feat(export): E1 — WsJsonClient extraction + bridge protocol client - #121
Merged
Conversation
📝 WalkthroughWalkthroughAdded a shared reconnecting WebSocket transport for Matter and bridge peers. Added Python bridge protocol models and ChangesBridge architecture
Estimated code review effort: 5 (Critical) | ~120 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 2 | ❌ 3❌ Failed checks (3 warnings)
✅ Passed checks (2 passed)
✨ Finishing Touches 💡 1📝 Generate docstrings 💡
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
The plugin half of BRIDGE_PROTOCOL.md: a client that speaks to the bridge node, built on the transport `matter_client.py` already had. Protocol only — no plugin.py wiring, no allow-list, no launchd (E2/E7). **`ws_json_client.py` (new, 359 lines)** — the transport pulled out of `matter_client.py`: run loop with `min(2**attempt, 30)` backoff, message_id → future correlation with a `wait_for` deadline and a `finally` pop, disconnect handling (in-flight futures fail with ConnectionError; `on_disconnect` only on a genuine drop of a live connection), and the once-per-streak repeated-failure latch. The handshake stays an overridable hook because the two peers' are genuinely different — server_info + start_listening vs hello + attach — as does frame classification, so `protocol.py`'s rename firewall stays intact: this module never names a wire field. One deliberate behaviour change (§3.4): an error response that no future is waiting for is now LOGGED at warning, not dropped. `set_state` is fire-and-forget, and an unnoticed failure there looks exactly like "the ecosystem shows stale state". Both clients gain it. `matter_client.py` drops to 147 lines — the URI, the handshake, the convenience wrappers — with an identical public API. `tests/test_matter_client.py` passes unchanged, no import edits needed. **`bridge_protocol.py` (new)** — our envelope: §3 commands, the §1.1 error-code domain, §4.2 roles, §5 events, and the normalised dataclasses the plugin consumes (BridgeCommand, StatusReport, PairingReport, FabricInfo, EndpointSpec). No rename firewall — we author both ends — so what protects us is `protocolVersion`, per the spec preamble. `build_attach` carries the §3.1 `intent: "replace_all"` rule: never emitted by default, so a stale client cannot un-export everything. **`bridge_client.py` (new)** — hello frame, version check, attach, StatusReport. Skew fails closed (§2): no attach, `on_version_skew`, and the run loop halts instead of spinning until `resume()`. Attach carries a *fresh* read of the injected endpoint provider on every connect, which is the reconcile-on-connect of PRD §5.4. `set_state` does not await its response — it is called from Indigo's device thread and must never block it — and a push while disconnected is dropped, not raised, because attach re-delivers everything (§6.2). **Golden frames reconciled.** `tests/fixtures/bridge_protocol/frames.json` is now the single shared location §7 asks for: the E0 frames plus the E1 commands and every §5 event. Commands the node does not implement yet sit in a `pending` section the TS suite skips (and asserts is well-formed); the Python suite asserts them in full, since the plugin half ships now. `npm test` copies the directory in, so both suites break together when a frame moves. Tests: 1022 → 1161 Python (all pre-existing pass untouched), 68 → 84 Node (69 pass, 15 skipped pending). pylint 9.26 → 9.33. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JxqGhP3DcENf68AZK21U4S
Three reviews converged on the same theme: the bridge client's failure paths were the least-examined code in the PR, and several of them were wrong in ways that only show up on a bad day. Fixes, with a regression test each. Critical - set_state's send can raise even while `connected` is True (the listen loop has not noticed the socket died). It now drops instead of throwing into Indigo's device thread. A push while HALTED warns the truth — nothing is reconciling — rather than promising "attach will reconcile". - Terminal attach refusals no longer retry forever. version_mismatch and mass_removal_refused halt with a distinguishable `halted_reason`; endpoint_map_invalid keeps the socket OPEN un-attached so §1.1's get_status/get_pairing/rebuild_endpoint_map — PRD §7's way out — stay reachable, and re-attaches once the rebuild succeeds. Everything else reconnects on the normal backoff, logging the error_code, not a traceback. New `on_attach_refused` callback and an `attached` property distinct from transport-level `connected`. - A peer that opens the socket and says nothing now hits a 5s HELLO_TIMEOUT in the shared base instead of waiting forever — the same latent hang existed in MatterClient. High - A halt latched inside a run-loop iteration wins over a resume() racing it; resume() refuses (and warns) while a run loop is