feat: ContentGuardianAgent, onboarding UX, Team Activity action wiring, docs, agent help modal

ContentGuardianAgent consolidation:
- Merge 3 duplicate classes into single source in specialized/content_guardian.py
- Watchdog audit_committee() with heuristic scoring, coverage gaps, overlaps, alerts
- Remove misleading rejection_rate() helper; use acceptance_rate directly
- Integrate audit + alerts + trend signals into today_workflow_service.py

Team Activity page:
- QualityAuditPanel: health ring, per-agent critiques, coverage gaps, overlaps
- TrendSignalsPanel: opportunity cards with urgency/impact/coverage bars
- AlertBanner: persistent dismiss via POST /alerts/{id}/mark-read
- AgentHelpModal: dialog showing all 8 agents with descriptions, tools, schedule
- QualityAuditPanel action buttons: Fill gap -> /content-planning, Resolve overlap, View CTA on alerts/issues
- TrendSignalsPanel action buttons: Create content from this trend -> /blog-writer with trend context state

Onboarding system:
- Step 4 validation: no auto-pass via basic_ready; requires persona data or explicit progression
- Step 5 validation: logs warning on auto-pass without integration data
- OnboardingCompletionService: single DB session, transactional task creation, upsert pattern
- Business-without-website: nullable website_url on SIFIndexingTask and MarketTrendsTask
- DeepCompetitorAnalysisExecutor: 5-min timeout, 10-competitor cap, asyncio.wait_for
- Persona generation: async with 30s timeout, falls back to scheduler
- OnboardingProgressService.reset_onboarding(): resets session + pauses all DB tasks
- OnboardingControlService.reset_onboarding(): also cancels APScheduler jobs
- FinalStep TaskSchedulingPanel: shows scheduled/failed tasks after completion, 8s auto-redirect
- onboarding_completed agent activity event logged to feed

Documentation:
- docs-site/features/onboarding/: overview, steps, scheduler-tasks, technical-reference (4 pages)
- docs-site/mkdocs.yml: added Onboarding System nav section
- docs-site/features/sif-agents/: overview, agent-directory, committee-system, content-guardian (4 pages)
- docs-site/features/team-activity/: overview, quality-audit, trend-signals, alert-system (4 pages)
- docs-site/features/todays-workflow/: updated overview, technical-architecture, workflow-guide, api-reference

backend/services/llm_providers/main_text_generation.py

@@ -375,9 +375,13 @@ def llm_text_gen(
                         system_prompt=system_instructions
                     )
             elif gpt_provider == "wavespeed":
-                llm_start = time.time()
+                t0 = time.time()
+                logger.warning(f"[llm_text_gen][{flow_tag}] wavespeed: Starting provider init for user {user_id}")
                 if json_struct:
+                    logger.warning(f"[llm_text_gen][{flow_tag}] wavespeed: Importing wavespeed_provider module (lazy import) for user {user_id}")
                     from services.llm_providers.wavespeed_provider import wavespeed_structured_json_response
+                    logger.warning(f"[llm_text_gen][{flow_tag}] wavespeed: Import done, making API call for user {user_id}, import_took={(time.time()-t0)*1000:.0f}ms")
+                    t1 = time.time()
                     response_text = wavespeed_structured_json_response(
                         prompt=prompt,
                         schema=json_struct,
@@ -387,7 +391,10 @@ def llm_text_gen(
                         system_prompt=system_instructions
                     )
                 else:
+                    logger.warning(f"[llm_text_gen][{flow_tag}] wavespeed: Importing wavespeed_provider module (lazy import) for user {user_id}")
                     from services.llm_providers.wavespeed_provider import wavespeed_text_response
+                    logger.warning(f"[llm_text_gen][{flow_tag}] wavespeed: Import done, making API call for user {user_id}, import_took={(time.time()-t0)*1000:.0f}ms")
+                    t1 = time.time()
                     response_text = wavespeed_text_response(
                         prompt=prompt,
                         model=model or "openai/gpt-oss-120b",
@@ -396,8 +403,9 @@ def llm_text_gen(
                         top_p=top_p,
                         system_prompt=system_instructions
                     )
-                llm_ms = (time.time() - llm_start) * 1000
-                logger.warning(f"[llm_text_gen][{flow_tag}] LLM API call took {llm_ms:.0f}ms for user {user_id} (wavespeed)")
+                api_took_ms = (time.time() - t1) * 1000
+                total_ms = (time.time() - t0) * 1000
+                logger.warning(f"[llm_text_gen][{flow_tag}] wavespeed: user={user_id} import_took={(t1-t0)*1000:.0f}ms api_took={api_took_ms:.0f}ms total={total_ms:.0f}ms")
             else:
                 logger.error(f"[llm_text_gen] Unknown provider: {gpt_provider}")
                 raise RuntimeError(f"Unknown LLM provider: {gpt_provider}. Supported providers: google, huggingface, wavespeed")

backend/services/llm_providers/wavespeed_provider.py

@@ -38,6 +38,7 @@ Last Updated: March 2026
 
 import os
 import sys
+import time as _time
 from pathlib import Path
 import json
 import re
@@ -46,15 +47,16 @@ from typing import Optional, Dict, Any, List
 from dotenv import load_dotenv
 
