delete-stale-todos

The agents-platform.md file was the largest document removed at 675 lines. It served as a directional companion for the managed VM hosting product, covering:

• Core tenancy model (one customer per VM, multiple agents per VM)
• Control plane vocabulary (CustomerVm, AgentInstance, AgentTemplate, etc.)
• Web dashboard requirements and SSR technology direction
• Authentication via Nostr login
• Routing and origin separation for agent UIs
• Update, drift control, and future eject mode
• Delivery sequence and explicit non-goals

As a living document that was meant to evolve, its removal signals that the planning phase it represented has concluded or moved to other artifacts.

This 81-line plan was marked as the highest priority call improvement and served as the prerequisite for the three other call plans. It covered:

• Replacing codec_opus.rs PCM passthrough with real Opus bindings
• Gap concealment and adaptive playout buffer logic
• Tuning and regression test coverage

Key files referenced: crates/pika-media/src/codec_opus.rs, rust/src/core/call_runtime.rs, crates/pika-media/src/jitter.rs.

At 272 lines, this was the most detailed call plan. It specified:

• A native capability bridge architecture keeping Rust as the media brain
• iOS implementation using AVAudioEngine with voice processing
• Android implementation using AudioRecord/AudioTrack with AEC/NS/AGC
• FFI contract changes (AudioPlayoutReceiver, send_audio_capture_frame)
• An eight-step incremental delivery sequence from contract freeze through manual QA
• RMP (Rust Multiplatform) alignment guardrails

The plan explicitly referenced files like CallAudioSessionCoordinator.swift, AndroidAudioFocusManager.kt, and call_runtime.rs.

This 85-line plan was the second priority in the call improvement sequence. It covered four workstreams:

1. Startup reliability via MediaFrameSubscription::wait_ready()
2. Reconnect and resubscribe state machine with bounded exponential backoff
3. Delivery semantics and telemetry hardening
4. Failure-mode test matrix for relay restart, delayed subscription, and disconnect recovery

Key files referenced: call_runtime.rs, crates/pika-media/src/network.rs, crates/pika-media/src/subscription.rs.

This 82-line plan was the third priority in the call improvement chain. It specified:

1. Decoder recovery path for iOS VideoDecoderRenderer instead of terminal failure on corrupt payloads
2. Keyframe discipline with configurable cadence and request-keyframe signaling
3. Video adaptation controls with quality levels driven by runtime stats
4. Validation and regression coverage in deterministic and E2E tests

Key files referenced: VideoDecoderRenderer.swift, VideoCaptureManager.swift, call_runtime.rs.

This 148-line document had already transitioned from an active plan to a closeout note. It cataloged:

• The production surfaces that shipped (group management, messaging, media, calls)
• Group observability via group_updated event family
• Group profile contract details
• Explicit non-goals (protocol JSON shape, event fanout, reconnect policy)
• Justification for closing the project

Since the document itself declared the project closed, removing it is a natural cleanup step.

This 108-line document proposed seven strategies for CI improvement:

1. Instrument per-binary timing
2. Consolidate runner flakes into one runner image per target
3. One binary per VM for pure tests
4. Deferred pikahut-dependent test parallelization
5. Persistent runner pool (pre-provisioned VMs on pika-build)
6. Longest-first scheduling with bin-packing
7. Identifying and splitting slow binaries

It included concrete projections (e.g., 71s sequential → 25s wall clock with 4 runners) and discussed Incus considerations and Apple path options.

This file contained the plan for implementing an iOS share extension for Pika. The full diff content was truncated in the input, but the file was 81 lines and has been removed as a stale planning artifact.

This planning document covered the Pika share feature design. Like the iOS share extension plan, its full content was truncated in the diff input. It has been removed as part of the stale TODO cleanup.

This document outlined a staged approach to CI pipeline improvements for pikaci. It has been removed alongside the related faster-ci.md document, suggesting that CI strategy has either been implemented or has moved to different planning artifacts.

This document covered plans for shared test fixtures in the pikahut testing infrastructure. Its removal indicates the testing strategy it described has either been realized or is no longer the intended approach.

This was a living document tracking the testing strategy across the project. As with the other removed files, it has been deleted as a stale planning artifact. Living documents that are no longer actively maintained become misleading over time, making this cleanup valuable for reducing documentation drift.