[FederatedLearning][Post-v1] Create post-v1 backlog tracking issue

[ooples]

Track deferred federated learning items listed in docs/FederatedLearning_OutOfScope_Backlog.md.\n\nThis issue is a container/tracker; individual issues should be created for each backlog item and linked here.

[coderabbitai]

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

✅ Successfully generated implementation plan.

💡 Use @coderabbitai <feedback> to replan

Examples

• Example 1
• Example 2

🔗 Similar Issues

Related Issues

• [Phase 3] Implement Federated Learning Framework #398

👤 Suggested Assignees

• ooples

---

🧪 Issue enrichment is currently in early access.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

[ooples]

Created initial post-v1 backlog issues from docs/FederatedLearning_OutOfScope_Backlog.md:\n- #537 Full on-device confidential training in iOS/Android TEEs\n- #538 PSI for vertical FL entity alignment\n- #539 Verifiable FL (ZK proofs)\n- #540 MPC beyond secure aggregation\n- #541 Federated learning over graphs (secure protocols)\n- #542 Production-grade Vertical Federated Learning (VFL)\n\nSource of truth for v1 scope: docs/FederatedLearning_V1_FrozenScope.md

[coderabbitai]

Implementation Plan

Approach

Create two federated learning documentation files that define v1 scope boundaries and track deferred advanced features, establishing a clear reference for what's included versus postponed for post-v1 development.

Observations

The project uses a phase-based documentation pattern for scope management (seen in JIT_ROADMAP.md and DistributedTrainingImplementations.md) with status indicators (✅/⏳/❌) and hierarchical issue tracking. The codebase currently has distributed training infrastructure but no federated learning implementation. Documentation lives in docs/ with clear separation between implementation guides, roadmaps, and design documents.

Assumptions

Assumption 1: What should be defined as the v1 federated learning scope given no current implementation exists?

Options Considered:

• Start with a comprehensive federated learning system including advanced privacy features
• Define a minimal v1 scope that builds on existing distributed training infrastructure with basic FedAvg

Chosen Option: Define a minimal v1 scope that builds on existing distributed training infrastructure with basic FedAvg

Rationale: Define a focused v1 scope that leverages existing distributed training patterns (client-server, parameter aggregation, synchronous/asynchronous modes) while deferring advanced cryptographic and privacy features. This aligns with industry standard FL v1 implementations and provides a clear upgrade path to advanced features. Based on web research, v1 should focus on horizontal federated learning with basic secure aggregation, while TEEs, PSI, ZK proofs, MPC, Graph FL, and VFL are appropriately deferred.

💡 User Tips

Regenerate the plan with different choices with @coderabbitai <feedback>.

Plan

Phase 1: Create Frozen Scope Documentation

Define the v1 federated learning scope boundaries to establish what features will be included in the initial release.

Create V1 Frozen Scope Document

Create docs/FederatedLearning_V1_FrozenScope.md following the project's documentation patterns.

• Structure the document with clear sections: Overview, Scope Boundaries, Core Components, Architecture, and Out of Scope Reference
• Define v1 scope to include: horizontal federated learning, basic FedAvg algorithm, client-server architecture, synchronous/asynchronous learning modes, and basic secure aggregation (building on existing distributed training infrastructure)
• Explicitly state that this scope is frozen and no additions will be made during v1 development
• Reference the existing distributed training components (CommunicationBackends, ShardedModel/Optimizer patterns) as the foundation for FL implementation
• Include a forward reference to docs/FederatedLearning_OutOfScope_Backlog.md for deferred features
• Follow the markdown formatting style seen in docs/JIT_ROADMAP.md with clear headers and structured sections

🤖 Prompt for AI agents

Create the v1 frozen scope document that defines what's included in the initial
federated learning release.

Create docs/FederatedLearning_V1_FrozenScope.md with:
- Overview, Scope Boundaries, Core Components, Architecture, and Out of Scope
Reference sections
- Define v1 scope: horizontal federated learning, basic FedAvg algorithm,
client-server architecture, synchronous/asynchronous learning modes, and basic
secure aggregation
- State that scope is frozen with no additions during v1 development
- Reference existing distributed training components (CommunicationBackends,
ShardedModel/Optimizer patterns) as foundation
- Include forward reference to docs/FederatedLearning_OutOfScope_Backlog.md
- Follow markdown formatting style from docs/JIT_ROADMAP.md

Phase 2: Create Backlog Documentation

Document the advanced federated learning features deferred to post-v1 releases.

