@herbertgao/pi-extensions 2026.10.0 → 2026.10.1

package/node_modules/@dietrichgebert/ponytail/scripts/uninstall.js

@@ -1,9 +1,10 @@
 #!/usr/bin/env node
 // ponytail — removes state ponytail wrote outside the plugin's own files:
-// the mode flag, the config file, the statusLine entry it added to
-// settings.json, and its entries in ~/.cursor/hooks.json. Plugin files
-// themselves are removed by each host's own uninstall command (see README);
-// this only cleans up what those commands can't see.
+// the mode flag, the config file, the statusline script copy, the
+// statusLine entry it added to settings.json, and its entries in
+// ~/.cursor/hooks.json. Plugin files themselves are removed by each host's
+// own uninstall command (see README); this only cleans up what those
+// commands can't see.
 
 const fs = require('fs');
 const os = require('os');
@@ -11,7 +12,7 @@ const path = require('path');
 const { getConfigPath, getClaudeDir } = require('../hooks/ponytail-config');
 const cursorHooks = require('./cursor-hooks');
 
-const STATUSLINE_SCRIPT = 'ponytail-statusline';
+const STATUSLINE_SCRIPT = /(?:^|[\s"'\\/])ponytail-statusline\.(?:sh|ps1)(?=$|[\s"'])/;
 
 function removeIfExists(filePath, label) {
   try {
@@ -23,8 +24,21 @@ function removeIfExists(filePath, label) {
 }
 
 removeIfExists(path.join(getClaudeDir(), '.ponytail-active'), 'mode flag');
+removeIfExists(path.join(getClaudeDir(), '.ponytail-statusline-nudged'), 'statusline nudge flag');
 removeIfExists(path.join(os.homedir(), '.cursor', '.ponytail-active'), 'Cursor mode flag');
+removeIfExists(path.join(os.homedir(), '.qoder', '.ponytail-active'), 'Qoder mode flag');
+for (const dir of [path.join(getClaudeDir(), 'ponytail-modes'), path.join(os.homedir(), '.cursor', 'ponytail-modes'), path.join(process.env.CODEBUDDY_CONFIG_DIR || path.join(os.homedir(), '.codebuddy'), 'ponytail-modes')]) {
+  if (fs.existsSync(dir)) {
+    fs.rmSync(dir, { recursive: true });
+    console.log(`Removed per-project mode flags: ${dir}`);
+  }
+}
+removeIfExists(path.join(process.env.CODEBUDDY_CONFIG_DIR || path.join(os.homedir(), '.codebuddy'), '.ponytail-active'), 'CodeBuddy mode flag');
 removeIfExists(getConfigPath(), 'config file');
+// Statusline script copies the activate hook keeps in the config dir (#1032).
+for (const name of ['ponytail-statusline.sh', 'ponytail-statusline.ps1']) {
+  removeIfExists(path.join(getClaudeDir(), name), 'statusline script');
+}
 
 // Cursor hooks (#817): drop only ponytail's entries from ~/.cursor/hooks.json,
 // keep every other hook the user configured there.
@@ -49,12 +63,12 @@ try {
   // (e.g. caveman && ponytail), keep the other plugin's command intact.
   // ponytail: splits on && / ; to detect other segments — good enough; a user
   // piping statuslines together is on their own.
-  if (typeof cmd === 'string' && cmd.includes(STATUSLINE_SCRIPT)) {
+  if (typeof cmd === 'string' && STATUSLINE_SCRIPT.test(cmd)) {
     const parts = cmd
       .split(/&&|;/)
       .map((s) => s.trim())
       .filter(Boolean);
-    const others = parts.filter((s) => !s.includes(STATUSLINE_SCRIPT));
+    const others = parts.filter((s) => !STATUSLINE_SCRIPT.test(s));
     if (others.length === 0) {
       delete settings.statusLine;
       fs.writeFileSync(settingsPath, JSON.stringify(settings, null, 2), 'utf8');

package/node_modules/@dietrichgebert/ponytail/skills/plugin.json

@@ -0,0 +1,5 @@
+{
+  "name": "ponytail",
+  "description": "Lazy senior dev mode. Forces the simplest solution that works: YAGNI, stdlib first, one line over fifty.",
+  "skills": "."
+}

package/node_modules/@dietrichgebert/ponytail/skills/ponytail-audit/SKILL.md

@@ -19,6 +19,7 @@ Same as ponytail-review:
 - `delete:` dead code, unused flexibility, speculative feature. Replacement: nothing.
 - `stdlib:` hand-rolled thing the standard library ships. Name the function.
 - `native:` dependency or code doing what the platform already does. Name the feature.
+- `reuse:` equivalent helper, util, or pattern already in this repo. Name the path.
 - `yagni:` abstraction with one implementation, config nobody sets, layer with one caller.
 - `shrink:` same logic, fewer lines. Show the shorter form.
 
