floe-guard 0.2.1 → 0.4.0

package/README.md

@@ -63,6 +63,27 @@ const adv = guard.advisory();
 const model = adv.nearLimit ? openai("gpt-4o-mini") : openai("gpt-4o");
 ```
 
+## Per-call spend log
+
+The guard keeps a typed, in-memory ledger of everything it priced: each
+`record()` / `settle()` appends one `SpendEvent`, and `recordTool()` lets paid
+non-LLM calls spend the same budget and land in the same log. The events sum to
+`spentUsd` (unless a `maxLogEvents` ring buffer has evicted old ones).
+
+```ts
+const guard = new BudgetGuard(1.0); // { maxLogEvents: N } caps memory
+guard.record("gpt-4o", 1_200, 350, { label: "researcher" });
+guard.recordTool("serpapi.search", 0.01, { label: "researcher" });
+
+guard.spendLog; // [{ timestamp, kind: "llm", modelOrTool: "gpt-4o", … }, …]
+process.stdout.write(guard.exportLog()); // JSONL, one event per line
+```
+
+`exportLog()` emits a stable snake_case schema —
+`{timestamp, kind: llm|tool, model_or_tool, prompt_tokens, completion_tokens,
+cost_usd, label?, reserved?}` — identical to the Python package's
+`export_log()`, so every agent produces the same shape regardless of stack.
+
 ## Compatibility
 
 `ai` is declared as a peer dependency with the range `>=4.0.0 <6.0.0`:

package/dist/index.cjs

@@ -22,7 +22,9 @@ var index_exports = {};
 __export(index_exports, {
   BudgetExceeded: () => BudgetExceeded,
   BudgetGuard: () => BudgetGuard,
+  DeadlineExceeded: () => DeadlineExceeded,
   FloeGuardError: () => FloeGuardError,
+  LatencyBudget: () => LatencyBudget,
   UnpriceableModelError: () => UnpriceableModelError,
   budgetGuardMiddleware: () => budgetGuardMiddleware,
   priceTokens: () => priceTokens,
@@ -61,6 +63,21 @@ var UnpriceableModelError = class extends FloeGuardError {
     this.model = model;
   }
 };
+function roundHalfUp(ms) {
+  return Math.floor(ms + 0.5);
+}
+var DeadlineExceeded = class extends FloeGuardError {
+  elapsedMs;
+  slaMs;
+  constructor(elapsedMs, slaMs) {
+    super(
+      `DEADLINE EXCEEDED \u2014 call blocked (elapsed ${roundHalfUp(elapsedMs)}ms of ${roundHalfUp(slaMs)}ms SLA)`
+    );
+    this.name = "DeadlineExceeded";
+    this.elapsedMs = elapsedMs;
+    this.slaMs = slaMs;
+  }
+};
 
 // src/pricing.ts
 var pricing_exports = {};
@@ -979,6 +996,9 @@ var BudgetGuard = class {
   lastCost = 0;
   /** USD held for in-flight calls (reserved, not yet settled). Counts toward the ceiling. */
   reserved = 0;
+  /** Per-call ledger, oldest first; a ring buffer when maxLogEvents is set. */
+  spendEvents = [];
+  maxLogEvents;
   /**
    * @param limitUsd the spend ceiling, in USD. `0` blocks the very first call.
    */
@@ -994,7 +1014,13 @@ var BudgetGuard = class {
         `nearLimitBps must be an integer in 0..10000, got ${nearLimitBps}`
       );
     }
+    if (options.maxLogEvents !== void 0 && (!Number.isInteger(options.maxLogEvents) || options.maxLogEvents < 0)) {
+      throw new RangeError(
+        `maxLogEvents must be a non-negative integer, got ${options.maxLogEvents}`
+      );
+    }
     this.limitUsd = limitUsd;
+    this.maxLogEvents = options.maxLogEvents;
     this.priceOverrides = options.priceOverrides;
     this.failClosed = options.failClosed ?? true;
     this.onBlock = options.onBlock ?? defaultOnBlock;
@@ -1056,7 +1082,10 @@ var BudgetGuard = class {
    * Release a reservation and record the actual cost. `record` is `settle` with
    * no reservation. Returns the USD cost of this call; unpriceable-model handling
    * matches {@link BudgetGuard.record}, and any held reservation is released even
-   * on the warn-and-skip path.