Create Out-of-Scope Backlog Document

Create docs/FederatedLearning_OutOfScope_Backlog.md listing deferred advanced features.

• Structure the document following the project's backlog pattern with sections for: Overview, Backlog Items, and Status Tracking
• Document each of the six deferred items using the titles from issues [FederatedLearning][Post-v1] Full on-device confidential training in iOS/Android TEEs #537-[FederatedLearning][Post-v1] Production-grade Vertical Federated Learning (VFL) #542:
  • Full on-device confidential training in iOS/Android TEEs (Trusted Execution Environments)
  • PSI for vertical FL entity alignment (Private Set Intersection)
  • Verifiable FL with zero-knowledge proofs
  • MPC beyond basic secure aggregation (Multi-Party Computation)
  • Federated learning over graphs with secure protocols
  • Production-grade Vertical Federated Learning (VFL)
• For each backlog item: provide a brief description of what the feature entails, explain why it's deferred (complexity, cryptographic requirements, or need for v1 foundation), and link to the corresponding tracking issue ([FederatedLearning][Post-v1] Full on-device confidential training in iOS/Android TEEs #537-[FederatedLearning][Post-v1] Production-grade Vertical Federated Learning (VFL) #542)
• Include status indicators (❌ Not Scheduled) following the pattern in docs/JIT_ROADMAP.md
• Add a reference back to docs/FederatedLearning_V1_FrozenScope.md to maintain bidirectional documentation links
• Note that these items will be prioritized for post-v1 releases once the foundational FL capabilities are stable

🤖 Prompt for AI agents

Create the out-of-scope backlog document that tracks deferred advanced federated
learning features.

Create docs/FederatedLearning_OutOfScope_Backlog.md with:
- Overview, Backlog Items, and Status Tracking sections
- Document six deferred items from issues #537-#542: Full on-device confidential
training in iOS/Android TEEs, PSI for vertical FL entity alignment, Verifiable
FL with zero-knowledge proofs, MPC beyond basic secure aggregation, Federated
learning over graphs with secure protocols, Production-grade Vertical Federated
Learning (VFL)
- For each item: provide brief description, explain why it's deferred
(complexity, cryptographic requirements, or need for v1 foundation), and link to
tracking issue
- Include status indicators (❌ Not Scheduled) following docs/JIT_ROADMAP.md
pattern
- Add reference back to docs/FederatedLearning_V1_FrozenScope.md
- Note that items will be prioritized post-v1 once foundational FL capabilities
are stable

---

🚀 Next Steps

🤖 All AI agent prompts combined

Task: 1

Create the v1 frozen scope document that defines what's included in the initial
federated learning release.

Create docs/FederatedLearning_V1_FrozenScope.md with:
- Overview, Scope Boundaries, Core Components, Architecture, and Out of Scope
Reference sections
- Define v1 scope: horizontal federated learning, basic FedAvg algorithm,
client-server architecture, synchronous/asynchronous learning modes, and basic
secure aggregation
- State that scope is frozen with no additions during v1 development
- Reference existing distributed training components (CommunicationBackends,
ShardedModel/Optimizer patterns) as foundation
- Include forward reference to docs/FederatedLearning_OutOfScope_Backlog.md
- Follow markdown formatting style from docs/JIT_ROADMAP.md
===============================================================================

Task: 2

Create the out-of-scope backlog document that tracks deferred advanced federated
learning features.

Create docs/FederatedLearning_OutOfScope_Backlog.md with:
- Overview, Backlog Items, and Status Tracking sections
- Document six deferred items from issues #537-#542: Full on-device confidential
training in iOS/Android TEEs, PSI for vertical FL entity alignment, Verifiable
FL with zero-knowledge proofs, MPC beyond basic secure aggregation, Federated
learning over graphs with secure protocols, Production-grade Vertical Federated
Learning (VFL)
- For each item: provide brief description, explain why it's deferred
(complexity, cryptographic requirements, or need for v1 foundation), and link to
tracking issue
- Include status indicators (❌ Not Scheduled) following docs/JIT_ROADMAP.md
pattern
- Add reference back to docs/FederatedLearning_V1_FrozenScope.md
- Note that items will be prioritized post-v1 once foundational FL capabilities
are stable

💡 Iterate on the plan with: @coderabbitai <feedback>

Example Feedback
- @coderabbitai You can skip phase 3. Add a simple unit test case for phase 2.
- @coderabbitai For assumption 1 go ahead with option 3 and replan.