implemented working error handling and retry system into AIservice using Spring Retry by avvvis · Pull Request #79 · JUCanEat/java-backend

avvvis · 2026-04-12T11:44:36Z

📄 Pull Request Description

Please fill out each section below.
The goal is that reviewers can understand everything without reading the code.

🧩 What was changed?

swapped primitive error handling into a retry mechanism from Spring framework

💡 Why was it changed?

previous mechanism was primitive and working poorly

⚙️ How was it implemented?

deleted try/except and a retry loop, emplaced it with a @Retry

⚠️ Side Effects or Risks

shouldnt break anything. Perhaps could add this retry into more places in the menu processing

kingazm · 2026-05-03T17:10:50Z

@avvvis have a look at conflics and seek review pls

Copilot

Pull request overview

This PR replaces the manual “try/catch + loop” retry logic in the production AI menu parsing service with Spring Retry annotations, and enables retry support at the application level.

Changes:

Added @EnableRetry to turn on Spring Retry support.
Annotated ProdAIService.parseMenuFromImage(...) with @Retryable and added an @Recover fallback.
Adjusted the prod AI service’s Spring profile activation (from !dev to prod).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 7 comments.

File	Description
`src/main/java/com/backend/services/ProdAIService.java`	Replaces manual retry loop with Spring Retry (`@Retryable` / `@Recover`), adds retry-attempt logging, and changes profile activation.
`src/main/java/com/backend/BackendApplication.java`	Enables Spring Retry globally via `@EnableRetry`.

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

kon-mtal · 2026-05-10T19:05:27Z

+import org.springframework.retry.annotation.EnableRetry;
 import org.springframework.scheduling.annotation.EnableScheduling;

 @SpringBootApplication
 @EnableScheduling
+@EnableRetry


@avvvis please address

kon-mtal · 2026-05-10T19:01:06Z

 @Service
 @RequiredArgsConstructor
 @Slf4j
-@Profile("!dev")
+@Profile({"prod"})
 public class ProdAIService implements MenuAIService {


Will be resolved by @avvvis on conflict resolution, please look to current version in main

kon-mtal · 2026-05-10T19:03:55Z

    public List<Dish> parseMenuFromImage(byte[] imageBytes) {
+        int attempt = RetrySynchronizationManager.getContext().getRetryCount() + 1;
+        log.info("Sending menu to AI - attempt {}/5", attempt);


kon-mtal · 2026-05-10T19:03:10Z

+    @Retryable(
+            retryFor = Exception.class,
+            maxAttempts = 5,
+            backoff = @Backoff(delay = 1000, multiplier = 2)  // 1s, 2s, 4s, 8s
+    )
+    @Override


kon-mtal · 2026-05-10T19:02:48Z

+    @Retryable(
+            retryFor = Exception.class,
+            maxAttempts = 5,
+            backoff = @Backoff(delay = 1000, multiplier = 2)  // 1s, 2s, 4s, 8s


@copilot create a PR which implements your suggested "only transient client exceptions from the AI call" and consider whether the trade-offs it introduces, in terms of say, complexity, do not outweigh the benefits.

kon-mtal · 2026-05-10T19:05:09Z

-    // Helper method to convert DishDTO to Dish entity
+    @Recover
+    public List<Dish> recover(Exception e, byte[] imageBytes) {
+        log.error("All GPT retry attempts exhausted: {}", e.getMessage());


kon-mtal · 2026-05-10T19:04:31Z

+    @Retryable(
+            retryFor = Exception.class,
+            maxAttempts = 5,
+            backoff = @Backoff(delay = 1000, multiplier = 2)  // 1s, 2s, 4s, 8s
+    )
+    @Override
    public List<Dish> parseMenuFromImage(byte[] imageBytes) {
+        int attempt = RetrySynchronizationManager.getContext().getRetryCount() + 1;
+        log.info("Sending menu to AI - attempt {}/5", attempt);


@kingazm should @avvvis implement tests or does this look good to you?

kon-mtal

Add dependencies, decide if you want to add tests, fix @retryable wrapping DB persistence code, fix conflicts, ask @kingazm or @haniazipser for final review.

kon-mtal · 2026-05-10T19:01:06Z

 @Service
 @RequiredArgsConstructor
 @Slf4j
-@Profile("!dev")
+@Profile({"prod"})
 public class ProdAIService implements MenuAIService {


Will be resolved by @avvvis on conflict resolution, please look to current version in main

kon-mtal · 2026-05-10T19:02:48Z

+    @Retryable(
+            retryFor = Exception.class,
+            maxAttempts = 5,
+            backoff = @Backoff(delay = 1000, multiplier = 2)  // 1s, 2s, 4s, 8s


