[WIP] Consolidate Cosmos DB retry logic into shared utility

[Copilot]

Thanks for asking me to work on this. I will get started on it and keep this PR's description up to date as I form a plan and make progress.

Issue: AI-186
Summary: AI-CLONE - Consolidate Cosmos DB retry logic into shared utility
Description: As of PR #53 (Conversation Capture), the codebase now has retry logic duplicated across four Cosmos DB store modules. This should be consolidated into a shared utility to improve maintainability and consistency.
Current State
Retry logic is now implemented in:

• slack-users-store.js (existing implementation with executeWithRetry() wrapper)
• feedback-store.js (added in PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53)
• interaction-store.js (added in PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53)
• conversation-capture-store.js (added in PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53)
  Duplicated Code
• Retryable error code detection (RETRYABLE_CODES, toNumericCode())
• Exponential backoff calculation with jitter (getDelayMs())
• Policy configuration (getRetryPolicy() with test/emulator/production variations)
• Safe error formatting (formatCosmosError())
  Proposed Solution
  Extract shared retry utilities to cosmos-utils.js (which already exists with isEmulatorTarget()):
• getRetryPolicy(connectionString, endpoint, isTest)
• getDelayMs(policy, attempt)
• toNumericCode(error)
• formatCosmosError(error)
• RETRYABLE_CODES and RECONNECT_CODES constants
  This consolidation would:
• Eliminate ~150 lines of duplicated code
• Ensure retry behavior is consistent across all stores
• Simplify testing and maintenance
• Make future retry enhancements (e.g., metrics/tracing) easier
  Acceptance Criteria
• Retry utilities exported from cosmos-utils.js with comprehensive JSDoc
• All four store modules refactored to use shared utilities
• All existing tests pass
• Unit tests added for cosmos-utils retry functions
  Related
  Follow-up to PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53 (Conversation Capture). Can be picked up after that merges.
  Labels: refactoring, technical-debt

Custom Fields:
Development: {}
Epic Link: AI-143
Rank: 1|i00l0v:
Sprint: Sprint 9 (future)

Parent Issue: AI-143
Parent Summary: R-MS5 MVP Agentic Workflow delivery of bug fixes
Parent Description: Goal: Establish agent-assisted software delivery on the Fiona repo by delivering one bug-fix workflow end-to-end — from a refined Jira ticket to a review-ready PR — with governance and no local compute. This validates the full pipeline before broadening to features, tech-debt, and Dependabot in Phase 1.
Approach: An implementation of Agentic Loop Engineering — governed reasoning loops where governance is the loop's base case. A hybrid architecture: versioned GitHub Actions own the Jira contract, readiness checks, and wall-clock governance; GitHub Copilot's hosted coding agent (steered by the automate-bug-fix skill) owns branch/tests/implementation/draft-PR.
h3. Effort (child stories)

• AI-150 — PRD: agentic workflow, end-state vision + governance (prerequisite; GitHub PR PRD: Agentic Workflow #67)
• AI-148 — TDD: MVP architecture + Phase 1 expansion (done, 69b6751)
• AI-98 — HITL bug-fix implementation. Elements may still be reused when implementing this epic, but the MVP follows the TDD's hybrid Copilot approach at deliberately limited scope (TDD Decision D-5). AI-98's full Claude-on-Foundry + Durable Functions pipeline is beyond MVP scope; its governance/telemetry learnings feed Phase 1.
  h3. July 31 MVP scope
• One supervised end-to-end run on a disposable ticket in the AI project
• Bug scoped to apps/fiona-slack (the app whose CI supplies the reused pre-check suite)
• Trigger: Jira Automation → GitHub repository_dispatch
  h3. Out of scope for MVP (Phase 1+)
• Prompt-injection defense (FR-CROSS-13) and protected-file scoping (FR-CROSS-14)
• Structured OpenTelemetry observability
• Feature, tech-debt, and Dependabot workflows

You have been assigned to a work item "AI-CLONE - Consolidate Cosmos DB retry logic into shared utility". Analyze the details of the work item and get started.
Summary: AI-CLONE - Consolidate Cosmos DB retry logic into shared utility
Description: As of PR #53 (Conversation Capture), the codebase now has retry logic duplicated across four Cosmos DB store modules. This should be consolidated into a shared utility to improve maintainability and consistency.
Current State
Retry logic is now implemented in:

• slack-users-store.js (existing implementation with executeWithRetry() wrapper)
• feedback-store.js (added in PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53)
• interaction-store.js (added in PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53)
• conversation-capture-store.js (added in PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53)
  Duplicated Code
• Retryable error code detection (RETRYABLE_CODES, toNumericCode())
• Exponential backoff calculation with jitter (getDelayMs())
• Policy configuration (getRetryPolicy() with test/emulator/production variations)
• Safe error formatting (formatCosmosError())
  Proposed Solution
  Extract shared retry utilities to cosmos-utils.js (which already exists with isEmulatorTarget()):
• getRetryPolicy(connectionString, endpoint, isTest)
• getDelayMs(policy, attempt)
• toNumericCode(error)
• formatCosmosError(error)
• RETRYABLE_CODES and RECONNECT_CODES constants
  This consolidation would:
• Eliminate ~150 lines of duplicated code
• Ensure retry behavior is consistent across all stores
• Simplify testing and maintenance
• Make future retry enhancements (e.g., metrics/tracing) easier
  Acceptance Criteria
• Retry utilities exported from cosmos-utils.js with comprehensive JSDoc
• All four store modules refactored to use shared utilities
• All existing tests pass
• Unit tests added for cosmos-utils retry functions
  Related
  Follow-up to PR [AI-129] Capture all conversations in Cosmos DB for evaluation #53 (Conversation Capture). Can be picked up after that merges.

User: You have been assigned to a work item "AI-CLONE - Consolidate Cosmos DB retry logic into shared utility". Analyze the details of the work item and get started.

Agent: I received your request and will let you know when the PR is ready.

Agent: checking if I have everything I need to get started.

User: Ed-Fi-Alliance-OSS/Fiona/

Branch naming

This task relates to issue AI-186. The branch name and pull request title MUST include AI-186 as a distinct token for traceability — this is required, not optional. The repository's branch-naming pattern is authoritative: the branch name must satisfy it in full, and you must never drop or rewrite a prefix the pattern requires — even when that prefix is given to you only as a pattern rather than spelled out here. Place AI-186 as early as the pattern permits: immediately after any required leading prefix, or as the first segment when the pattern requires none. Follow it with a short kebab-case description of the task. Do not use AI-186 as the entire branch name, do not prepend it with a leading hyphen, and do not place it at the end. Only if the pattern cannot admit AI-186 at all should you comply with the pattern and omit it.

Include a link to the original issue in the pull request description: https://edfi.atlassian.net/browse/AI-186