8a81085a72
Anthropic accepts at most 4 blocks carrying cache_control per request. When the client had already spent that budget, the re-anchor added a 5th marker and the request was rejected with a non-retryable 400 that the failure path treated as an account problem, retrying the same malformed body across the whole pool until every account locked. anchorClaudeCache now normalizes bare-object content, strips the invalid cache_control carried by defer_loading tools, pins the 1h head anchors on the last system block and last cacheable tool, then trims an over-budget body to 4 markers. The trim holds those head anchors and fills the remaining slots with the tail-most message markers: a plain "keep the last four in document order" rule drops the anchors first even though they lead document order, and skipping the re-anchor at a spent budget left system/tools on the 5m default instead of 1h. Some clients send content as a single block object rather than a one-element array. Such a turn was dropped or zeroed on every leg that reads messages, silently losing conversation history. normalizeMessageContent wraps it as a one-block array on all four paths, and hasValidContent keeps it.
9Router Embeddings Tests
Unit tests for the /v1/embeddings endpoint implementation.
Setup
Install test dependencies from the tests/ directory:
cd tests/ && npm install
Running Tests
From the tests/ directory:
npm test
Or run vitest directly with npx:
npx vitest run --reporter=verbose --config ./vitest.config.js
Test Files
| File | What it tests |
|---|---|
unit/embeddingsCore.test.js |
open-sse/handlers/embeddingsCore.js — core logic: body builder, URL router, headers, handler flow |
unit/embeddings.cloud.test.js |
cloud/src/handlers/embeddings.js — cloud worker handler: auth, validation, rate limits, CORS |
Coverage Summary (59 tests)
embeddingsCore.test.js (36 tests)
buildEmbeddingsBody: single string, array, encoding_format, default floatbuildEmbeddingsUrl: openai, openrouter, openai-compatible-*, unsupported providersbuildEmbeddingsHeaders: per-provider header sets, fallback to accessTokenhandleEmbeddingsCoreinput validation: missing, wrong type, null, emptyhandleEmbeddingsCoresuccess: response format, CORS, Content-Type, callbackshandleEmbeddingsCoreerrors: 400/429/500, network error, invalid JSONhandleEmbeddingsCoretoken refresh: 401 retry, graceful fallback
embeddings.cloud.test.js (23 tests)
- CORS OPTIONS: 200 response, empty body, correct headers
- Authentication: missing key, bad format, old-format key, wrong key value, valid key
- Body validation: invalid JSON, missing model, missing input, bad model
- Happy path: single string, array, correct delegation, CORS header, machineId override
- Rate limiting: all accounts rate-limited → 503 + Retry-After, no credentials → 400
- Error propagation: non-fallback errors passed through, 429 exhausts accounts
- machineId override: validates key, rejects wrong key