@@ -26,11 +27,15 @@ Same as ponytail-review:
 
 Deps the stdlib or platform already ships, single-implementation interfaces,
 factories with one product, wrappers that only delegate, files exporting one
-thing, dead flags and config, hand-rolled stdlib.
+thing, dead flags and config, hand-rolled stdlib, helpers duplicating an
+equivalent that already lives in this repo.
+Before emitting `delete:`, grep the whole tree for the symbol, including tests,
+fixtures and string or dynamic references.
 
 ## Output
 
-One line per finding, ranked: `<tag> <what to cut>. <replacement>. [path]`.
+One line per finding, numbered and ranked: `<N>. <tag> <what to cut>. <replacement>. [path]`,
+so the user can say "fix 2 and 5".
 End with `net: -<N> lines, -<M> deps possible.` Nothing to cut: `Lean already. Ship.`
 
 ## Boundaries

package/node_modules/@dietrichgebert/ponytail/skills/ponytail-debt/SKILL.md

@@ -17,7 +17,7 @@ can't quietly become permanent.
 Grep the repo for comment markers, skipping `node_modules`, `.git`, and build
 output:
 
-`grep -rnE '(#|//) ?ponytail:' .`  (add other comment prefixes if your stack uses them)
+`grep -rnE --exclude-dir=.git --exclude-dir=node_modules --exclude-dir=dist --exclude-dir=build '(#|//|/[*]) ?ponytail:' .`  (add other comment prefixes if your stack uses them)
 
 Each hit is one ledger row. The comment prefix keeps prose that merely mentions
 the convention out of the ledger.

package/node_modules/@dietrichgebert/ponytail/skills/ponytail-gain/SKILL.md

@@ -2,7 +2,7 @@
 name: ponytail-gain
 description: >
   Show ponytail's measured impact as a compact scoreboard: less code, less
-  cost, more speed, from the benchmark medians. One-shot display, not a
+  cost, more speed, from the agentic benchmark averages. One-shot display, not a
   persistent mode, and not a per-repo number. Trigger: /ponytail-gain,
   "ponytail gain", "what does ponytail save", "show ponytail impact",
   "ponytail scoreboard".
@@ -13,24 +13,26 @@ description: >
 Display this scoreboard when invoked. One-shot: do NOT change mode, write flag
 files, or persist anything.
 
-The figures are the published benchmark medians (5 everyday tasks: email
-validator, debounce, CSV sum, countdown timer, rate limiter; three models:
-Haiku, Sonnet, Opus). They are measured, not computed from the current repo.
-Source: `benchmarks/` and the README.
+The figures are the published agentic benchmark averages: a headless Claude
+Code session (Haiku 4.5) doing 12 feature tickets on a real FastAPI + React
+repo, each task averaged over 4 runs, against the same agent without the
+skill, plus 6 safety tasks. They are measured, not computed from the current
+repo. Source: `benchmarks/results/2026-06-18-agentic.md` and the README.
 
 ## Scoreboard
 
-Render plain ASCII bars. The bar length shows the measured range; the label
-carries the exact figure:
+Render plain ASCII bars. The bar length shows ponytail as a share of the
+no-skill baseline; the label carries the exact figure:
 
 ```
-  ponytail gain                     benchmark median · 5 tasks · 3 models
+  ponytail gain                  benchmark average · 12 tasks · Haiku 4.5
 
-  Lines of code   no-skill  ████████████████████  100%
-                  ponytail  ██▌·················    6–20%   ▼ 80–94%
-  Cost            no-skill  ████████████████████  100%
-                  ponytail  █████▌··············   23–53%  ▼ 47–77%
-  Speed           ponytail  ▸ 3–6× faster
+  no-skill        ████████████████████  100%
+  Lines of code   █████████···········   46%   ▼ 54%
+  Tokens          ████████████████····   78%   ▼ 22%
+  Cost            ████████████████····   80%   ▼ 20%
+  Time            ███████████████·····   73%   ▼ 27%
+  Safety kept     100%  (validation, error handling, security)
 
   This repo:  /ponytail-debt  (shortcuts you deferred)
               /ponytail-audit (what's still cuttable)
@@ -38,7 +40,7 @@ carries the exact figure:
 
 ## Honesty boundary
 
-These are benchmark medians, not this repo. NEVER print a per-repo savings
+These are benchmark averages, not this repo. NEVER print a per-repo savings
 number ("you saved X lines/tokens here"): the unbuilt version was never
 written, so there is no real baseline to subtract from in a live repo. The
 only real per-repo figures come from `/ponytail-debt` (a counted ledger), and

package/node_modules/@dietrichgebert/ponytail/skills/ponytail-help/SKILL.md

@@ -32,9 +32,9 @@ Level sticks until changed or session end.
 | **ponytail-gain** | `/ponytail-gain` | Measured-impact scoreboard: less code, less cost, more speed. |
 | **ponytail-help** | `/ponytail-help` | This card. |
 
-Codex uses `@ponytail`, `@ponytail-review`, and `@ponytail-help`; Claude Code
-and OpenCode use the slash-command forms above (OpenCode ships all six as
-slash commands).
+In Codex CLI and the IDE extension, invoke skills with `$ponytail:ponytail`,
+`$ponytail:ponytail-review`, or `$ponytail:ponytail-help`. Claude Code and OpenCode use the
+slash-command forms above (OpenCode ships all six as slash commands).
 
 ## Deactivate
 

package/node_modules/@dietrichgebert/ponytail/skills/ponytail-review/SKILL.md

@@ -15,31 +15,36 @@ to cut, what replaces it. The diff's best outcome is getting shorter.
 
 ## Format
 
-`L<line>: <tag> <what>. <replacement>.`, or `<file>:L<line>: ...` for
-multi-file diffs.
+`<N>. L<line>: <tag> <what>. <replacement>.`, or `<N>. <file>:L<line>: ...` for
+multi-file diffs. Number findings `1.`, `2.`, ... across the whole report, so the
+user can say "fix 2 and 5".
 
 Tags:
 
 - `delete:` dead code, unused flexibility, speculative feature. Replacement: nothing.
 - `stdlib:` hand-rolled thing the standard library ships. Name the function.
 - `native:` dependency or code doing what the platform already does. Name the feature.
+- `reuse:` equivalent helper, util, or pattern already in this repo. Name the path.
 - `yagni:` abstraction with one implementation, config nobody sets, layer with one caller.
 - `shrink:` same logic, fewer lines. Show the shorter form.
 
+
 ## Examples
 
 ❌ "This EmailValidator class might be more complex than necessary, have you
 considered whether all these validation rules are needed at this stage?"
 
-✅ `L12-38: stdlib: 27-line validator class. "@" in email, 1 line, real validation is the confirmation mail.`
+✅ `1. L12-38: stdlib: 27-line validator class. "@" in email, 1 line, real validation is the confirmation mail.`
+
+✅ `2. L4: native: moment.js imported for one format call. Intl.DateTimeFormat, 0 deps.`
 
-✅ `L4: native: moment.js imported for one format call. Intl.DateTimeFormat, 0 deps.`
+✅ `3. L18-29: reuse: slugify helper duplicates src/lib/slug.ts slugify. Delete it, import the existing one.`
 
-✅ `repo.py:L88: yagni: AbstractRepository with one implementation. Inline it until a second one exists.`
+✅ `4. repo.py:L88: yagni: AbstractRepository with one implementation. Inline it until a second one exists.`
 
-✅ `L52-71: delete: retry wrapper around an idempotent local call. Nothing replaces it.`
+✅ `5. L52-71: delete: retry wrapper around an idempotent local call. Nothing replaces it.`
 
-✅ `L30-44: shrink: manual loop builds dict. dict(zip(keys, values)), 1 line.`
+✅ `6. L30-44: shrink: manual loop builds dict. dict(zip(keys, values)), 1 line.`
 
 ## Scoring
 

package/node_modules/@narumitw/pi-caffeinate/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@narumitw/pi-caffeinate",
-  "version": "0.49.8",
+  "version": "0.49.9",
   "description": "Pi extension that keeps the computer awake while the agent is running.",
   "type": "module",
   "license": "MIT",
@@ -35,8 +35,8 @@
     "@earendil-works/pi-coding-agent": "*"
   },
   "devDependencies": {
-    "@earendil-works/pi-coding-agent": "0.86.0",
-    "@types/node": "26.6.2",
+    "@earendil-works/pi-coding-agent": "1.0.0",
+    "@types/node": "26.6.4",
     "esbuild": "0.28.2",
     "typescript": "7.0.2"
   },
@@ -47,6 +47,6 @@
   },
   "dependencies": {
     "@narumitw/pi-tui-kit": "^0.59.0",
-    "dbus-native": "^0.15.2"
+    "dbus-native": "^0.16.0"
   }
 }

package/node_modules/pi-typesafe/README.md

@@ -142,7 +142,7 @@ if (!answer.ok) return { skipped: answer.errorCode === "budget" };  // never thr
 
 Your extension owns its own user consent and budget; `/typesafe enable` applies only to this package's tool. Check `authState()` rather than your own consent flag before you report that judgments are on.
 
-Judgments can also go through OpenRouter: `createTypeSafe({ backend: "openrouter" })` sends them to `openrouter.ai` with the key from `OPENROUTER_API_KEY`. Command Code serves the same protocol: `createTypeSafe({ backend: "commandcode" })` sends them to `api.commandcode.ai` with the key from `COMMANDCODE_API_KEY` and the model `typesafe/jev`. Or pass a `BackendEndpoint` object as `backend` to name any host that serves the protocol: it brings its own key variable, never reads `TYPESAFE_API_KEY`, and `backendHost()` reports its destination host for your consent text. These backends have no login store, so `/typesafe login` does not apply to them. Pass the same `backend` to `authState`, `keySituation`, and `ensureApiKey`, or the status you report describes the TypeSafe key while the requests use another one. The `/typesafe` commands and the `typesafe_evaluate` tool always use the TypeSafe backend. Every export — the client, `ask`, batching, the usage ledger, auth state, and the `pi-typesafe/calibrate` and `pi-typesafe/ui` entry points — is in [docs/api.md](docs/api.md).
+Judgments can also go through OpenRouter: `createTypeSafe({ backend: "openrouter" })` sends them to `openrouter.ai` with the key from `OPENROUTER_API_KEY`. Command Code serves the same protocol: `createTypeSafe({ backend: "commandcode" })` sends them to `api.commandcode.ai` with the key from `COMMANDCODE_API_KEY` and the model `typesafe/jev`. So does Liquid AI: `createTypeSafe({ backend: "liquid" })` sends them to `api.liquid.ai` with the key from `LIQUID_API_KEY` and the model `d1:free`, which is free; pass `model: "d1"` for the paid model, priced at $0.040 per million input tokens. Liquid counts the state once per question, where TypeSafe counts it once per request, so a request with N questions costs about N times the state on Liquid. Speed, measured on 2026-10-03 with one request at a time: `d1:free` took a median of about 14 seconds and sometimes failed with HTTP 504 after about 46 seconds, while `d1` took about 0.5 seconds. With `d1:free`, raise `timeoutMs` above its 15-second default, or pass `model: "d1"`. Or pass a `BackendEndpoint` object as `backend` to name any host that serves the protocol: it brings its own key variable, never reads `TYPESAFE_API_KEY`, and `backendHost()` reports its destination host for your consent text. These backends have no login store, so `/typesafe login` does not apply to them. Pass the same `backend` to `authState`, `keySituation`, and `ensureApiKey`, or the status you report describes the TypeSafe key while the requests use another one. The `/typesafe` commands and the `typesafe_evaluate` tool always use the TypeSafe backend. Every export — the client, `ask`, batching, the usage ledger, auth state, and the `pi-typesafe/calibrate` and `pi-typesafe/ui` entry points — is in [docs/api.md](docs/api.md).
 
 ## Development
 

package/node_modules/pi-typesafe/dist/backends.d.ts

@@ -1,6 +1,6 @@
-export type TypeSafeBackend = "typesafe" | "openrouter" | "commandcode";
+export type TypeSafeBackend = "typesafe" | "openrouter" | "commandcode" | "liquid";
 export interface BackendConfig {
-    /** Human name for status lines: "TypeSafe", "OpenRouter", "Command Code". */
+    /** Human name for status lines: "TypeSafe", "OpenRouter", "Command Code", "Liquid AI". */
     label: string;
     host: string;
     /** The environment variable that carries this backend's key. Absent means the TypeSafe key resolution applies. */
@@ -16,6 +16,11 @@ export interface BackendConfig {
     /** Whether the model list checks the key. A public list accepts any key, so it proves nothing. Absent means it does. */
     modelsVerifyKey?: boolean;
 }
+/** A registry entry: a backend config that can also name a price per model. Caller-supplied endpoints carry none. */
+export interface RegistryBackendConfig extends BackendConfig {
+    /** USD per million input tokens, by the model id as it is sent. A model with no entry is priced at DEFAULT_USD_PER_MTOK. */
+    prices?: Readonly<Record<string, number>>;
+}
 /** A caller-supplied endpoint that serves the Jev decisions protocol. Passed per call; never added to the registry. */
 export interface BackendEndpoint extends BackendConfig {
     /** Required. The environment variable with this endpoint's key. Must not be TYPESAFE_API_KEY. */
@@ -26,7 +31,7 @@ export interface BackendEndpoint extends BackendConfig {
 /** A registry name or a caller-supplied endpoint. */
 export type BackendSpec = TypeSafeBackend | BackendEndpoint;
 /** A backend resolved and validated: what the client will actually use. */
-export interface ResolvedBackend extends BackendConfig {
+export interface ResolvedBackend extends RegistryBackendConfig {
     /** Registry name, absent for a caller-supplied endpoint. */
     readonly name?: TypeSafeBackend;
     /** Origin only: scheme, host, and port. */
@@ -42,9 +47,9 @@ export declare const DEFAULT_BACKEND: TypeSafeBackend;
 /** The environment variable and login store that the default backend reads. */
 export declare const TYPESAFE_KEY_ENV = "TYPESAFE_API_KEY";
 /** Registry of known judgment backends. Callers pass a `BackendEndpoint` for a host this registry does not name. */
-export declare const DECISIONS_BACKENDS: Record<TypeSafeBackend, BackendConfig>;
+export declare const DECISIONS_BACKENDS: Record<TypeSafeBackend, RegistryBackendConfig>;
 /** The registry entry for a backend name; a `configuration` error for a name the registry does not know. */
-export declare function backendConfig(name: TypeSafeBackend): BackendConfig;
+export declare function backendConfig(name: TypeSafeBackend): RegistryBackendConfig;
 /**
  * The model id to send for a caller's `model` on this backend. OpenRouter routes a bare Jev id under the `typesafe`
  * author: `jev-latest` becomes its alias form `~typesafe/jev-latest`, and a bare `jev-<major>.<minor>` — with or
@@ -55,6 +60,11 @@ export declare function backendConfig(name: TypeSafeBackend): BackendConfig;
 export declare function backendModelId(backend: TypeSafeBackend, model: string): string;
 /** The model a client sends when the caller names none: the backend's own default, in the form that backend accepts. */
 export declare function defaultModelId(backend: TypeSafeBackend): string;
+/**
+ * The registry's price for a model, in USD per million input tokens, or undefined when the backend names none.
+ * `model` is the id as sent. A caller-supplied endpoint has no prices.
+ */
+export declare function backendPrice(backend: ResolvedBackend, model: string): number | undefined;
 /**
  * Whether a backend's key comes from the TypeSafe resolution (`TYPESAFE_API_KEY`, then the login store) or only from
  * its own environment variable. Only the TypeSafe backend has a login store; every other backend is environment-only.

package/node_modules/pi-typesafe/dist/backends.js

@@ -26,6 +26,15 @@ export const DECISIONS_BACKENDS = {
         modelsIdField: "id",
         modelsVerifyKey: false,
     },
+    liquid: {
+        label: "Liquid AI",
+        host: "https://api.liquid.ai",
+        keyEnv: "LIQUID_API_KEY",
+        path: "/decisions/v1/systemone",
+        modelsPath: "/decisions/v1/models",
+        modelsVerifyKey: true,
+        prices: { d1: 0.04, "d1:free": 0 },
+    },
 };
 /** The registry entry for a backend name; a `configuration` error for a name the registry does not know. */
 export function backendConfig(name) {
@@ -39,6 +48,7 @@ const DEFAULT_MODEL = {
     typesafe: "jev-latest",
     openrouter: "typesafe/jev-1.13",
     commandcode: "typesafe/jev",
+    liquid: "d1:free",
 };
 /**
  * The model id to send for a caller's `model` on this backend. OpenRouter routes a bare Jev id under the `typesafe`
@@ -61,6 +71,14 @@ export function backendModelId(backend, model) {
 export function defaultModelId(backend) {
     return backendModelId(backend, DEFAULT_MODEL[backend]);
 }
+/**
+ * The registry's price for a model, in USD per million input tokens, or undefined when the backend names none.
+ * `model` is the id as sent. A caller-supplied endpoint has no prices.
+ */
+export function backendPrice(backend, model) {
+    const prices = backend.prices;
+    return prices !== undefined && Object.hasOwn(prices, model) ? prices[model] : undefined;
+}
 /**
  * Whether a backend's key comes from the TypeSafe resolution (`TYPESAFE_API_KEY`, then the login store) or only from
  * its own environment variable. Only the TypeSafe backend has a login store; every other backend is environment-only.

package/node_modules/pi-typesafe/dist/client.d.ts

@@ -2,8 +2,8 @@ import type { Fetch, Questions, SystemOneRequest, SystemOneResult } from "@types
 import type { BackendSpec } from "./backends.js";
 import type { BatchEvaluation, BatchOptions } from "./batch.js";
 import type { BlockedCap, SpendCaps, UsageLedger, UsageReport } from "./usage.js";
-export { backendHost, resolveBackend, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./backends.js";
-export type { BackendConfig, BackendEndpoint, BackendSpec, ResolvedBackend, TypeSafeBackend } from "./backends.js";
+export { backendHost, backendPrice, resolveBackend, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./backends.js";
+export type { BackendConfig, BackendEndpoint, BackendSpec, RegistryBackendConfig, ResolvedBackend, TypeSafeBackend } from "./backends.js";
 export interface TypeSafeOptions {
     /** Defaults to TYPESAFE_API_KEY, then the key saved by `/typesafe login`; never returned. */
     apiKey?: string;
@@ -23,7 +23,7 @@ export interface TypeSafeOptions {
     maxInputTokensPerDay?: number;
     /** Estimated spend per local day, in US dollars. Unlimited by default. */
     maxUsdPerDay?: number;
-    /** Price used for the cost estimate and the USD cap. Default: DEFAULT_USD_PER_MTOK. */
+    /** Price in USD per million input tokens, used for every request's cost and the USD cap; 0 means free. Default: the backend's price for the model sent, else DEFAULT_USD_PER_MTOK. */
     usdPerMTok?: number;
     /** The usage ledger; defaults to the store next to the key. Injected by tests. */
     ledger?: UsageLedger;
@@ -50,6 +50,7 @@ export interface SpendReport {
     readonly session: UsageSnapshot;
     readonly today: UsageReport;
     readonly caps: SpendCaps;
+    /** The price of this client's default model. A request that names another model may be priced differently. */
     readonly usdPerMTok: number;
     readonly blocked?: BlockedCap;
 }

package/node_modules/pi-typesafe/dist/client.js

@@ -1,12 +1,12 @@
 import { TypeSafeClient } from "@typesafe-ai/sdk";
 import { recordAuthFailure, recordAuthVerified } from "./auth.js";
-import { TYPESAFE_KEY_ENV, backendModelId, resolveBackend, usesTypesafeKey } from "./backends.js";
+import { TYPESAFE_KEY_ENV, backendModelId, backendPrice, resolveBackend, usesTypesafeKey } from "./backends.js";
 import { evaluateAll, evaluateMany } from "./batch.js";
 import { keySituation } from "./credentials.js";
 import { TypeSafeIntegrationError, safeError } from "./errors.js";
 import { DEFAULT_MAX_INPUT_BYTES, assertWithinByteLimit, prepareEvaluationRequest } from "./schema.js";
 import { DEFAULT_USD_PER_MTOK, capsFromEnvironment, estimateUsd, mergeCaps, openUsageLedger } from "./usage.js";
-export { backendHost, resolveBackend, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./backends.js";
+export { backendHost, backendPrice, resolveBackend, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./backends.js";
 /** The paths the SDK appends to whatever base URL it is given. */
 const SDK_PATH = "/v1/systemone";
 const SDK_MODELS_PATH = "/v1/models";
@@ -63,6 +63,12 @@ function positiveInteger(value, label) {
     }
     return value;
 }
+function nonNegativeNumber(value, label) {
+    if (!Number.isFinite(value) || value < 0) {
+        throw new TypeSafeIntegrationError("configuration", `${label} must be a non-negative number.`);
+    }
+    return value;
+}
 function positiveNumber(value, label) {
     if (!Number.isFinite(value) || value <= 0) {
         throw new TypeSafeIntegrationError("configuration", `${label} must be a positive number.`);
@@ -138,7 +144,7 @@ export function createTypeSafe(options = {}) {
     const timeout = positiveInteger(options.timeoutMs ?? 15_000, "timeoutMs");
     const maxInputBytes = positiveInteger(options.maxInputBytes ?? DEFAULT_MAX_INPUT_BYTES, "maxInputBytes");
     const maxRequests = positiveInteger(options.maxRequests ?? DEFAULT_MAX_REQUESTS, "maxRequests");
-    const usdPerMTok = positiveNumber(options.usdPerMTok ?? DEFAULT_USD_PER_MTOK, "usdPerMTok");
+    const callerRate = options.usdPerMTok === undefined ? undefined : nonNegativeNumber(options.usdPerMTok, "usdPerMTok");
     const caps = mergeCaps({
         maxRequests,
         ...(options.maxRequestsPerDay === undefined ? {} : { maxRequestsPerDay: positiveInteger(options.maxRequestsPerDay, "maxRequestsPerDay") }),
@@ -165,12 +171,17 @@ export function createTypeSafe(options = {}) {
         logLevel: "off",
         ...(transport ? { fetch: transport } : {}),
     });
-    const ledger = options.ledger ?? openUsageLedger({ usdPerMTok });
+    // The caller's price wins; else the backend's price for the model sent; else the default. Each request is priced alone.
+    const rateFor = (sent) => callerRate ?? backendPrice(backend, sent) ?? DEFAULT_USD_PER_MTOK;
+    const usdPerMTok = rateFor(model);
+    // The ledger's own rate only estimates input tokens that carry no recorded cost, so it follows the caller's price alone.
+    const ledger = options.ledger ?? openUsageLedger(callerRate === undefined ? {} : { usdPerMTok: callerRate });
+    let sessionUsd = 0;
     const usage = { requestsStarted: 0, requestsSucceeded: 0, requestsFailed: 0, inputTokens: 0, outputTokens: 0 };
     // Auth state writes are deduplicated: one verification record per process, one record per distinct failure.
     let verificationRecorded = false;
     let lastFailureRecorded;
-    const snapshot = () => ({ ...usage, estimatedUsd: estimateUsd(usage.inputTokens, usdPerMTok) });
+    const snapshot = () => ({ ...usage, estimatedUsd: Math.round(sessionUsd * 1e6) / 1e6 });
     const blocked = () => (usage.requestsStarted >= maxRequests ? undefined : ledger.blocked(caps));
     const typesafe = {
         getUsage: snapshot,
@@ -203,7 +214,8 @@ export function createTypeSafe(options = {}) {
         async evaluate(input, callOptions = {}) {
             const validated = prepareEvaluationRequest(input, { maxInputBytes });
             // A per-request model meets the same mapping as the client default; the schema already limited the caller's own id.
-            const body = JSON.stringify({ ...validated, model: validated.model === undefined ? model : mapModel(validated.model) });
+            const sentModel = validated.model === undefined ? model : mapModel(validated.model);
+            const body = JSON.stringify({ ...validated, model: sentModel });
             assertWithinByteLimit(body, maxInputBytes);
             if (callOptions.signal?.aborted)
                 throw new TypeSafeIntegrationError("aborted", "TypeSafe request cancelled before submission.");
@@ -227,7 +239,10 @@ export function createTypeSafe(options = {}) {
                 usage.requestsSucceeded += 1;
                 usage.inputTokens += result.usage.input_tokens;
                 usage.outputTokens += result.usage.output_tokens;
-                ledger.recordSuccess(result.usage.input_tokens, result.usage.output_tokens);
+                // Unrounded: a small request must still count toward the cap.
+                const cost = (result.usage.input_tokens * rateFor(sentModel)) / 1e6;
+                sessionUsd += cost;
+                ledger.recordSuccess(result.usage.input_tokens, result.usage.output_tokens, cost);
                 if (!verificationRecorded) {
                     verificationRecorded = true;
                     recordAuthVerified();

package/node_modules/pi-typesafe/dist/index.d.ts

@@ -1,5 +1,5 @@
-export { createTypeSafe, backendHost, resolveBackend, DEFAULT_MAX_REQUESTS, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./client.js";
-export type { TypeSafe, TypeSafeOptions, EvaluationOptions, Evaluation, UsageSnapshot, SpendReport, TypeSafeBackend, BackendConfig, BackendEndpoint, BackendSpec, ResolvedBackend, } from "./client.js";
+export { createTypeSafe, backendHost, backendPrice, resolveBackend, DEFAULT_MAX_REQUESTS, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./client.js";
+export type { TypeSafe, TypeSafeOptions, EvaluationOptions, Evaluation, UsageSnapshot, SpendReport, TypeSafeBackend, BackendConfig, BackendEndpoint, BackendSpec, RegistryBackendConfig, ResolvedBackend, } from "./client.js";
 export { ask, DEFAULT_ASK_TIMEOUT_MS } from "./ask.js";
 export type { AskAnswer, AskOptions, Judge } from "./ask.js";
 export { fanOut, DEFAULT_CONCURRENCY, evaluateMany, evaluateAll, chunkEvaluationRequest } from "./batch.js";

package/node_modules/pi-typesafe/dist/index.js

@@ -1,4 +1,4 @@
-export { createTypeSafe, backendHost, resolveBackend, DEFAULT_MAX_REQUESTS, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./client.js";
+export { createTypeSafe, backendHost, backendPrice, resolveBackend, DEFAULT_MAX_REQUESTS, DECISIONS_BACKENDS, DEFAULT_BACKEND } from "./client.js";
 export { ask, DEFAULT_ASK_TIMEOUT_MS } from "./ask.js";
 export { fanOut, DEFAULT_CONCURRENCY, evaluateMany, evaluateAll, chunkEvaluationRequest } from "./batch.js";
 export { authState, authStatePath, clearAuthState, describeAuth, recordAuthFailure, recordAuthVerified } from "./auth.js";

package/node_modules/pi-typesafe/dist/usage.d.ts

@@ -40,6 +40,10 @@ export declare function usagePath(): string;
 /** Input-token cost, rounded to a micro-dollar so the number stays readable. */
 export declare function estimateUsd(inputTokens: number, usdPerMTok: number): number;
 export declare function emptyTotals(): UsageTotals;
+/** Longest an update waits for the ledger lock. A request that cannot get it keeps its delta and adds it with the next update. */
+export declare const LOCK_WAIT_MS = 200;
+/** A lock this old belongs to a process that died while holding it; the next update takes it over. */
+export declare const LOCK_STALE_MS = 5000;
 /** The caps a headless run may set without code: `PI_TYPESAFE_MAX_USD_PER_DAY` and its siblings. */
 export declare function capsFromEnvironment(env?: NodeJS.ProcessEnv): SpendCaps;
 /** Environment first for day caps, then the explicit option; a cap is never raised by the environment. */
@@ -49,12 +53,15 @@ export interface UsageLedgerOptions {
     path?: string;
     /** Clock for the local day and for rollover; injectable for tests. */
     now?: () => Date;
+    /** Rate for input tokens that have no recorded cost: requests recorded without one, and days from an older ledger file. Default: DEFAULT_USD_PER_MTOK. */
     usdPerMTok?: number;
 }
 /**
- * One day of persisted request, token, and cost totals, plus the caps that stop the next request. The in-memory copy is
- * authoritative for this process; the file is the cross-process, across-restart record. Reads are defensive, writes are
- * atomic and best-effort, and a day rolls over on the local date, so a long eval cannot accumulate forever unnoticed.
+ * One day of persisted request, token, and cost totals, plus the caps that stop the next request. The file is the
+ * record for every process: each update adds its delta to the file's current content under a lock, and `today()`,
+ * `blocked()`, and `describe()` read the file, so a cap counts every process that writes it. A delta that cannot be
+ * written in time (the lock is held, or the file is unwritable) stays in this process and is added with its next
+ * update. Reads are defensive, writes are atomic and best-effort, and a day rolls over on the local date.
  */
 export interface UsageLedger {
     readonly path: string;
@@ -63,11 +70,14 @@ export interface UsageLedger {
     today(): UsageReport;
     /** Count the attempt before it is submitted; a request that never returns still counts. */
     recordStart(): void;
-    recordSuccess(inputTokens: number, outputTokens: number): void;
+    /** `usd` is this request's cost at its own price; without it the request is estimated at the ledger's rate. */
+    recordSuccess(inputTokens: number, outputTokens: number, usd?: number): void;
     recordFailure(): void;
     /** The reached day cap that blocks the next request, or undefined. The caller owns the caps. */
     blocked(caps: SpendCaps): BlockedCap | undefined;
     /** One line for status output: today's requests, tokens, and cost, with the caps that apply. */
     describe(caps?: SpendCaps): string;
+    /** Add to the file any delta this process still holds. Updates do this themselves; call it before the process exits. */
+    flush?(): void;
 }
 export declare function openUsageLedger(options?: UsageLedgerOptions): UsageLedger;