speakeasy-api / speakeasy-api/kit
Include structured retry exhaustion accounting and correlation IDs in terminal failures
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 40
- Forks
- 4
- Avg merge
- 4h 18m
- Merged PRs (30d)
- 148
Description
Summary
Terminal provider failures do not expose a complete structured retry summary. Attempt count can appear in rendered text and fatal records already have session/event identity, but cumulative backoff and correlation data are not propagated together as typed exhaustion metadata to the failing tool boundary.
Split from #35 after the transient overload retry policy was fixed by #32.
Desired behavior
On retry exhaustion, provide a structured, sanitized summary containing:
- total attempts;
- cumulative backoff or sleep time;
- total elapsed retry time;
- final retry/exhaustion category; and
- session ID and event ID, or an equivalent stable correlation reference to the fatal record.
The parent should not need to parse a rendered error string or search ~/.kit/errors to correlate the tool failure with its fatal record. This issue is about terminal accounting; live progress belongs in a separate issue.
Suggested coverage
- Use deterministic time to verify attempts, cumulative delays, and elapsed time.
- Verify the parent-visible structured failure correlates with the schema-v2 fatal record.
- Cover
subagent(...)throughcompose(background=true). - Assert no provider body, credentials, private prompts, or customer identifiers are exposed.
Environment
- Originally observed with Kit
0.1.96onDarwin 25.6.0 arm64. - Confirmed as a remaining gap on Kit
0.1.105onDarwin 25.5.0 arm64. - Surface: ACP; provider route:
openai-subscription.
Related work
- #35
- #32
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start at the ACP provider failure boundary and the schema-v2 fatal record, then trace how subagent(...) flows through compose(background=true). Add deterministic coverage for attempts, delays, elapsed time, correlation, and sanitization; done means the parent receives typed terminal metadata without sensitive provider data.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- backend-api-design
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 52/100