Define Code Bridge and browser control parity boundary

[shiny-code-bot]

Finish Line

Code Bridge and browser/control behavior has a tested Codex-fork overlay boundary before any bridge implementation is restored.

Current Status

State: Created from #397 parity ledger.

Deleted evidence from the pre-#390 tree includes code-rs/browser/, code-rs/browser/tests/local_navigation.rs, code-rs/core/src/bridge_client.rs, TUI browser history cells, and browser-session VT100 snapshots.

Current main has app-server/Desktop remote-control compatibility, but not the Every Code local Code Bridge control channel as a proven overlay.

Acceptance Criteria

• Inventory old Code Bridge/browser tests and behavior probes from fa0c33944f.
• Classify each as port, rewrite, covered, defer, or retire.
• Define the additive overlay boundary for console/error/pageview/screenshot/control events.
• Add parity fixtures for local navigation/control behavior before implementation ports.
• Do not mutate Desktop compatibility responses to carry Every Code bridge behavior without validation.
• End code-bearing slices with ./build-fast.sh passing cleanly.

Relationships

• Child of Stabilize Every Code parity on Codex CLI fork baseline #397.
• Related to Validate Codex-base Every Code in copied Codex Desktop app #387 because app-server/Desktop compatibility shapes the control boundary.

[cbusillo]

Merged the Code Bridge/browser boundary slice in #415.

Validation evidence:

• git diff --check
• ./build-fast.sh
• scoped Claude planning and review on the Define Code Bridge and browser control parity boundary #399 boundary map
• PR check: Blob size policy passed

What landed:

• docs/code-bridge-browser-parity.md classifies the old pre-feat: replace Rust substrate with Codex base #390 bridge/browser evidence as port/rewrite/defer work.
• The ledger and feature inventory now link to the Define Code Bridge and browser control parity boundary #399 boundary map.
• The boundary explicitly keeps Code Bridge as an additive Every Code overlay and separates bridge control events from Codex Desktop remote-control status/probe/request paths.

Keeping #399 open for the fixture slice. Next action: add fake bridge metadata/server/control fixtures that prove discovery, stale heartbeat handling, subscription normalization, control request/result matching, and separation from Desktop remote-control paths before implementation ports.

[cbusillo]

Merged the first Codex-aligned Code Bridge fixture slice in #417.

Validation evidence:

• cargo test -p codex-app-server --test all code_bridge_boundary -- --nocapture
• git diff --check
• ./build-fast.sh
• PR check: Blob size policy passed

What landed:

• Added an app-server v2 boundary fixture proving Code Bridge-shaped control method names (codeBridge/control, code_bridge/control) are rejected by the Codex JSON-RPC app-server path.
• Verified remoteControl/status/read still returns the current Desktop compatibility response (disabled, no environment id).
• Updated the Define Code Bridge and browser control parity boundary #399 boundary doc to prefer Codex-aligned app-server/protocol/command shapes when Codex already has a similar surface, preserving Every Code behavior through the smallest additive overlay needed for gaps.

Keeping #399 open. Next action remains the fake bridge metadata/server fixture slice for discovery, stale heartbeat handling, subscription normalization, and control request/result matching outside the Desktop remote-control path.

[cbusillo]

PR #418 landed the next Codex-aligned Code Bridge fixture slice.

What changed:

• Added codex_core::code_bridge overlay primitives for .code/code-bridge.json discovery, subscription normalization, and bridge control request/result handling.
• Added fake local WebSocket bridge fixtures covering consumer auth, subscribe frames, control_request forwarding, successful control_result, and failed control_result surfacing.
• Kept the Desktop/app-server remote-control boundary intact: bridge control frames assert no JSON-RPC method and remain separate from remoteControl/* dispatch.
• Addressed scoped Claude review findings around llmFilter casing, persistent-socket result waits, and failed result propagation.

Validation:

• cargo test -p codex-core code_bridge --lib -- --nocapture
• cargo test -p codex-core --test all code_bridge -- --nocapture
• git diff --check
• ./build-fast.sh
• GitHub Blob size policy passed on Add fake Code Bridge core fixtures #418

Keeping this issue open for the remaining local navigation/browser/TUI snapshot fixture slice.

[cbusillo]

PR #419 landed the local Code Bridge event fixture slice.

What changed:

• Added typed passive bridge event payloads for pageview, console, error, and screenshot in codex_core::code_bridge.
• Added collect_events to authenticate, subscribe, and collect bridge events without using Desktop/app-server remoteControl/*.
• Extended the fake WebSocket bridge fixture with a real loopback HTTP page, then emitted local pageview, console, and screenshot payloads for that page.
• Addressed Claude and Gemini/Antigravity review findings: pageId assertions, structural remote-control guard comment, partial event return on timeout, narrower wire-level enum visibility, camelCase clientId, and base64-aware control screenshot byte length.

Validation:

• cargo test -p codex-core code_bridge --lib -- --nocapture
• cargo test -p codex-core --test all code_bridge -- --nocapture
• git diff --check
• ./build-fast.sh
• GitHub Blob size policy passed on Add fake Code Bridge local event fixtures #419
• Scoped Claude review and Gemini/Antigravity review completed, findings addressed

Keeping this issue open for the next implementation-facing step: wire the Code Bridge tool/consumer surface to these fixtures, then add TUI snapshot coverage once the event storage/transcript shape is stable.