+   * on the warn-and-skip path. A priced call appends one {@link SpendEvent} to
+   * {@link BudgetGuard.spendLog} (`label` tags it, e.g. with an agent/task name);
+   * the warn-and-skip path accrues nothing and logs nothing, so the ledger stays
+   * in lockstep with `spentUsd`.
    */
   settle(model, promptTokens, completionTokens, options = {}) {
     const reserved = options.reserved ?? 0;
@@ -1093,6 +1122,18 @@ var BudgetGuard = class {
       this.spentUsd = this.limitUsd;
     }
     this.lastCost = cost;
+    this.appendEvent({
+      timestamp: Date.now() / 1e3,
+      kind: "llm",
+      modelOrTool: model,
+      promptTokens,
+      completionTokens,
+      costUsd: cost,
+      ...options.label !== void 0 ? { label: options.label } : {},
+      // 0 means "no reservation" (the plain record() path) — omit rather than
+      // log a meaningless zero.
+      ...reserved ? { reserved } : {}
+    });
     return cost;
   }
   /**
@@ -1105,9 +1146,40 @@ var BudgetGuard = class {
   record(model, promptTokens, completionTokens, options = {}) {
     return this.settle(model, promptTokens, completionTokens, {
       reserved: 0,
-      price: options.price
+      price: options.price,
+      label: options.label
     });
   }
+  /**
+   * Accrue a non-LLM cost (a paid tool/API call) against the same ceiling.
+   *
+   * Tools with direct dollar costs — search APIs, scrapers, sandboxes — spend the
+   * same budget the LLM calls do; `recordTool` folds them into `spentUsd` (so
+   * `check()` / `reserve()` see them) and appends a `kind: "tool"`
+   * {@link SpendEvent} to {@link BudgetGuard.spendLog}. The caller supplies the
+   * cost: tools have no token usage to price. Deliberately does NOT update the
+   * next-call estimate — that predicts the next *LLM* call, and a tool's price
+   * would skew it. Returns `costUsd`.
+   */
+  recordTool(tool, costUsd, options = {}) {
+    if (!Number.isFinite(costUsd) || costUsd < 0) {
+      throw new RangeError(`costUsd must be a finite, non-negative number, got ${costUsd}`);
+    }
+    this.spentUsd += costUsd;
+    if (this.spentUsd - this.limitUsd > 0 && this.spentUsd - this.limitUsd < EPS) {
+      this.spentUsd = this.limitUsd;
+    }
+    this.appendEvent({
+      timestamp: Date.now() / 1e3,
+      kind: "tool",
+      modelOrTool: tool,
+      promptTokens: null,
+      completionTokens: null,
+      costUsd,
+      ...options.label !== void 0 ? { label: options.label } : {}
+    });
+    return costUsd;
+  }
   /**
    * Drop an in-flight reservation without recording spend (e.g. the call failed
    * before producing usage). Safe to call with `0`.
@@ -1123,6 +1195,46 @@ var BudgetGuard = class {
   get remainingUsd() {
     return Math.max(0, this.limitUsd - this.spentUsd - this.reserved);
   }
+  /**
+   * The per-call spend ledger, oldest first — one {@link SpendEvent} per priced
+   * `record()` / `settle()` / `recordTool()`. Returns a snapshot copy: mutating
+   * it cannot corrupt the ledger.
+   */
+  get spendLog() {
+    return [...this.spendEvents];
+  }
+  /**
+   * The spend ledger as JSONL — one event per line, newline-terminated.
+   *
+   * The schema is stable and language-independent (snake_case keys, fixed order;
+   * optional fields omitted when absent), identical to the Python package's
+   * `export_log()`, so heterogeneous agents produce logs you can concatenate and
+   * analyse as one stream. (The *schema* is the contract, not the bytes: the two
+   * runtimes may render the same float differently, e.g. JS `0.0000025` vs
+   * Python `2.5e-06`.) Empty ledger yields `""`.
