fix(assistant): retry provider timeouts

[omarluq]

Summary

• retry provider/client deadline timeouts that look like upstream HTTP timeouts
• preserve partial streamed progress in the UI when provider attempts ultimately fail
• add regression coverage for timeout classification and failed streamed progress

Validation

• mise exec -- go test ./internal/assistant ./internal/terminal
• mise exec -- task ci

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 2d8d7f3a-a03c-4a2b-8ff7-7eb468c3feb8

📥 Commits

Reviewing files that changed from the base of the PR and between 9ed22cb and 1783deb.

📒 Files selected for processing (2)

• internal/assistant/retry.go
• internal/assistant/retry_timeout_test.go

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes

  • Refined retry behavior for timeout/deadline errors to retry only when messages indicate provider-side timeouts.
  • Preserve and restore streamed prompt progress on failures so partial content remains visible.
  • Improve error returned when persisting partial progress fails for clearer diagnostics.

• Tests

  • Added tests covering deadline-exceeded retry behavior and visibility of streamed progress when prompts fail.

Walkthrough

The PR refines retry classification for deadline-exceeded errors, changes partial-prompt error propagation to return a distinct persistence-failure error when saving partial progress fails, and preserves streamed message progress on prompt cancellation/failure by snapshotting and selectively restoring streamed blocks.

Changes

Error handling and retry classification

Layer / File(s)	Summary
Deadline-exceeded retry classification internal/assistant/retry.go, internal/assistant/retry_timeout_test.go	ShouldRetryModelError no longer treats context.DeadlineExceeded as immediately non-retryable; it lowercases the error message, applies nonRetryableProviderMessage, and delegates DeadlineExceeded to retryableDeadlineExceeded, which returns retryable only for specific timeout/HTTP-stream phrases. Tests cover provider timeout vs. caller-deadline cases.

Partial progress error propagation

Layer / File(s)	Summary
Partial progress error propagation internal/assistant/runtime.go, internal/assistant/runtime_test.go	Removes errors.Join usage; respondWithPartialProgress returns a dedicated persistence-failure error when saving partial progress fails or otherwise returns the original prompt error. Test now asserts the exact error string.

Streaming block preservation on error

Layer / File(s)	Summary
Streaming block restoration internal/terminal/async_events.go, internal/terminal/render_test.go	Captures app.streamingBlocks before resetting state and restores previously streamed assistant/tool/custom/thinking blocks via applyFailedPromptStreamedBlocks, skipping user/branch/compaction roles. Tests verify partial progress remains visible and app.streamingBlocks is cleared.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• omarluq/librecode#42: Modifies internal/assistant/retry.go retry classification and internal/assistant/runtime.go partial-progress persistence, overlapping with deadline-exceeded and error-join behavior changes.

Poem

A deadline tapped at the burrow door,
I sniffed the message, then checked once more.
Some timeouts bounce back to try again,
While caller clocks quietly remain.
Streams keep their footprints, saved by my paw — progress stays when the prompt meets a flaw. 🐰

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'fix(assistant): retry provider timeouts' directly reflects the main objective of the PR: to retry provider timeout errors. It is concise and accurately summarizes the primary change.
Description check	✅ Passed	The description clearly relates to the changeset, outlining three specific objectives: retrying provider timeouts, preserving streamed progress on failures, and adding test coverage. These align with the file changes across retry logic, runtime handling, and new tests.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/retry-provider-timeouts

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[codecov-commenter]

Codecov Report

❌ Patch coverage is 62.50000% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 60.72%. Comparing base (f64fe71) to head (1783deb).

Files with missing lines	Patch %	Lines
internal/terminal/async_events.go	50.00%	6 Missing and 1 partial ⚠️
internal/assistant/runtime.go	16.66%	4 Missing and 1 partial ⚠️

❌ Your patch status has failed because the patch coverage (62.50%) is below the target coverage (70.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #46      +/-   ##
==========================================
+ Coverage   60.68%   60.72%   +0.03%     
==========================================
  Files         172      172              
  Lines       16901    16928      +27     
==========================================
+ Hits        10257    10280      +23     
  Misses       5624     5624              
- Partials     1020     1024       +4     

Flag	Coverage Δ
unittests	60.72% <62.50%> (+0.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@internal/assistant/retry.go`:
- Around line 119-122: The function retryableDeadlineExceeded is using a
call-site label ("request provider response") as a retry signal which causes
wrapped caller deadlines to be treated as provider retries; remove the check for
"request provider response" from retryableDeadlineExceeded and instead rely only
on provider-specific indicators (e.g. "client.timeout exceeded" and "awaiting
headers") or, better, update retryableDeadlineExceeded to inspect/unwrap the
provided error (using errors.Is/errors.As) to detect provider-specific error
types rather than matching call-site messages; adjust the function signature and
callers as needed to use the unwrapped error checks and add a clarifying comment
in retryableDeadlineExceeded about avoiding call-site labels as retry signals.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 6a816dde-a03f-4001-993b-17ac62483e09

📥 Commits

Reviewing files that changed from the base of the PR and between f64fe71 and 9ed22cb.

📒 Files selected for processing (6)

• internal/assistant/retry.go
• internal/assistant/retry_timeout_test.go
• internal/assistant/runtime.go
• internal/assistant/runtime_test.go
• internal/terminal/async_events.go
• internal/terminal/render_test.go

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud