feat: frontend redesign, dashboard API, auto-create applications

- Complete frontend visual overhaul with indigo/purple gradients,
  animations (fadeIn, slideUp, shimmer), glass morphism, hover effects
- Dashboard now fetches live stats and recent analyses from new
  /dashboard/stats and /dashboard/recent-analyses API endpoints
- Analysis service auto-creates Application record (status: wishlist)
  after each successful agent run, linking analysis + cover letter
- Updated CLAUDE.md with current project state, completed phases,
  and key changes log

Author: last-26

CLAUDE.md

@@ -6,7 +6,9 @@ NextHire is an **agentic AI system** that helps job seekers analyze job postings
 
 This is NOT a simple "prompt-in, text-out" wrapper. It is a **multi-step agent** that plans, executes tools, evaluates its own output, and iterates. The agent uses **multi-model routing** (cheap model for simple tasks, powerful model for reasoning) and streams its execution steps to the frontend in real-time.
 
-**Live Architecture:** Next.js frontend → FastAPI backend → LangGraph Agent → AWS Bedrock (Claude Haiku 4.5 + Claude Sonnet)
+**Live Architecture:** Next.js frontend → FastAPI backend → LangGraph Agent → AWS Bedrock (Claude Haiku 4.5 + Claude Sonnet 4)
+
+**Current Status (March 2026):** Core pipeline fully operational. Analysis flow works end-to-end: CV upload + job description → 7-node agent pipeline → ATS score, gap analysis, cover letter → auto-creates kanban application. Dashboard shows live stats. CI/CD passes.
 
 ---
 
@@ -200,6 +202,7 @@ NextHire/
 │   │   │       ├── cover_letter.py    # POST /api/v1/cover-letter
 │   │   │       ├── cv_review.py       # POST /api/v1/cv-review
 │   │   │       ├── agent.py           # GET /api/v1/agent/stream/{run_id} (SSE)
+│   │   │       ├── dashboard.py       # GET /api/v1/dashboard/stats, /recent-analyses
 │   │   │       └── health.py          # GET /api/v1/health
 │   │   │
 │   │   ├── models/                    # SQLAlchemy ORM models
@@ -272,7 +275,7 @@ NextHire/
 ├── frontend/
 │   ├── Dockerfile
 │   ├── package.json
-│   ├── next.config.ts
+│   ├── next.config.mjs
 │   ├── tailwind.config.ts
 │   ├── tsconfig.json
 │   ├── postcss.config.mjs
@@ -568,36 +571,36 @@ docker compose exec backend pytest -v  # Tests in container
 
 ## Implementation Order (Phases)
 
-### Phase 1: Foundation (Day 1-2)
-- [ ] