checkpoint

⚒️ Generated with [Fabro](https://fabro.sh)

stages/004-preflight_lint@1/output.log

@@ -0,0 +1 @@
+blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126

stages/004-preflight_lint@1/script_timing.json

@@ -0,0 +1,8 @@
+{
+  "output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126",
+  "exit_code": 0,
+  "duration_ms": 165282,
+  "termination": "exited",
+  "output_bytes": 0,
+  "live_streaming": false
+}

stages/004-preflight_lint@1/status.json

@@ -0,0 +1,6 @@
+{
+  "outcome": "succeeded",
+  "notes": "Script completed: cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1",
+  "failure_reason": null,
+  "timestamp": "2026-07-23T17:02:21.005635430Z"
+}

stages/005-implement@1/prompt.md

@@ -0,0 +1,134 @@
+Goal: <proposed_plan>
+# Shared-Checkout Parallel Execution
+
+## Summary
+
+Simplify parallel execution so every branch:
+
+- Receives an isolated context fork.
+- Executes concurrently against the same sandbox and Git checkout.
+- Creates no Git branch, worktree, checkpoint, or branch commit.
+- Always waits for every branch; remove `first_success` and `join_policy`.
+- Produces a collision-free result containing its context updates.
+
+Fan-in remains an explicit join node. Without a prompt it is a no-op barrier; with a prompt it behaves like a standard prompt stage that reduces the aggregated results without selecting a branch.
+
+## Implementation Changes
+
+### Parallel execution
+
+- Remove all parallel-specific Git setup, checkpointing, committing, cleanup, and fast-forwarding.
+- Give each branch the same `Arc<dyn Sandbox>` and inherited `internal.work_dir`.
+- Retain `max_parallel`, branch-scoped contexts, IDs, hooks, and lifecycle events.
+- Preserve outgoing-edge order in the final results regardless of completion order.
+- Define the parallel outcome as:
+  - `succeeded` when every branch succeeds.
+  - `failed` when every branch fails.
+  - `partially_succeeded` for every mixed or partial result.
+- Preserve branch ID/index when handler or task execution fails.
+- Keep the current single-node-per-branch topology.
+
+### Results and fan-in
+
+Introduce a shared typed result:
+
+```rust
+ParallelBranchResult {
+    id: String,
+    status: String,
+    context_updates: BTreeMap<String, serde_json::Value>,
+}
+```
+
+- Store the ordered array in `parallel.results` and keep `parallel.branch_count`.
+- Include updates from successful and failed branches; never merge them into the parent context at top level.
+- Offload large leaf values using existing artifact storage while keeping `parallel.results` itself a structured array.
+- Resolve nested textual `response.*` and `command.output` blob references when building downstream prompt context.
+- Emit the same typed results through `parallel.completed` and project them into `StageProjection.parallel_results`.
+- A promptless fan-in verifies results exist and succeeds with a joined-branches note.
+- A prompted fan-in uses standard prompt execution and records `response.<fan_in_id>`, `last_response`, usage, and prompt events. It performs no ranking or selection.
+
+Remove:
+
+- Heuristic and LLM branch-selection code.
+- `parallel.fan_in.best_id`, `best_outcome`, and `best_head_sha`.
+- Per-result `head_sha` and `score`.
+- The special fan-in model-usage mode.
+- Fan-in selected-branch notes and UI.
+
+### Delete obsolete Git/worktree infrastructure
+
+- Delete the parallel-only `WorktreeSandbox`, its options/callbacks, exports, tests, and sandbox trait path helpers.
+- Remove `EngineServices`’ parallel-only `GitState` and pipeline wiring.
+- Remove now-unused branch/worktree/fast-forward Git helpers and parallel ref sanitization.
+- Remove the parallel-base-checkpoint notice code.
+- Remove the now-unemitted `git.branch`, `git.worktree.added`, and `git.worktree.removed` event variants and property types.
+- Leave the server’s independent Git checkout/worktree implementation unchanged.
+- Continue normal run-level checkpointing after the parallel node; any shared workspace changes are captured together.
+
