Add exponential backoff retry logic for AI service failures

[Copilot]

AI service calls (Gemini, Hugging Face) fail silently on transient errors (rate limits, network timeouts), returning empty or fallback responses without retry attempts.

Changes

Core retry infrastructure

• backend/retry_utils.py: Exponential backoff decorator with configurable retry parameters (default: 3 retries, 1s→2s→4s delays)
• Structured logging for retry attempts and final failures

AI service integration

• Gemini services (ai_service.py, gemini_summary.py): Wrapped API calls with retry decorator
  • Action plan generation
  • Chat assistant
  • MLA summary generation
• Hugging Face service (hf_service.py): Wrapped all detection endpoints with retry decorator
  • Vandalism, infrastructure, flooding detection

Implementation pattern

@exponential_backoff_retry(max_retries=3, base_delay=1.0, max_delay=10.0)
async def _generate_action_plan_with_retry(issue_description: str, category: str) -> dict:
    model = genai.GenerativeModel('gemini-1.5-flash')
    response = await model.generate_content_async(prompt)
    return json.loads(response.text.strip())

async def generate_action_plan(issue_description: str, category: str, image_path: Optional[str] = None) -> dict:
    try:
        return await _generate_action_plan_with_retry(issue_description, category)
    except Exception as e:
        logger.error(f"All retries exhausted: {e}", exc_info=True)
        return _get_fallback_action_plan(issue_description, category)

Testing

• 10 unit tests covering retry timing, exhaustion, fallback patterns
• Integration tests validating end-to-end behavior

Behavior

• Transient failures trigger automatic retry with exponential backoff
• After retry exhaustion, returns well-formed fallback responses (never empty/error states)
• No changes to function signatures or response formats

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Incomplete Error Handling for AI Service Failures</issue_title>
<issue_description>If AI APIs fail (e.g., rate limits, network issues), the application continues without fallback logic, potentially returning empty or incorrect responses.
Severity: High
Recommendation: Implement retry logic with exponential backoff and default responses for AI-dependent features.</issue_description>

<agent_instructions>monitor the assignees code quality and overalll projecct flow shouldnt break</agent_instructions>

Comments on the Issue (you are @copilot in this section)

• Fixes Incomplete Error Handling for AI Service Failures #70

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Updated (UTC)
❌ Deployment failed View logs	fixingbharat	bd49c76	Jan 09 2026, 04:53 PM

[Copilot]

Pull request overview

This pull request adds exponential backoff retry logic to AI service calls to handle transient failures like rate limits and network timeouts. The implementation introduces a new retry utility module with decorators that wrap existing AI service calls from Gemini and Hugging Face APIs. Upon retry exhaustion, services return well-formed fallback responses instead of failing silently.

Changes:

• Created core retry infrastructure with exponential backoff decorator supporting configurable retry parameters
• Integrated retry logic into all Gemini AI services (action plan generation, chat assistant, MLA summary generation)
• Integrated retry logic into all Hugging Face detection endpoints (vandalism, infrastructure, flooding)
• Added comprehensive unit and integration tests covering retry timing, exhaustion, and fallback patterns

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
backend/retry_utils.py	New module providing async and sync exponential backoff retry decorators with structured logging
backend/ai_service.py	Wrapped Gemini action plan and chat functions with retry decorators, added fallback helpers
backend/gemini_summary.py	Wrapped MLA summary generation with retry decorator, added fallback response
backend/hf_service.py	Wrapped Hugging Face API requests with retry decorator, improved error handling and logging
tests/test_retry_logic.py	Unit tests for retry decorator covering success, failure, timing, and exception handling scenarios
tests/manual_integration_test.py	Manual integration tests demonstrating retry behavior with simulated API failures

---

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

[Copilot]

Import of 'Optional' is not used.

Suggested change

	from typing import TypeVar, Callable, Optional, Tuple, Type
	from typing import TypeVar, Callable, Tuple, Type

[Copilot]

Import of 'asyncio' is not used.

Suggested change

import asyncio

[Copilot]

Import of 'AsyncMock' is not used.
Import of 'MagicMock' is not used.
Import of 'patch' is not used.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

---

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

[Copilot]

The base64 import at module level is unnecessary since it's only used within the _make_request_with_retry function. Moving it there (line 34) would be more appropriate, or keep it at module level and remove the duplicate import location.

[Copilot]

The json import at module level is redundant since it was already used in the original code within the try block. Consider keeping it at module level throughout for consistency.

Suggested change

import json

[Copilot]

The timing tolerances (±0.02-0.05s) may be too strict for CI/CD environments with variable load. Consider widening these ranges or using a more flexible approach to verify exponential growth pattern rather than absolute timing values.