@copilot create a PR which implements your suggested "only transient client exceptions from the AI call" and consider whether the trade-offs it introduces, in terms of say, complexity, do not outweigh the benefits.

kon-mtal · 2026-05-10T19:03:10Z

+    @Retryable(
+            retryFor = Exception.class,
+            maxAttempts = 5,
+            backoff = @Backoff(delay = 1000, multiplier = 2)  // 1s, 2s, 4s, 8s
+    )
+    @Override


kon-mtal · 2026-05-10T19:03:55Z

    public List<Dish> parseMenuFromImage(byte[] imageBytes) {
+        int attempt = RetrySynchronizationManager.getContext().getRetryCount() + 1;
+        log.info("Sending menu to AI - attempt {}/5", attempt);


kon-mtal · 2026-05-10T19:04:31Z

+    @Retryable(
+            retryFor = Exception.class,
+            maxAttempts = 5,
+            backoff = @Backoff(delay = 1000, multiplier = 2)  // 1s, 2s, 4s, 8s
+    )
+    @Override
    public List<Dish> parseMenuFromImage(byte[] imageBytes) {
+        int attempt = RetrySynchronizationManager.getContext().getRetryCount() + 1;
+        log.info("Sending menu to AI - attempt {}/5", attempt);


@kingazm should @avvvis implement tests or does this look good to you?

kon-mtal · 2026-05-10T19:05:09Z

-    // Helper method to convert DishDTO to Dish entity
+    @Recover
+    public List<Dish> recover(Exception e, byte[] imageBytes) {
+        log.error("All GPT retry attempts exhausted: {}", e.getMessage());


kon-mtal · 2026-05-10T19:05:27Z

+import org.springframework.retry.annotation.EnableRetry;
 import org.springframework.scheduling.annotation.EnableScheduling;

 @SpringBootApplication
 @EnableScheduling
+@EnableRetry


@avvvis please address

Copilot · 2026-05-10T19:06:34Z

@kon-mtal I've opened a new pull request, #88, to work on those changes. Once the pull request is ready, I'll request review from you.

@recover

Addresses Copilot review on ProdAIService and fixes two bugs surfaced during testing. Changes pom.xml — Added spring-retry and spring-boot-starter-aop to give @EnableRetry the classpath it needs. ProdAIService.java retryFor narrowed to transient errors only (ResourceAccessException, HttpServerErrorException); deterministic failures fail fast via a new non-retryable MenuExtractionException. Bounded backoff with maxDelay = 10000. ChatClient built once in the constructor; null-safe RetrySynchronizationManager.getContext(); MIME type detected from image bytes; defensive parseCategory (no crash on unexpected GPT output); saveAll instead of per-dish loop; @recover preserves the stack trace and rethrows. Bugs fixed during testing Infinite retry on unreadable images — empty AI responses were being retried 5× per call. Now thrown as non-retryable MenuExtractionException. RabbitMQ redelivery loop — an inner @transactional on parseMenuFromImage was marking the listener's outer transaction rollback-only on failure, causing the listener's commit to fail → message NACKed → requeued forever. Removed; the listener owns the single transaction boundary. Reviewer comments CommentHow resolvedMissing Spring Retry / AOP depsAdded to pom.xmlretryFor = Exception.class too broadNarrowed to transient onlyRetryable over GPT + DB save → duplicatesSave-time failures no longer trigger retrygetContext() can be nullNull-checked@Recover drops stack traceLogs full throwable, rethrows@Profile changeReverted to !devTests for retry behaviorDeferred

avvvis · 2026-06-13T12:38:40Z

Closing as this PR is outdated.

implemented working error handling and retry system suing Spring Retry

2516c60

avvvis temporarily deployed to dev April 12, 2026 11:44 — with GitHub Actions Inactive

kingazm requested a review from kon-mtal April 12, 2026 17:39

kon-mtal requested a review from Copilot May 10, 2026 18:55

Copilot started reviewing on behalf of kon-mtal May 10, 2026 18:56 View session

Copilot AI reviewed May 10, 2026

View reviewed changes

kon-mtal requested changes May 10, 2026

View reviewed changes

Copilot AI mentioned this pull request May 10, 2026

Restrict AI menu parsing retries to transient AI exceptions only #88

Draft

avvvis closed this Jun 13, 2026

Uh oh!

Conversation

avvvis commented Apr 12, 2026

📄 Pull Request Description

🧩 What was changed?

💡 Why was it changed?

⚙️ How was it implemented?

⚠️ Side Effects or Risks

Uh oh!

kingazm commented May 3, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kon-mtal left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI commented May 10, 2026

Uh oh!

avvvis commented Jun 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

kon-mtal left a comment •

edited

Loading