feat: tryChain + createHealthTracker — walk the chain, know when it's down - #18
Merged
Conversation
… down The chain and catalog modules already answer WHICH model to call and whether the vendor still lists it. What every hand-rolled adopter still had to build itself, differently, was the loop that actually walks the chain on failure, and a record of whether the whole thing came back empty — the gap a real outage fell through on 2026-08-28: an app's only key died, the chat routes caught it and answered a friendly 200, and /health reported healthy because it only ever checked the database. tryChain() tries each link in order and returns the first success, throwing ChainExhaustedError (naming every link's failure, not just the last) when none work. It ships no HTTP client, same as the rest of this package — attempt() is supplied by the caller and makes the real request; this only decides which link goes next. createHealthTracker() is a factory rather than a module-level global, so a consumer gets old-style shared-process state for free by creating exactly one, and tests never fight over shared state. Health only records once per whole chain walk: a success on link two is a success, not a "degraded" report on the fallback working as designed. Version 0.4.0 -> 0.5.0. Also fixes stale catomean/* org references in README and package.json left over from the bitbaum rename. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
tryChain()walks aLink[]in order and returns the first success, throwingChainExhaustedError(naming every link's failure) when all fail — closes the gap between "here's an ordered list" and an actual fallback nobody hand-rolls the same way twice. No HTTP client added:attempt()is supplied by the caller.createHealthTracker()is a small state machine (unknown → ok/degraded/down) as a factory, not a global — a consumer gets the old singleton behaviour for free (export const llmHealth = createHealthTracker()), and it's independently testable.lib/llm-health.ts+generateWithBestProvider, built this week fixing a real outage: a dead key answered a friendly 200 and/healthreported "healthy" because it only checked the database. This generalizes the fix so it doesn't stay a one-repo patch.catomean/*org refs inREADME.md/package.jsonleft over from thebitbaumrename.Test plan
npm run verifygreen locally (lint + typecheck + build + 64 tests, 15 new)