feat: add retry logic to health check by vlordier · Pull Request #69 · Liquid4All/cookbook

vlordier · 2026-03-06T22:44:46Z

Summary

Improves on PR #58 (health monitoring) by adding retry logic:

Add 3 attempts with exponential backoff (200ms, 400ms, 800ms) to health_check()
Handles transient failures (temporary network issues, slow server startup)
Reduces false negatives when model server is temporarily unavailable

This makes health monitoring more reliable, especially during model server restarts.

Testing

Clippy clean

- Tool Result Compression: smart summarization for large tool outputs (directory listings, search results, JSON data) - Request Deduplication: 500ms debounce on send button to prevent duplicate requests from rapid clicks - Config Hot Reload: poll config file for changes, reload without restart, show toast notification - Error Boundary: timeout wrapper (120s) around tool execution, graceful error messages instead of crashing agent loop

- Add 13 tests for settings.rs (AppSettings, SamplingConfig, config hot reload) - Add 12 tests for chat.rs compression functions (truncate, compress directory, search, JSON) - Total tests: 392 (up from 365) - Coverage: 42.5% (up from 39.5%)

- Add ModelStatus tests (serialization, healthy/unhealthy states) - Add SamplingOverrides serialization tests - Add InferenceClient tests for: - LM Studio URL construction and model selection - Tool call format (NativeJson vs Pythonic) - Fallback chain exhaustion (AllModelsUnavailable) - is_retriable for various HTTP status codes - Error repair from malformed tool calls - Total tests: 417 (up from 392)

- Add tests for data_dir(), cache_dir(), resolve_db_path() - Add rotate_log_file() tests (creates rotated copies, handles missing files) - Add filter_by_enabled_servers() tests (filters correctly, handles missing/invalid config) - Add resolve_vision_model() tests (returns None without config) - Add filter_tools_by_allowlist() test (doesn't panic without config) - Add load_override_file() tests (parses config, returns empty for missing) - Total tests: 430 (up from 417) - Coverage: 43.3% (up from 42.5%)

- Add helpful debug message when LM Studio port connection fails - Distinguish between connection errors and timeouts in health_check

- Document default LM Studio port (1234) in environment configuration

- Add GitHub Actions workflow for shellcheck - Add shellcheck validation to pre-commit hook for staged .sh files This improves on PR Liquid4All#55 by adding automated shellcheck validation.

- Add 3 attempts with exponential backoff to health_check() - Improves reliability of PR Liquid4All#58 health monitoring by handling transient failures This helps avoid false negatives when model server is temporarily slow.

Copilot

Pull request overview

This PR is titled "feat: add retry logic to health check" but the actual diff contains a large initial commit of the entire LocalCowork application codebase, including Rust backend (Tauri), TypeScript/React frontend, configuration files, and CI workflows. The PR description mentions retry logic for health checks, but no health check file or retry logic is visible in the diffs.

Changes:

Full initial codebase commit for the LocalCowork Tauri desktop app (Rust backend + React frontend)
Frontend store (settingsStore.ts), UI components (MessageInput.tsx, App.tsx), and Tauri configuration
Rust modules for MCP client, inference engine, agent core, and Tauri IPC commands

Reviewed changes

Copilot reviewed 51 out of 58 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
`src/stores/settingsStore.ts`	Zustand store for settings, config watching, permissions, and sampling config
`src/components/Chat/MessageInput.tsx`	Chat input component with debounce and auto-resize
`src/App.tsx`	Root app component with onboarding, settings panel, and config reload notification
`src-tauri/tauri.conf.json`	Tauri app configuration including window, security, and bundle settings
`src-tauri/src/mcp_client/`	MCP client modules: types, transport, lifecycle, registry, errors, discovery, client
`src-tauri/src/inference/`	Inference client modules: types, errors, config, streaming, mod
`src-tauri/src/commands/`	Tauri IPC commands: session, python_env, ollama, model_download, hardware, filesystem, mod
`src-tauri/src/agent_core/`	Agent core modules: types, tool_router, tokens, response_analysis, permissions, plan_parser, plan_templates, tool_prefilter, errors, mod
`src-tauri/src/main.rs`	Tauri app entrypoint
`src-tauri/Cargo.toml`	Rust dependencies
`src-tauri/build.rs`	Tauri build script
`src-tauri/mcp-servers.json`	Empty MCP server config
`src-tauri/entitlements.plist`	macOS entitlements
`src-tauri/capabilities/default.json`	Tauri capability permissions
`examples/localcowork/.github/workflows/shellcheck.yml`	CI workflow for shellcheck
`examples/localcowork/.git-hooks/pre-commit`	Pre-commit hook with shellcheck integration
`examples/localcowork/.env.example`	Environment variable documentation