+## Public Interfaces and Documentation
+
+- Remove `join_policy` from DOT documentation, examples, fixtures, events, and UI.
+- Add a validation error directing users to remove any `join_policy` attribute.
+- Update `parallel.started` to contain only visit and branch count.
+- Update `parallel.branch.completed` to remove `head_sha`.
+- Define `ParallelBranchResult` in OpenAPI and change `StageProjection.parallel_results` from untyped objects to that typed array; regenerate Rust and TypeScript clients.
+- Update the parallel UI to show branch status and links without commit SHAs or a join-policy badge.
+- Replace the fan-in trophy/selection UI with joined-state information and the optional standard reducer transcript.
+- Rewrite the active parallel strategy and public docs around shared-checkout concurrency:
+  - Read-only behavior is best effort.
+  - Concurrent writes are allowed but entirely user-managed.
+  - Fabro performs no write enforcement, detection, or warnings.
+  - Results are available through `parallel.results`, not a workspace `parallel_results.json`.
+  - A fan-in prompt synthesizes results but never selects workspace state.
+- Remove `join_policy` from all checked-in tutorial/demo workflows and generated documentation fixtures.
+
+## Test Plan
+
+- Unit-test that every branch receives the same sandbox working directory while retaining independent contexts.
+- Verify deterministic result ordering and complete per-branch context updates, including failed branches and structured/command outputs.
+- Cover all-success, mixed, partial, all-failed, zero-branch, `max_parallel`, dry-run, and run-cancellation behavior.
+- In a temporary Git repository, run parallel branches that write distinct files and assert:
+  - Both files remain in the shared checkout.
+  - No `fabro/run/parallel/*` refs exist.
+  - No parallel worktrees, branch commits, worktree events, or fast-forward commands occur.
+- Test promptless fan-in as a no-op join and prompted fan-in as a standard reducer that sees every branch result and emits a normal response.
+- Update event serialization, store projection, API round-trip, web parser, and renderer tests for the new typed payloads and removed fields.
+- Remove the obsolete host and Daytona parallel-Git-selection tests; retain provider-independent shared-sandbox coverage.
+- Verify with:
+  - `cargo build -p fabro-api`
+  - `cd lib/packages/fabro-api-client && bun run generate && bun run typecheck`
+  - `cargo build --workspace`
+  - `ulimit -n 4096 && cargo nextest run --workspace`
+  - `cargo +nightly-2026-04-14 fmt --check --all`
+  - `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`
+  - `cd apps/fabro-web && bun test && bun run typecheck && bun run build`
+  - `cargo --locked dev docs check`
+
+## Assumptions and Unresolved Questions
+
+- The application is greenfield; removed fields, events, and attributes receive no compatibility layer.
+- Parallel branches remain single-node executions.
+- `max_parallel` remains supported.
+- Shared-workspace write races are explicitly the user’s responsibility.
+- No Git workspace warning or read-only enforcement will be added.
+- Unresolved questions: none.
+</proposed_plan>
+
+
+## Completed stages
+- **toolchain**: succeeded
+  - Script: `command -v cargo >/dev/null || { curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y && sudo ln -sf $HOME/.cargo/bin/* /usr/local/bin/; }; cargo --version 2>&1`
+  - Output:
+    ```
+    cargo 1.96.0 (30a34c682 2026-05-25)
+    ```
+- **preflight_compile**: succeeded
+  - Script: `cargo check -q --workspace 2>&1`
+  - Output: (empty)
+- **preflight_lint**: succeeded
+  - Script: `cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1`
+  - Output: (empty)
+
+
+Read the plan file referenced in the goal and implement every step. Make all the code changes described in the plan. Use red/green TDD.

stages/005-implement@1/provider_used.json

@@ -0,0 +1,6 @@
+{
+  "mode": "agent",
+  "provider": "openrouter",
+  "model": "openai/gpt-5.6-sol",
+  "reasoning_effort": "xhigh"
+}