend per flow, model, trigger, repo and day, what each declared subscription actually saves against API rates, and a what-if that re-prices a flow under another model. Every dollar carries its class: a plan-covered run never renders as $0.00, and an estimate is always marked as one.
+- **Insights, the one analytics page.** Press `i` on the panel (or type `/dispatch insights`) and a single self contained page opens in your browser: the budget dials (the one lever that actually changes what all of this costs), per-plan verdicts against API rates, daily, cumulative and per-flow spend charts, the four breakdowns (flow, trigger, model, repo), and the whole trigger and flow topology with spend badged onto the triggers that earned it. Every dollar carries its class: a plan-covered run never renders as $0.00, and an estimate is always marked as one. Design your agent loops as triggers and skills, see the loops you actually built, and see what each one costs, all in one place ([`docs/insights.md`](https://github.com/edgehero/pi-dispatch/blob/main/docs/insights.md)).
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/edgehero/pi-dispatch/main/docs/images/costs-view.png?v=0.11.0" alt="The COSTS view: per-plan verdicts against API rates, a daily spend sparkline, spend by flow, trigger, model and repo, and subscription amortization" width="820">
-</p>
-
-- **The topology, as a graph.** The GRAPH view (`g`) and `/dispatch graph` show what triggers what, what chained to what in the recorded runs, and what a skill's own text says it might chain to, with orphan skills and dangling triggers flagged, each cron's next fire or overdue state, and each trigger's window spend. `/dispatch graph html` draws the same picture Node-RED style in your browser, from one self contained file on disk: no server, no port, and the page reloads itself when you regenerate it.
-
-<p align="center">
-  <img src="https://raw.githubusercontent.com/edgehero/pi-dispatch/main/docs/images/graph-view.png?v=0.11.0" alt="The trigger and flow graph: triggers wired to their flows, an observed chain edge with its count and recency, a potential mention, a skill with its prose loop grouped inside it, cron re-arm loops, an orphan skill dimmed, the forge group naming the repos its runs hit, and the legend stating the chain caps and honesty counters" width="820">
-</p>
-
-- **One page for both questions.** `/dispatch insights html` writes a single self contained page: the topology with spend badged onto its triggers, beside the cost analytics drawn as charts (daily spend, the four breakdowns, plan verdict cards). Design your agent loops as triggers and skills, see the loops you actually built, and see what each one costs, all in one place ([`docs/insights.md`](https://github.com/edgehero/pi-dispatch/blob/main/docs/insights.md)).
-
-<p align="center">
-  <img src="https://raw.githubusercontent.com/edgehero/pi-dispatch/main/docs/images/insights-view.png?v=0.11.0" alt="The insights page: KPI tiles, a plan verdict card, the daily spend chart, the four breakdowns with plan-covered buckets drawn as chips instead of dollar bars, and the topology with spend badges" width="820">
+  <img src="https://raw.githubusercontent.com/edgehero/pi-dispatch/main/docs/images/insights-view.png?v=0.12.0" alt="The insights page: KPI tiles, a plan verdict card, the daily spend chart, the four breakdowns with plan-covered buckets drawn as chips instead of dollar bars, and the topology with spend badges" width="820">
 </p>
 
 - **Triggers, editable live.** cron, label, comment and pull_request triggers with colored drill-ins showing what fires each one, what it runs, and its trust model. Added, edited and deleted without a restart. Triggers that run third-party code or a custom image are badged; opting in or out of either stays an edit to the reviewed `triggers.json`, which neither the console nor a model-callable tool will make for you.

docs/costs.md

@@ -1,41 +1,24 @@
 # Cost analytics
 
 Every run records what it spent (`docs`: run history; `specs`: `REQ-TOKEN-ACCOUNTING-AND-CAPS`). The
-costs surface makes that history analyzable: what a flow costs, what a month costs per model, what a
+cost analytics make that history analyzable: what a flow costs, what a month costs per model, what a
 subscription is actually saving, and what a flow *would* cost on a different model. It informs; it
 changes nothing — no auto-switching, no vendor API calls, no database (`REQ-COST-ANALYTICS`,
 `DES-COST-FOLD-BY-SCAN`).
 
-## The COSTS view
-
-Press `c` on the dashboard (`/dispatch`). The view is verdict-first:
-
-```
-┌ pi-dispatch ── COSTS · Aug 2026 (mtd) ──────────────────────────────────────────┐
-│ VERDICT  kimi-allegro is SAVING ~$4.89 est. this month                          │
-│   plan price (prorated) $99.00 → $39.60 · plan runs @ API ~$44.49 est.          │
-│ daily  ▁▁▂▃▂▅▇▃▂▁·▁▂▂▃█▄▂▁▁▂▃▂▁▁▂▄▃▂▁  Σ ≥$12.41 · max $3.10/d                  │
-│   FLOW              RUNS  TOKENS  COST           API-EQUIV                      │
-│ › triage             41   12.4M  ≥$8.02         —                               │
-│   nightly-sync       28    6.1M  plan:kimi      ~$3.90 est.                     │
-│ plans   kimi-allegro $99/mo · 28 runs · ~$3.54 est./run amortized               │
-│         peak 5h rolling window: 9 runs, 12.4M tok — limit undisclosed by vendor │
-│ ~ estimates at pi-ai 0.80.7 · 2 runs unmetered · 3 not repriceable              │
-│ [↑↓] row [f] table [t] 7d/30d/mtd [w] what-if [esc] back                        │
-└─────────────────────────────────────────────────────────────────────────────────┘
-```
-
-Keys: `t` cycles the window (7d / 30d / month-to-date), `f` cycles the rollup table (by flow, by
-model, by trigger, by repo), `w` opens the what-if on the selected flow (press again to cycle
-candidate models; `/` type-to-filters the full priced catalog), `Esc` backs out one layer at a time.
-
-The **by-trigger** table answers "which trigger burns the most" and "what did its failures cost":
-each row is a `triggers.json` entry (attributed by the same persisted index-and-type join the graph
-view uses), with a FAIL count and, when failures spent anything, the failed share beside the total.
-Runs no trigger claims stay visible as their own rows, never blended in: `(chained runs)` for
-children spawned by another run, `(manual/local)` for CLI dispatches, `(unattributed)` for forge runs
-whose recorded trigger no longer matches the current file. The **by-repo** table groups spend by the
-target repository (issue and MR numbers stripped), with `local:<folder>` targets as their own rows.
+The surface is the **insights page** ([`insights.md`](insights.md)): `/dispatch insights` writes and
+opens one self-contained file with the plan verdicts, the daily and cumulative spend charts, the
+per-flow trend panels, and the four breakdowns (by flow, by trigger, by model, by repo) — beside the
+trigger/flow topology those numbers come from. This document explains the semantics behind every
+dollar that page draws.
+
+The **by-trigger** breakdown answers "which trigger burns the most" and "what did its failures
+cost": each row is a `triggers.json` entry (attributed by the persisted index-and-type join the
+topology uses). Runs no trigger claims stay visible as their own rows, never blended in:
+`(chained runs)` for children spawned by another run, `(manual/local)` for CLI dispatches,
+`(unattributed)` for forge runs whose recorded trigger no longer matches the current file. The
+**by-repo** breakdown groups spend by the target repository (issue and MR numbers stripped), with
+`local:<folder>` targets as their own rows.
 
 ## How to read the numbers
 
@@ -100,17 +83,17 @@ second caveat too: same token profile, different tokenizers — directional only
 history gets one offer: the `$0.5–$5/job` band recorded at `OQ-002`, scaled by the flow's run count and
 labeled `unmeasured (OQ-002)`.
 
-## Without the TUI
+## The surfaces
 
-- `/dispatch insights html` — the browser surface: the same fold drawn as charts (daily spend,
-  the four breakdowns, plan verdict cards) beside the trigger/flow topology, in one self-contained
-  file ([`insights.md`](insights.md)).
-- `/dispatch costs [7d|30d|mtd]` — the same fold, plain text, same labels.
-- `/dispatch costs whatif <provider>/<model> --flow <flow>` — scripting-friendly what-if; unknown
-  models get closest-match suggestions (this is the full-catalog path). `--flow` is **required**: the
-  estimate scores one flow's median run, not a portfolio, so the command refuses without it.
+- `/dispatch insights [7d|30d|mtd]` — the page: the fold drawn as charts beside the trigger/flow
+  topology, in one self-contained file your browser opens from disk ([`insights.md`](insights.md)).
+  Over SSH the file still writes and its URL still prints.
+- `/dispatch insights whatif <provider>/<model> --flow <flow>` — the what-if: re-prices a flow's
+  recorded token profiles under another model. Unknown models get closest-match suggestions, and
+  tab completion offers the full priced catalog. `--flow` is **required**: the estimate scores one
+  flow's median run, not a portfolio, so the command refuses without it.
 - The `dispatch_costs` tool returns the fold as JSON in which **every monetary value carries its
-  `class`** — a model reading it can no more launder an estimate into a fact than the screen can.
+  `class`** — a model reading it can no more launder an estimate into a fact than the page can.
 
 ## Environment