live, and documents that it does not restart one. - Frame containment: an undecodable or non-object frame is logged (truncated) and dropped, connection kept. Every exit path from a connection attempt now closes the socket — the broad-exception path leaked it and turned the 1s backoff into a connect/attach hammer. - The repeated-failure diagnostic fires from the handshake/broad-except path too, so handshake failure streaks reach the supervisor. - Un-awaited sends keep a bounded FIFO note (message_id → "set_state dev N"), so an unmatched error names the device instead of a bare message_id. Medium - A dict frame that is neither event nor response is logged, not silently dropped; event-dispatch failures name the event and payload in both clients; asyncio.TimeoutError is caught distinctly (from 3.11 it IS an OSError) so a timeout never logs "connection lost: " with an empty reason; upsert_endpoint without endpointNumber and a window without pairing codes raise instead of defaulting to 0/"". Fixtures and spec - The golden `attach` frame is now lawful (empty request set → empty live set); the E0 node's deviation is asserted explicitly on the TS side and noted in the fixture. `unknown_command` uses a genuinely unknown name. - New error frames for malformed_args, unknown_role, commissioning_window_failed and internal; a fabrics_changed "deleted"; a command for an unexported device; rebuild_endpoint_map now demonstrates reallocation instead of echoing attach. - Coverage is real: ROLE_STATE_KEYS/ROLE_COMMANDS in bridge_protocol drive assertions that every §4.2 role, command name and state key, and every §1.1 error code, is carried by a frame — deleting any of them fails the suite. - bridge-node: EventFrame.event and EndpointSummary.role are unions now, so the `satisfies` mirror catches misspellings; the seven orphan event frames are swept by fixtures.test.ts. BRIDGE_PROTOCOL §4.3 documents DriftEntry. python: 1287 passed (was 1178). bridge-node: 105 tests, 0 fail, 31 skipped (pending grew 15→31, all frames the node genuinely lacks handlers for). pylint 9.37/10. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JxqGhP3DcENf68AZK21U4S
simons-plugins
force-pushed
the
feat/e1-bridge-client
branch
from
August 4, 2026 20:37
5680908 to
f6c81f0
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (5)
indigo-matter.indigoPlugin/Contents/Server Plugin/ws_json_client.py (1)
393-398: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winLog the reconcile failure with explicit exception info.
_on_reconcile_doneis a done-callback, so no exception is being handled at that point.logging.Logger.exceptionderives the traceback fromsys.exc_info(), which is empty here, so a stdlib logger recordsNoneType: Noneinstead of the real traceback. Pass the exception explicitly and add a message that names the hook.♻️ Proposed change
- exc = task.exception() - if exc is not None: - self.logger.exception(exc) # on_connect failed — surface, don't swallow + exc = task.exception() + if exc is not None: + # exc_info is explicit: this runs in a done-callback, so sys.exc_info() + # is empty and logger.exception() alone would log "NoneType: None". + self.logger.error("%s: on_connect reconcile failed: %s", self.PEER, exc, + exc_info=exc)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/ws_json_client.py around lines 393 - 398, Update _on_reconcile_done so the non-cancelled task exception is logged with an explicit error message naming the reconcile hook and the captured exc value, rather than using logger.exception without exception context. Preserve the existing cancellation return and exception check.indigo-matter.indigoPlugin/Contents/Server Plugin/matter_client.py (2)
91-94: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winMove the
sdk_versionwire field name intoprotocol.py.Line 93 reads the matter-server field
"sdk_version"as a literal in this module. The coding guidelines require matter-server wire field names to stay inprotocol.py. Add a constant there and use it here.♻️ Proposed change
- version = (self.server_info or {}).get("sdk_version", "version unknown") + version = (self.server_info or {}).get(protocol.KEY_SDK_VERSION, "version unknown")Add to
protocol.pyalongside the other key constants:KEY_SDK_VERSION = "sdk_version"As per coding guidelines: "Keep matter-server wire field names confined to
protocol.py, the rename firewall."🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/matter_client.py around lines 91 - 94, Define a KEY_SDK_VERSION constant in protocol.py alongside the existing protocol key constants, then update the version lookup in the matter client’s connection logging to use that constant instead of the literal "sdk_version".Source: Coding guidelines
88-90: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winRegister a send context for the un-awaited
start_listening.
start_listeningis sent without a pending future, so its response reaches_log_unmatched. If matter-server answers with an error, the warning names only themessage_id._remember_send_contextexists for exactly this case. Record the frame's id so the log line names the request.♻️ Proposed change
- await self._send_frame(self.proto.build_request(protocol.CMD_START_LISTENING)) + frame = self.proto.build_request(protocol.CMD_START_LISTENING) + self._remember_send_context(self._message_id_of(frame), protocol.CMD_START_LISTENING) + await self._send_frame(frame)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/matter_client.py around lines 88 - 90, Update the start_listening send path in the surrounding connection method to register the frame’s message ID with _remember_send_context before awaiting _send_frame, so any unmatched error response is logged with the request context. Preserve the existing fire-and-forget behavior and avoid changing the reconciliation flow.tests/test_ws_json_client.py (1)
402-414: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winAdd a test for the unmatched-error log line.
The class docstring states that an unmatched error must name the request. The only test here checks the size bound of
_send_context. The_log_unmatchedwarning path is the behaviour the shared transport guarantees, and it is untested in this file. Add a scenario that sends a fire-and-forget request, answers it with an error frame, and asserts the warning contains both the recorded context and the error code.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_ws_json_client.py` around lines 402 - 414, Add a test method to TestUnmatchedContext covering the _log_unmatched warning path: send a fire-and-forget request, respond with an error frame, and assert the logger warning includes both the recorded request context and the error code. Reuse the existing bridge_client, FakeWebSocket, and mock_logger test fixtures and preserve the existing bounded-context test.tests/test_bridge_client.py (1)
699-724: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueDrop the unused
fakebinding.Both tests unpack
fakeand never use it. Ruff reports RUF059 on lines 702 and 722. Do not name it_fake, because that shadows the module-level helper at line 73.🧹 Proposed fix
- fake, client = self._recovering(mock_logger) + _unused, client = self._recovering(mock_logger) client._on_attach_refused = lambda code, details: refusals.append(code)- fake, client = self._recovering(mock_logger) + _unused, client = self._recovering(mock_logger) task = asyncio.create_task(client.run())🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_bridge_client.py` around lines 699 - 724, Remove the unused fake binding from both test scenarios by unpacking the _recovering result without assigning its first value, while retaining the client binding and avoiding the name _fake. Update the assignments in test_the_connection_is_kept_open_un_attached and test_the_three_recovery_commands_still_work.Source: Linters/SAST tools
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/BRIDGE_PROTOCOL.md`:
- Around line 314-316: Update the §4.3 example’s endpointCount value to match
the documented endpoints.length rule, or add entries until the list contains 12
endpoints; keep the example’s endpointCount and endpoints array consistent.
In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/bridge_client.py:
- Around line 383-393: Update the _SEND_ERRORS handler around _send_frame in the
set_state flow to discard the remembered send context for the failed frame
before logging the failure. Use the frame’s message ID with the existing
send-context removal mechanism, ensuring no context remains when the send cannot
produce a response.
In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/bridge_protocol.py:
- Around line 330-358: Update parse_status, parse_fabrics, and parse_command to
validate required fields before indexing or converting values, and raise
BridgeProtocolError(ERR_MALFORMED_ARGS, ...) for missing or malformed required
data instead of allowing KeyError or TypeError to escape. Preserve existing
normalization for valid payloads and ensure malformed status/command responses
follow the attach refusal and failed-event handling paths.
---
Nitpick comments:
In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/matter_client.py:
- Around line 91-94: Define a KEY_SDK_VERSION constant in protocol.py alongside
the existing protocol key constants, then update the version lookup in the
matter client’s connection logging to use that constant instead of the literal
"sdk_version".
- Around line 88-90: Update the start_listening send path in the surrounding
connection method to register the frame’s message ID with _remember_send_context
before awaiting _send_frame, so any unmatched error response is logged with the
request context. Preserve the existing fire-and-forget behavior and avoid
changing the reconciliation flow.
In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/ws_json_client.py:
- Around line 393-398: Update _on_reconcile_done so the non-cancelled task
exception is logged with an explicit error message naming the reconcile hook and
the captured exc value, rather than using logger.exception without exception
context. Preserve the existing cancellation return and exception check.
In `@tests/test_bridge_client.py`:
- Around line 699-724: Remove the unused fake binding from both test scenarios
by unpacking the _recovering result without assigning its first value, while
retaining the client binding and avoiding the name _fake. Update the assignments
in test_the_connection_is_kept_open_un_attached and
test_the_three_recovery_commands_still_work.
In `@tests/test_ws_json_client.py`:
- Around line 402-414: Add a test method to TestUnmatchedContext covering the
_log_unmatched warning path: send a fire-and-forget request, respond with an
error frame, and assert the logger warning includes both the recorded request
context and the error code. Reuse the existing bridge_client, FakeWebSocket, and
mock_logger test fixtures and preserve the existing bounded-context test.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: b932579b-e18f-4110-acb4-7d249dffa723
📒 Files selected for processing (23)
CLAUDE.mdbridge-node/package.jsonbridge-node/src/node.tsbridge-node/src/protocol.tsbridge-node/src/ws-server.tsbridge-node/test/fixture-shapes.tsbridge-node/test/fixtures.test.tsbridge-node/test/fixtures/e0-frames.jsonbridge-node/test/protocol.test.tsbridge-node/test/stub-bridge.tsdocs/BRIDGE_PROTOCOL.mddocs/HANDOVER.mdindigo-matter.indigoPlugin/Contents/Info.plistindigo-matter.indigoPlugin/Contents/Server Plugin/bridge_client.pyindigo-matter.indigoPlugin/Contents/Server Plugin/bridge_protocol.pyindigo-matter.indigoPlugin/Contents/Server Plugin/matter_client.pyindigo-matter.indigoPlugin/Contents/Server Plugin/ws_json_client.pytests/conftest.pytests/fakes.pytests/fixtures/bridge_protocol/frames.jsontests/test_bridge_client.pytests/test_bridge_protocol_frames.pytests/test_ws_json_client.py
💤 Files with no reviewable changes (1)
- bridge-node/test/fixtures/e0-frames.json
Comment on lines
+314
to
+316
| `endpoints[].role` is one of the §4.2 enum; `endpointCount` is | ||
| `endpoints.length` (it is sent explicitly so a client can log the size without | ||
| walking the list). |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
The new endpointCount rule contradicts the §4.3 example.
The added text states that endpointCount is endpoints.length. The example above it shows "endpointCount": 12 with one entry in endpoints. Update the example to 1, or extend the example list, so the shape and the rule agree.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/BRIDGE_PROTOCOL.md` around lines 314 - 316, Update the §4.3 example’s
endpointCount value to match the documented endpoints.length rule, or add
entries until the list contains 12 endpoints; keep the example’s endpointCount
and endpoints array consistent.
Comment on lines
+383
to
+393
| self._remember_send_context(self.proto.message_id_of(frame), | ||
| f"set_state dev {indigo_device_id}") | ||
| try: | ||
| await self._send_frame(frame) | ||
| except _SEND_ERRORS as exc: | ||
| # The listen loop has not noticed the socket is gone yet, so | ||
| # ``connected`` is still True and the check above passed. Dropping is | ||
| # still right (attach re-delivers), but it must not surface as an | ||
| # exception on Indigo's device thread. | ||
| self.logger.debug("bridge node send failed (%s); dropping set_state for %s", | ||
| exc, indigo_device_id) |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Inspect the send-context bookkeeping in the shared transport.
fd -t f 'ws_json_client.py' --exec ast-grep outline {} --items all
fd -t f 'ws_json_client.py' --exec rg -n -C4 '_send_context|_remember_send_context'Repository: simons-plugins/indigo-matter
Length of output: 6706
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Inspect the relevant state-correlation and send methods, plus tests around missing responses.
printf '\n== outlines ==\n'
for f in indigo-matter.indigoPlugin/Contents/Server\ Plugin/bridge_client.py indigo-matter.indigoPlugin/Contents/Server\ Plugin/ws_json_client.py tests/test_bridge_client.py tests/test_ws_json_client.py; do
if [ -f "$f" ]; then
echo "--- $f ---"
ast-grep outline "$f" --match '_handle_frame|_send_context|_remember_send_context|_send_frame|set_state|UnmatchedContext|missing|no' --view expanded || true
fi
done
printf '\n== targeted source snippets ==\n'
sed -n '330,410p' indigo-matter.indigoPlugin/Contents/Server\ Plugin/bridge_client.py
printf '\n--- ws_json_client relevant ---\n'
sed -n '330,425p' indigo-matter.indigoPlugin/Contents/Server\ Plugin/ws_json_client.py
printf '\n== targeted tests ==\n'
rg -n -C 8 '_send_context|missing response|unmatched|no response|set_state|_send_frame|_ask|_request' indigo-matter.indigoPlugin/Contents/Server\ Plugin testsRepository: simons-plugins/indigo-matter
Length of output: 50385
Discard the send context when send fails.
When _send_frame raises, no response for that message_id can arrive, and the unmatched-response path never removes the send-context entry. Remove it in the _SEND_ERRORS handler before logging the failure.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/bridge_client.py around
lines 383 - 393, Update the _SEND_ERRORS handler around _send_frame in the
set_state flow to discard the remembered send context for the failed frame
before logging the failure. Use the frame’s message ID with the existing
send-context removal mechanism, ensuring no context remains when the send cannot
produce a response.
Comment on lines
+330
to
+358
| def parse_fabrics(data: Any) -> list: | ||
| """Normalise a list of §4.3 ``FabricInfo`` objects.""" | ||
| return [ | ||
| FabricInfo( | ||
| fabric_index=int(item[ARG_FABRIC_INDEX]), | ||
| label=str(item.get("label", "")), | ||
| vendor_id=int(item.get("vendorId", 0)), | ||
| ) | ||
| for item in (data or []) | ||
| ] | ||
|
|
||
|
|
||
| def parse_status(result: Any) -> StatusReport: | ||
| """Normalise a ``StatusReport`` payload (§4.3).""" | ||
| data = result or {} | ||
| return StatusReport( | ||
| commissioned=bool(data.get("commissioned", False)), | ||
| fabrics=parse_fabrics(data.get("fabrics")), | ||
| endpoint_count=int(data.get("endpointCount", 0)), | ||
| endpoints=[ | ||
| EndpointSummary( | ||
| indigo_device_id=int(item[ARG_INDIGO_DEVICE_ID]), | ||
| endpoint_number=int(item["endpointNumber"]), | ||
| role=str(item.get("role", "")), | ||
| ) | ||
| for item in (data.get("endpoints") or []) | ||
| ], | ||
| drift=parse_drift(data.get("drift")), | ||
| ) |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Inspect exception handling around bridge_protocol parsers in the client.
fd -t f 'bridge_client.py' | while IFS= read -r f; do
rg -n -C 6 'parse_status|parse_command|parse_fabrics|parse_drift|except ' "$f"
doneRepository: simons-plugins/indigo-matter
Length of output: 5891
🏁 Script executed:
#!/bin/bash
set -euo pipefail
printf '%s\n' '--- locate files ---'
fd -t f 'bridge_protocol.py|bridge_client.py' .
printf '%s\n' '--- bridge_protocol relevant parser definitions ---'
f=$(fd -t f 'bridge_protocol.py' . | head -n 1)
wc -l "$f"
sed -n '320,425p' "$f"
printf '%s\n' '--- client attach/event parsing calls and handlers ---'
c=$(fd -t f 'bridge_client.py' . | head -n 1)
wc -l "$c"
sed -n '160,190p' "$c"
sed -n '230,280p' "$c"
sed -n '320,345p' "$c"
sed -n '410,445p' "$c"
sed -n '455,472p' "$c"Repository: simons-plugins/indigo-matter
Length of output: 12611
Fail closed on missing required status/command fields.
parse_status, parse_fabrics, and parse_command raise KeyError or TypeError when required fields such as ARG_INDIGO_DEVICE_ID, endpointNumber, or ARG_FABRIC_INDEX are absent. These exceptions can bypass the attach refusal path and can appear as generic failed events. Convert these missing-field cases to BridgeProtocolError(ERR_MALFORMED_ARGS, ...).
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@indigo-matter.indigoPlugin/Contents/Server` Plugin/bridge_protocol.py around
lines 330 - 358, Update parse_status, parse_fabrics, and parse_command to
validate required fields before indexing or converting values, and raise
BridgeProtocolError(ERR_MALFORMED_ARGS, ...) for missing or malformed required
data instead of allowing KeyError or TypeError to escape. Preserve existing
normalization for valid payloads and ensure malformed status/command responses
follow the attach refusal and failed-event handling paths.
This was referenced Aug 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Milestone E1 (PRD §9): the plugin-side half of the local protocol, per docs/BRIDGE_PROTOCOL.md.
ws_json_client.py(new, 359): transport core extracted frommatter_client.py— run loop, exponential backoff, message_id↔future correlation, disconnect semantics, repeated-failure diagnostic latch. Handshake stays an overridable hook. Spec-required behaviour change: unmatched error responses are now logged (were silently dropped) — BRIDGE_PROTOCOL §3.4's fire-and-forgetset_statedepends on it.matter_client.py311 → 147: matter-server specialisation only.tests/test_matter_client.py(446 lines) passes with zero edits.bridge_protocol.py(new, 506): our wire contract — envelope helpers, normalised dataclasses (BridgeCommand/StatusReport/PairingReport/FabricInfo), §1.1 error codes, the §3.1intent: replace_allrule.bridge_client.py(new, 310): hello-frame handshake with version skew failing closed (halt + resume, no spurious reconnect), attach-reconcile on every connect via an injected endpoint provider, async methods for the full command set, fire-and-forgetset_state, event callbacks (command/fabrics_changed/commissioned/window_closed/drift_detected).tests/fixtures/bridge_protocol/frames.jsonconsumed by BOTH suites; the 13 exchanges the node doesn't implement yet sit in apendingsection the TS suite skip-tracks.Tests
Python 1022 → 1161; Node 68 → 84 (69 pass + 15 pending-skips). pylint 9.26 → 9.33 (note: the previously-recorded 9.90 was a stale pylint cache artifact; 9.26 is the real main baseline).
Spec calls made (documented in code)
factory_resetresult assumed{};driftentry shape taken from the node'sDriftEntry; empty attach never auto-carriesreplace_all(the §3.1 guard is the point); attach pumped inline during handshake (8s, inside the node's 10s unattached close).🤖 Generated with Claude Code
https://claude.ai/code/session_01JxqGhP3DcENf68AZK21U4S
Summary by CodeRabbit
New Features
Documentation
Bug Fixes
Tests
Chores