… outbound)

Port GoogleChatAdapter Part B to match upstream adapter-gchat/src/index.ts:

- handle_webhook accepts both direct HTTP MESSAGE/ADDED_TO_SPACE/REMOVED_FROM_SPACE
  events and Pub/Sub push envelopes, funneling both into _dispatch_event.
- New pubsub.py helper detects envelope shape and unwraps CloudEvents metadata
  so the Workspace Events Pub/Sub flow routes through the same handler chain.
- Outbound surface: post_message (plain text + Card v2), edit_message with
  update_mask=text,cards_v2, delete_message, add_reaction / remove_reaction
  (emoji → Unicode via DEFAULT_EMOJI_MAP), fetch_messages with pageToken/pageSize
  pagination, fetch_channel_info, fetch_channel_messages, list_threads, open_dm,
  start_typing, stream (placeholder → periodic edit → final flush).
- open_modal raises chat.NotImplementedError with feature="modals" (Google Chat
  has no Slack-style modals; upstream uses Card v2 inline instead).
- REST dispatch via pluggable _rest_client (tests inject SimpleNamespace +
  AsyncMock; production falls back to an httpx-backed attribute-walk shim).
- ADDED_TO_SPACE triggers create_space_subscription when pubsub_topic is set;
  REMOVED_FROM_SPACE tears the cached subscription down.
- Adapter Protocol conformance: adds initialize, disconnect, post_channel_message,
  subscribe/unsubscribe, get_channel_visibility; declares lock_scope="thread"
  and persist_message_history=False class attrs.

Tests: new tests/test_dispatch.py with 16 cycles covering 3.2–3.8 plus the
Phase 0 conformance test now flips GREEN (152 gchat tests pass).

E2E skeleton at examples/e2e/gchat/echo.py — mirrors slack/echo.py shape,
ast.parses cleanly, exits with clear message when required env vars missing.

…keleton

…eton

- LinearAdapter now conforms to chat.types.Adapter Protocol
  (added lock_scope, persist_message_history, disconnect, subscribe,
  unsubscribe, post_channel_message, fetch_channel_info,
  fetch_channel_messages, list_threads, open_dm, open_modal, is_dm,
  get_channel_visibility).
- Unsupported surfaces raise chat.errors.NotImplementedError with
  feature= kwarg (matches Teams/WhatsApp pattern).
- remove_reaction upgraded from a silent warn no-op to a pinned
  chat.NotImplementedError(feature='removeReaction') stub; add_reaction
  remains fully implemented (Linear reactionCreate).
- test_dispatch.py pins Comment webhook round-trip via Chat.handle_webhook
  (HMAC-SHA256 verified end-to-end, no monkeypatch) and the Author
  dataclass shape (guards against author-dict regression).
- examples/e2e/linear/echo.py skeleton — LINEAR_API_KEY +
  LINEAR_WEBHOOK_SECRET env vars, fails clearly when missing.
- docs/parity.md: Linear 'react' column updated to 'partial' with a
  dedicated stub-listing subsection for the 7 Linear stubs.

…keleton

…f-test

Add a parametrised cross-adapter dispatch test that routes canned webhook
bodies through ``Chat.handle_webhook(<name>, body, headers)`` for all 8
shipped adapters (slack / gchat / discord / github / whatsapp / teams /
linear / telegram) and asserts the expected handler fires. Body fixtures
live under ``tests/__fixtures__/`` with provenance comments; signature-
bearing adapters (Slack / GitHub / WhatsApp / Linear) compute HMAC
signatures dynamically against known secrets, while Ed25519 (Discord)
and JWT (Teams / GChat) verifiers are monkeypatched at factory time.

Also add a parity-doc self-test that walks ``packages/`` and asserts
every ``## Dispatch surface`` row in ``docs/parity.md`` corresponds to a
real ``chat-adapter-<name>`` module — catches stale parity entries.

Closes DES-196.

Three bugs surfaced running examples/e2e/slack/echo.py against a real
workspace:

- examples/e2e/_common.py: FastAPI could not resolve the `Request` type
  annotation on the webhook handler — imports lived inside the function
  scope but `from __future__ import annotations` stringifies hints, so
  FastAPI fell back to treating `request` as a query parameter and
  returned 422 before the body was ever parsed. Move the fastapi /
  uvicorn / Response imports to module level behind a defensive ImportError
  guard.

- packages/chat-adapter-slack/src/chat_adapter_slack/adapter.py:
  SlackAdapter.initialize() now calls auth.test() to populate
  bot_user_id / bot_id when they weren't passed via config. Mirrors
  upstream's initialize — without this, Chat._detect_mention can never
  match `<@U_BOT>` on plain `message.channels` events and the
  `on_new_mention` handler never fires for non-app_mention traffic.

- examples/e2e/slack/echo.py: handlers take (thread, message, context),
  matching Chat._dispatch_to_handlers; use `author.user_id` instead of
  the nonexistent `author.id`.

Plus ruff RUF100 / format cleanup across the sibling echo skeletons.

214 Slack tests still green. Manual E2E verified live — Slack
url_verification 200 + @-mention dispatch + in-thread subscribe/echo.

Adds the `--mode webhook|socket` flag the original plan flagged as a
nice-to-have. `examples/e2e/_common.py` gains a `run_socket_client`
helper that calls `Chat.initialize` (which in turn drives
`SlackAdapter.connect`, opening the websocket), then blocks on
SIGINT/SIGTERM and cleans up via `adapter.disconnect`.

`--mode socket` requires SLACK_APP_TOKEN; defaults to webhook. Picks up
$SLACK_MODE as a default if no flag is passed.

Live-verified end-to-end: websocket connects, `auth.test` resolves
identity, `apps.connections.open` succeeds, Slack routes @-mention +
thread-message events through the socket, handlers fire, bot replies
in-thread — no ngrok required.