+   */
+  exportLog() {
+    return this.spendEvents.map((e) => {
+      const row = {
+        timestamp: e.timestamp,
+        kind: e.kind,
+        model_or_tool: e.modelOrTool,
+        prompt_tokens: e.promptTokens,
+        completion_tokens: e.completionTokens,
+        cost_usd: e.costUsd
+      };
+      if (e.label !== void 0) row.label = e.label;
+      if (e.reserved !== void 0) row.reserved = e.reserved;
+      return `${JSON.stringify(row)}
+`;
+    }).join("");
+  }
+  appendEvent(event) {
+    this.spendEvents.push(Object.freeze(event));
+    if (this.maxLogEvents !== void 0 && this.spendEvents.length > this.maxLogEvents) {
+      this.spendEvents.splice(0, this.spendEvents.length - this.maxLogEvents);
+    }
+  }
   /**
    * Context-aware spend advisory for this budget — see {@link BudgetAdvisory}.
    *
@@ -1154,6 +1266,68 @@ function defaultOnBlock(spentUsd, limitUsd) {
   );
 }
 
+// src/latency.ts
+var LatencyBudget = class {
+  slaMs;
+  nearDeadlineBps;
+  onBlock;
+  clock;
+  startedAt;
+  /** The budget starts counting at construction — build it when the request
+   *  (and its SLA) starts. */
+  constructor(slaMs, options = {}) {
+    if (!(Number.isFinite(slaMs) && slaMs > 0)) {
+      throw new RangeError("slaMs must be > 0");
+    }
+    const nearDeadlineBps = options.nearDeadlineBps ?? 8e3;
+    if (!Number.isInteger(nearDeadlineBps) || nearDeadlineBps < 0 || nearDeadlineBps > 1e4) {
+      throw new RangeError("nearDeadlineBps must be an integer 0..10000");
+    }
+    this.slaMs = slaMs;
+    this.nearDeadlineBps = nearDeadlineBps;
+    this.onBlock = options.onBlock;
+    this.clock = options.clock ?? (() => performance.now());
+    this.startedAt = this.clock();
+  }
+  /** Milliseconds since construction (monotonic). */
+  get elapsedMs() {
+    return this.clock() - this.startedAt;
+  }
+  /** Milliseconds left before the SLA, floored at 0 — the readable signal a
+   *  router uses to pick a faster fallback or truncate work mid-chain. */
+  get remainingMs() {
+    return Math.max(0, this.slaMs - this.elapsedMs);
+  }
+  /**
+   * Throw {@link DeadlineExceeded} when the projected elapsed time (now +
+   * `expectedMs` for the upcoming call) would blow the SLA. Call it
+   * immediately before each tool/model call; pass 0 to only gate on time
+   * already spent.
+   */
+  check(expectedMs = 0) {
+    if (!(Number.isFinite(expectedMs) && expectedMs >= 0)) {
+      throw new RangeError("expectedMs must be >= 0");
+    }
+    const elapsed = this.elapsedMs;
+    if (elapsed + expectedMs > this.slaMs) {
+      this.onBlock?.(elapsed, this.slaMs);
+      throw new DeadlineExceeded(elapsed, this.slaMs);
+    }
+  }
+  /** The soft near-deadline signal — symmetric to `BudgetGuard.advisory()`. */
+  advisory() {
+    const elapsed = this.elapsedMs;
+    const usedBps = elapsed > 0 ? Math.min(1e4, Math.round(elapsed * 1e4 / this.slaMs)) : 0;
+    return {
+      nearDeadline: usedBps >= this.nearDeadlineBps,
+      usedBps,
+      remainingMs: Math.max(0, this.slaMs - elapsed),
+      slaMs: this.slaMs,
+      elapsedMs: elapsed
+    };
+  }
+};
+
 // src/middleware.ts
 function usageTokens(modelId, usage) {
   const u = usage;
@@ -1247,7 +1421,9 @@ function budgetGuardMiddleware(guard) {
 0 && (module.exports = {
   BudgetExceeded,
   BudgetGuard,
+  DeadlineExceeded,
   FloeGuardError,
+  LatencyBudget,
   UnpriceableModelError,
   budgetGuardMiddleware,
   priceTokens,