MCP consolidation:
- Route all MCP-capable connectors (Slack, Jira, Linear, ClickUp, Airtable,
Notion, Confluence, interim Gmail/Calendar, custom MCP) through a single
`mcp_discovery` subagent. Drive/OneDrive/Dropbox stay native to enrich the KB.
- Deprecate Discord/Teams/Luma: no viable official MCP server.
Google-only web search:
- Remove the main-agent `web_search` tool and the SearXNG platform service;
all public web search now flows through the `google_search` subagent via task().
- Deprecate the Tavily/SearXNG/Linkup/Baidu search connectors (HTTP 410 on
create, "Deprecated" badge); guide heavy users to the custom MCP connector.
- Remove web search from anonymous chat (pure Q&A).
- Tear SearXNG out of docker compose + install scripts; drop tavily-python
and linkup-sdk deps and their config/env vars.
Fix:
- metrics._package_version() now swallows any metadata lookup failure. A
malformed editable-install distribution with no `Version` field raised
KeyError deep in importlib.metadata, and since it runs on every
record_subagent_invoke_duration call it was crashing every task()
delegation. Verified end-to-end against live GPT-5.4.
Co-authored-by: Cursor <cursoragent@cursor.com>
Closes the create loop in chat: the agent describes user intent → the
drafter sub-LLM produces an AutomationCreate JSON → this card surfaces
a structured preview → approve persists; reject cancels. Edits flow
through chat refinement (re-call with a refined intent), not in-card,
so the card stays simple and the multi-turn checkpointer carries the
context.
Tool UI (components/tool-ui/automation/):
- create-automation.tsx — entry dispatcher + ApprovalCard chrome
(pending/processing/complete/rejected via useHitlPhase) + SavedCard
(links to the detail page) + InvalidCard (lists drafter validation
issues) + ErrorCard (verbatim message). Rejection result is hidden
because the approval card itself shows the rejected phase inline.
- automation-draft-preview.tsx — structured preview body: name +
description + goal, triggers (humanised cron + tz + static-input
keys), plan steps (step_id → action), and a collapsible raw JSON
for power users.
Wiring:
- components/tool-ui/index.ts — re-export.
- features/chat-messages/timeline/tool-registry/registry.ts —
register create_automation → CreateAutomationToolUI (dynamic import,
same pattern as other connector tools).
- contracts/enums/toolIcons.tsx — Workflow icon + "Create automation"
display name so fallback chrome (and timeline headers) are honest.
Shared util:
- lib/automations/describe-cron.ts — lifted from the route slice's
lib/ folder since both the dashboard slice and the new approval card
now render schedule descriptions. Slice imports updated; the now-
empty slice lib/ folder is gone.
Backend prompt fragments:
- main_agent/system_prompt/.../create_automation/description.md and
the tool's docstring no longer promise in-card edits. They make the
refinement path explicit: if the user wants changes after seeing the
draft, they reply in chat and the agent calls the tool again with a
refined intent.
v1 deliberately excludes:
- In-card edit form / right-side edit panel — defer until we see real
demand. The chat refinement loop covers the common case.
- approve_always / persistent allow rules — automations are a single
artifact, not a repeated mutation, so the "trust this kind of call"
affordance doesn't apply.
The 'Always Allow' button is now driven entirely by the server-supplied
allowed_decisions palette. The card no longer peeks at
context.mcp_connector_id to decide whether to render the button, and no
longer fires a separate trust-tool HTTP call on click - one
{type: 'approve_always'} dispatch is enough; the agent middleware
handles the in-memory promotion and (for MCP tools) the database save
via its trusted_tool_saver callback.
Drops the dead trustMCPTool / untrustMCPTool service helpers - they had
no remaining callers after this rework. The backing HTTP routes are
kept on the server as a programmatic surface.
Renames the SurfSense HITL extension decision-type from "always" to
"approve_always" so it sits in the same verb-first family as "approve",
"reject", and "edit". The Python constant is now SURFSENSE_DECISION_APPROVE_ALWAYS;
the wire value, the permission-domain decision_type, and the FE union members
all match (no wire/internal mismatch).
Both the multi_agent_chat permission middleware and the legacy new_chat one
accept the new wire value; the FE types.ts union is updated accordingly.
The "context.always" payload key is intentionally left untouched - it's the
patterns-to-promote field, semantically distinct from the decision type.
