Spaces:
Running
Running
Commit History
-
Mission M-E — Observability & Tests
Ashraf Al-Kassem +1: Claude Sonnet 4.6- ADK callback handler (callbacks.py): logs all lifecycle events to RuntimeEventLog - agent.{name}.started/completed, agent.llm_request/response, agent.error - tool.{name}.started/completed, tool error events - _redact_sensitive() strips access_token, api_key, password, token fields - runner.py: attaches LeadPilotCallbackHandler to agent after build_orchestrator() - tracks turn duration + records agent.turns_processed / avg_turn_duration_ms metrics - metrics_service.py: adds gauge() method (overwrites, not additive) - admin.py: GET /executions/{instance_id}/trace — returns chronological event timeline - 8 new tests: test_callbacks.py (5 unit), test_e2e_agent_flow.py (3 E2E) - Frontend: execution trace timeline page at /admin/executions/[executionId] - Frontend: monitoring/page.tsx — Agent Metrics tab + View Trace links in Executions tab - docs/missions/event-types.md: full event taxonomy reference Total tests: 360 (all passing) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthr…87e5ad2
-
feat: Mission 21 — Settings Center + Data-Driven Dropdowns
Ashraf Al-Kassem +1: Claude Opus 4.6- Add PATCH /auth/me for user profile updates with audit logging - Add 8 new catalog endpoints (timezones, languages, ai-models, dedupe-strategies, event-sources, event-outcomes, audit-actions, contact-fields) - Create LookupsProvider context + useLookups() hook for bulk catalog preload - Restructure Settings page with Profile section (name edit, email verification badge) - Replace 10 hardcoded dropdowns across 5 frontend pages with catalog data - Sidebar shows real user name + computed initials from /auth/me - 141 tests passing (+23 new), zero TS build errors Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>c626151 -
feat: Mission 19+20 — Catalog Standardization + Prompt Studio Knowledge Base v2 + Dynamic Lead Qualification
Ashraf Al-Kassem +1: Claude Opus 4.6Mission 19: Catalog registry with standardized provider/trigger/action metadata, validation, and frontend catalog client. Mission 20: Single source-of-truth prompt compiler replacing two divergent AI context paths. Knowledge files now injected into AI context. Automation AI steps use goal/tasks/extra_instructions. Dynamic lead qualification with workspace-level questions/statuses and admin defaults. - 118 tests passing (was 57), frontend build clean Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>9d062e9