 # Fix the environment loading path - load from backend directory
+_mod_start = _time.time()
 current_dir = Path(__file__).parent.parent  # services directory
 backend_dir = current_dir.parent  # backend directory
 env_path = backend_dir / '.env'
 
 if env_path.exists():
     load_dotenv(env_path)
-    print(f"Loaded .env from: {env_path}")
+    _dotenv_ms = (_time.time() - _mod_start) * 1000
+    print(f"Loaded .env from: {env_path} (took {_dotenv_ms:.0f}ms)")
 else:
-    # Fallback to current directory
     load_dotenv()
     print(f"No .env found at {env_path}, using current directory")
 
@@ -64,6 +66,7 @@ from utils.logger_utils import get_service_logger
 # Use service-specific logger to avoid conflicts
 logger = get_service_logger("wavespeed_provider")
 
+_import_start = _time.time()
 from tenacity import (
     retry,
     retry_if_exception,
@@ -80,6 +83,8 @@ except ImportError:
     NotFoundError = Exception
     logger.warn("OpenAI library not available. Install with: pip install openai")
 
+logger.warning(f"[wavespeed_provider] Module import completed in {(_time.time()-_import_start)*1000:.0f}ms (openai_available={OPENAI_AVAILABLE})")
+
 # Default WaveSpeed models for fallback
 WAVESPEED_FALLBACK_MODELS = [
     "openai/gpt-oss-120b",
@@ -276,12 +281,13 @@ def wavespeed_text_response(
         if not api_key:
             raise Exception("WAVESPEED_API_KEY not found in environment variables")
             
+        _t0 = _time.time()
         # Initialize WaveSpeed client
         client = OpenAI(
             base_url="https://llm.wavespeed.ai/v1",
             api_key=api_key,
         )
-        logger.info("✅ WaveSpeed client initialized for text response")
+        logger.warning(f"[wavespeed_text_response] OpenAI client init took {(_time.time()-_t0)*1000:.0f}ms")
 
         # Prepare input for the API
         messages = []
@@ -311,6 +317,7 @@ def wavespeed_text_response(
         
         logger.info("🚀 Making WaveSpeed API call (chat completion)...")
         
+        _api_t0 = _time.time()
         # Call exactly the requested model; no retries, no fallbacks, no variants
         response = client.chat.completions.create(
             model=model,
@@ -319,6 +326,7 @@ def wavespeed_text_response(
             top_p=top_p,
             max_tokens=max_tokens
         )
+        logger.warning(f"[wavespeed_text_response] API call took {(_time.time()-_api_t0)*1000:.0f}ms")
         
         # Extract text from response
         generated_text = response.choices[0].message.content
@@ -422,13 +430,15 @@ def wavespeed_structured_json_response(
         
         if not api_key:
             raise Exception("WAVESPEED_API_KEY not found in environment variables")
-            
+        
+        _fn_start = _time.time()
         # Initialize OpenAI client with WaveSpeed base URL
         client = OpenAI(
             base_url="https://llm.wavespeed.ai/v1",
             api_key=api_key,
         )
-        logger.info("✅ WaveSpeed client initialized for structured JSON response")
+        _client_init_ms = (_time.time() - _fn_start) * 1000
+        logger.warning(f"[wavespeed_structured_json_response] OpenAI client init took {_client_init_ms:.0f}ms")
 
         # Prepare input for the API
         messages = []
@@ -463,11 +473,13 @@ def wavespeed_structured_json_response(
         json_schema_str = json.dumps(schema, indent=2)
         messages[-1]["content"] += f"\n\nJSON Schema:\n{json_schema_str}"
         
+        _api_start = _time.time()
         try:
             response = None
             last_error = None
             for candidate_model in _fallback_model_sequence(model, fallback_models):
                 try:
+                    logger.info(f"[wavespeed_structured_json_response] Calling model={candidate_model}...")
                     response = client.chat.completions.create(
                         model=candidate_model,
                         messages=messages,
@@ -475,8 +487,10 @@ def wavespeed_structured_json_response(
                         max_tokens=max_tokens,
                         response_format={"type": "json_object"} # Try to enforce JSON mode if supported
                     )
+                    _api_ms = (_time.time() - _api_start) * 1000
                     if candidate_model != model:
                         logger.warning("WaveSpeed structured generation switched to fallback model: {}", candidate_model)
+                    logger.warning(f"[wavespeed_structured_json_response] First API call completed in {_api_ms:.0f}ms (model={candidate_model})")
                     break
                 except NotFoundError as nf_err:
                     last_error = nf_err