@bitbaum/ai-kit 0.9.0 → 0.11.0

package/dist/complete.d.ts

@@ -141,6 +141,19 @@ export interface CompleteOptions {
     tools?: unknown[];
     /** Extra body fields for a vendor-specific parameter. Merged last, so it can override. */
     extraBody?: Record<string, unknown>;
+    /**
+     * Extra request headers. Merged last, so a caller can override `content-type`
+     * — but NOT `authorization`, which stays the key this module resolved.
+     *
+     * Vendors ask for these and quietly change behaviour without them:
+     * OpenRouter reads `HTTP-Referer` and `X-Title` for app attribution in its
+     * public rankings, and an app that stops sending them simply disappears from
+     * that list with no error anywhere. Without this option, adopting `complete`
+     * would mean silently dropping them, which is exactly the kind of small,
+     * invisible regression that makes a shared engine feel worse than the
+     * hand-rolled client it replaced.
+     */
+    extraHeaders?: Record<string, string>;
     /** Called on each link's failure before moving on — e.g. to log which id rotted. */
     onLinkFailure?: (link: Link, error: Error) => void;
     /** Injected for tests. Defaults to global `fetch`. */

package/dist/complete.js

@@ -191,7 +191,15 @@ async function callLink(link, options, key) {
     try {
         res = await doFetch(`${link.provider.baseUrl}/chat/completions`, {
             method: "POST",
-            headers: { "content-type": "application/json", authorization: `Bearer ${key}` },
+            headers: {
+                "content-type": "application/json",
+                ...options.extraHeaders,
+                // Last on purpose. A caller may add or override any header it likes,
+                // but not this one: silently sending someone else's credential — or
+                // none — would turn a typo in a caller's header map into an auth
+                // failure blamed on the vendor.
+                authorization: `Bearer ${key}`,
+            },
             body: JSON.stringify(body),
             signal: deadline.signal,
         });

package/dist/liveness.d.ts

@@ -39,6 +39,7 @@
  * is "does the pipe carry water", not "is the model any good".
  */
 import { type CompleteOptions } from "./complete.js";
+import type { Link } from "./chain.js";
 import type { HealthTracker } from "./health.js";
 export interface LivenessResult {
     /** Did a model answer? */
@@ -64,6 +65,22 @@ export interface LivenessOptions extends Omit<CompleteOptions, "messages" | "max
      * Default 10 minutes. Set 0 to disable caching — only for a test.
      */
     minIntervalMs?: number;
+    /**
+     * Resolve the chain when a probe actually runs, rather than once at
+     * construction.
+     *
+     * For an app whose provider list lives in a DATABASE — an admin screen with
+     * enabled/default rows and per-provider keys — a chain fixed at construction
+     * is a chain frozen at process start. The probe would then keep reporting on
+     * a configuration the operator changed twenty minutes ago, which is the
+     * opposite of "the truth about right now".
+     *
+     * It is called only on a real probe, never on a cache hit, so a monitor
+     * polling this route does not also poll the database.
+     *
+     * Takes precedence over `chain` when both are given.
+     */
+    resolveChain?: () => Link[] | Promise<Link[]>;
     /** Injected for tests. Defaults to `Date.now`. */
     now?: () => number;
 }

package/dist/liveness.js

@@ -90,9 +90,13 @@ export function createLivenessProbe(options = {}) {
             }
             const started = now();
             try {
+                // Resolved here, not at construction, and only on a real probe — so a
+                // monitor polling this route does not also poll whatever backs it.
+                const chain = options.resolveChain ? await options.resolveChain() : options.chain;
                 const result = await complete({
                     timeoutMs: PROBE_TIMEOUT_MS,
                     ...options,
+                    chain,
                     messages: PROBE_MESSAGES,
                     maxTokens: PROBE_MAX_TOKENS,
                     temperature: 0,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bitbaum/ai-kit",
-  "version": "0.9.0",
+  "version": "0.11.0",
   "description": "One install for the AI layer of an app: which model to call, what to do when the vendor retires it, how to walk the fallback chain and know when none of it worked, how to read the three kinds of 429, a fair daily budget across users, headless AI form filling — and now the model registry (one SSOT for every callable id, with the paid/free boundary as a field) and the grounding harness (facts, contract, deterministic fabrication check).",
   "license": "MIT",
   "author": "Mao Nakamoto",

package/src/complete.ts

@@ -146,6 +146,19 @@ export interface CompleteOptions {
   tools?: unknown[];
   /** Extra body fields for a vendor-specific parameter. Merged last, so it can override. */
   extraBody?: Record<string, unknown>;
+  /**
+   * Extra request headers. Merged last, so a caller can override `content-type`
+   * — but NOT `authorization`, which stays the key this module resolved.
+   *
+   * Vendors ask for these and quietly change behaviour without them:
+   * OpenRouter reads `HTTP-Referer` and `X-Title` for app attribution in its
+   * public rankings, and an app that stops sending them simply disappears from
+   * that list with no error anywhere. Without this option, adopting `complete`
+   * would mean silently dropping them, which is exactly the kind of small,
+   * invisible regression that makes a shared engine feel worse than the
+   * hand-rolled client it replaced.
+   */
+  extraHeaders?: Record<string, string>;
   /** Called on each link's failure before moving on — e.g. to log which id rotted. */
   onLinkFailure?: (link: Link, error: Error) => void;
   /** Injected for tests. Defaults to global `fetch`. */
@@ -322,7 +335,15 @@ async function callLink(
   try {
     res = await doFetch(`${link.provider.baseUrl}/chat/completions`, {
       method: "POST",
-      headers: { "content-type": "application/json", authorization: `Bearer ${key}` },
+      headers: {
+        "content-type": "application/json",
+        ...options.extraHeaders,
+        // Last on purpose. A caller may add or override any header it likes,
+        // but not this one: silently sending someone else's credential — or
+        // none — would turn a typo in a caller's header map into an auth
+        // failure blamed on the vendor.
+        authorization: `Bearer ${key}`,
+      },
       body: JSON.stringify(body),
       signal: deadline.signal,
     });

package/src/liveness.ts

@@ -40,6 +40,7 @@
  */
 
 import { complete, type CompleteOptions } from "./complete.js";
+import type { Link } from "./chain.js";
 import { ChainExhaustedError } from "./attempt.js";
 import type { HealthTracker } from "./health.js";
 
@@ -71,6 +72,22 @@ export interface LivenessOptions extends Omit<
    * Default 10 minutes. Set 0 to disable caching — only for a test.
    */
   minIntervalMs?: number;
+  /**
+   * Resolve the chain when a probe actually runs, rather than once at
+   * construction.
+   *
+   * For an app whose provider list lives in a DATABASE — an admin screen with
+   * enabled/default rows and per-provider keys — a chain fixed at construction
+   * is a chain frozen at process start. The probe would then keep reporting on
+   * a configuration the operator changed twenty minutes ago, which is the
+   * opposite of "the truth about right now".
+   *
+   * It is called only on a real probe, never on a cache hit, so a monitor
+   * polling this route does not also poll the database.
+   *
+   * Takes precedence over `chain` when both are given.
+   */
+  resolveChain?: () => Link[] | Promise<Link[]>;
   /** Injected for tests. Defaults to `Date.now`. */
   now?: () => number;
 }
@@ -140,9 +157,14 @@ export function createLivenessProbe(options: LivenessOptions = {}): LivenessProb
 
       const started = now();
       try {
+        // Resolved here, not at construction, and only on a real probe — so a
+        // monitor polling this route does not also poll whatever backs it.
+        const chain = options.resolveChain ? await options.resolveChain() : options.chain;
+
         const result = await complete({
           timeoutMs: PROBE_TIMEOUT_MS,
           ...options,
+          chain,
           messages: PROBE_MESSAGES,
           maxTokens: PROBE_MAX_TOKENS,
           temperature: 0,