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-06T22:46:10Z

+    if (interval) {
+      clearInterval(interval);


stopConfigWatch destructures configWatchInterval from state but then checks the undeclared variable interval, which will cause a runtime ReferenceError. The condition on line 142 should reference configWatchInterval instead of interval.

Suggested change

if (interval) {

clearInterval(interval);

if (configWatchInterval) {

clearInterval(configWatchInterval);

Copilot · 2026-03-06T22:46:10Z

+  const [value, setValue] = useState("");
+  const textareaRef = useRef<HTMLTextAreaElement>(null);
+  const lastSendTimeRef = useRef<number>(0);
+  const [isDebouncing, setIsDebouncing] = useState(false);


isDebouncing is set to true when a duplicate send is detected, but there is no path that sets it back to false in that case — the setTimeout callback on line 39 does call setIsDebouncing(false), however the debounce guard returns early before lastSendTimeRef is updated, so the spinner will show for 500ms even though no send was in progress. More critically, isDebouncing is never set to true for a successful send, so the "Please wait..." placeholder and spinner only appear during the debounce-rejection window, not during an actual in-progress send. The state name isDebouncing is misleading for this behavior; consider renaming or rethinking the intended UX.

Copilot · 2026-03-06T22:46:10Z

+          <button className="toast-close" aria-label="Dismiss">
+            ×
+          </button>


The "Dismiss" button inside the toast has no onClick handler of its own. The parent <div> on line 48 has onClick={clearConfigReloadNotification}, but since the button does not call stopPropagation, clicking the button will trigger the parent's handler via event bubbling — this works, but it also means any click anywhere on the toast dismisses it, which may not be intended. More importantly, the button itself should have an explicit onClick handler so its behavior is clear and not accidentally dependent on event propagation.

Paulescu · 2026-04-28T12:00:41Z

Hi @vlordier

The intent of this PR is not clear to me.
Could you please explain it to me using your own words?

Pau

vlordier added 8 commits March 6, 2026 21:46

test: add unit tests for PR features

8799d20

- Add 13 tests for settings.rs (AppSettings, SamplingConfig, config hot reload) - Add 12 tests for chat.rs compression functions (truncate, compress directory, search, JSON) - Total tests: 392 (up from 365) - Coverage: 42.5% (up from 39.5%)

fix: improve LM Studio connection error handling

d726844

- Add helpful debug message when LM Studio port connection fails - Distinguish between connection errors and timeouts in health_check

docs: add LM Studio endpoint to .env.example

7c32538

- Document default LM Studio port (1234) in environment configuration

ci: add shellcheck to CI and pre-commit hook

1b7bec3

- Add GitHub Actions workflow for shellcheck - Add shellcheck validation to pre-commit hook for staged .sh files This improves on PR Liquid4All#55 by adding automated shellcheck validation.

feat: add retry logic to health check

0c774a7

- Add 3 attempts with exponential backoff to health_check() - Improves reliability of PR Liquid4All#58 health monitoring by handling transient failures This helps avoid false negatives when model server is temporarily slow.

Copilot AI review requested due to automatic review settings March 6, 2026 22:44

Copilot AI reviewed Mar 6, 2026

View reviewed changes

vlordier closed this Apr 28, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: add retry logic to health check#69

feat: add retry logic to health check#69
vlordier wants to merge 8 commits intoLiquid4All:mainfrom
vlordier:feat/health-check-retry

vlordier commented Mar 6, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 6, 2026

Uh oh!

Copilot AI Mar 6, 2026

Uh oh!

Copilot AI Mar 6, 2026

Uh oh!

Paulescu commented Apr 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

vlordier commented Mar 6, 2026

Summary

Testing

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

Paulescu commented Apr 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants