[COOL IDEA™] Add transactional write-enabled FUSE/CAS projection for causal copy workflows

[flyingrobots]

Problem

A read-only FUSE projection is tracked in #66, but it does not support write semantics needed by higher-level copy or migration workflows where a verified CAS view should also accept mutation traces and replay-safe commit events.

Invariant impact

Any write-capable projection must preserve authenticated source identity, explicit causal intent, and replay-safe recovery behavior (no silent patch application, no unbounded rehydration). Writes must be append-only at the projection boundary unless an explicit transaction boundary is proven.

Proposed boundary

Keep should export a narrow integration boundary that:

• presents verified projections with deterministic naming,
• records mutation intents as first-class causal events,
• maps projection writes to segment/record publication with explicit ordering and recovery checks.

Alternatives considered

1. Keep filesystem-only tooling around current read APIs only (insufficient for write/copy use-cases).
2. Keep using ad-hoc scripts for write projection.
3. Add dedicated write-enabled projection adapter with explicit transaction journal and replay validator.

Evidence and acceptance criteria

• Add a tracked design/rationale update explaining recovery and crash boundaries.
• Validate with at least one end-to-end test: mount-style projected view, write path, remap to verified CAS layout, verify unchanged identity of untouched bytes and deterministic replay on reopen.
• Add explicit negative tests for untrusted mutation attempts and partial writes.

Failure modes

• Partial FUSE write visibility without durable intent log commit.
• Replay reordering vs existing causal lineage.
• Power-loss consistency gaps between mount-level transaction and durable publication state.

[coderabbitai]

🔗 Related PRs

#50 - Implement bounded streaming CAS [merged]
#54 - Define the durable segment store protocol [merged]
#68 - Define retention roots and GC liveness [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

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

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!