v1.6.1: fix live design crash (llm_call telemetry to stdlib logger)#69
Merged
Conversation
…r (v1.6.1) The per-call telemetry log passed structlog-style kwargs (model=, tokens=, ...) to the stdlib logger returned by get_logger(). Under the CLI, which enables root INFO via configure_logging(), this raised TypeError: Logger._log() got an unexpected keyword argument 'model' right after the model responded and tokens were billed, so every live `cloudwright design`/`modify` produced a traceback and no spec. Latent since v1.1.0: pytest leaves the root logger at WARNING, so info() short- circuits and the line never emitted or crashed under tests. Switch both the Anthropic and OpenAI providers to %-style stdlib formatting. Side effect: LLM telemetry (model, latency, tokens, cache hits) now actually emits. Regression: test_llm_telemetry_logging.py runs both providers under configure_logging() and asserts no crash.
xmpuspus
added a commit
to xmpuspus/ai-workflow-benchmark
that referenced
this pull request
Jul 8, 2026
awb run --dry-run paid the adapter auth preflight, a live model call that can take 30 seconds; recording the demo surfaced it. The preview now prints without touching the adapter, pinned by a test that makes adapter resolution explode under --dry-run. demos/from-pr.gif is a real vhs recording against a real merged PR (xmpuspus/cloudwright#69): mine, inspect the pinned SHA and rubric, validate, load with run --tasks-dir. Embedded in the README's Benchmark Your Own Setup section.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Hotfix for a ship-blocking crash the July 2026 product audit surfaced.
cloudwright design(and non-streammodify) crashed on a live API key with:The per-call telemetry line passed structlog-style kwargs (
model=,tokens=, ...) to the stdlib logger returned byget_logger(). The CLI enables root INFO viaconfigure_logging(), so the call raised right after the model responded and tokens were billed, producing no spec.Latent since v1.1.0: pytest leaves the root logger at WARNING, so
info()short-circuits and the line never emitted or crashed under tests. Demos use mock LLMs; release smoke tests run keyless. Three masks aligned.Fix
Both the Anthropic and OpenAI providers now format the
llm_callline with%-style stdlib logging. Side effect: per-call LLM telemetry (model, latency, tokens, cache hits) now actually emits.Test
packages/core/tests/test_llm_telemetry_logging.pyruns both providers underconfigure_logging()(root at INFO) and asserts no crash. Confirmed RED before the fix, GREEN after. Full core suite 1372 passed at CI parity. A fresh-context adversarial review cleared the placeholder/arg counts, logger type, and test coverage.Version bumped to 1.6.1 across all four packages, core extras pins, and server.json.
Reviewed by Xavier Puspus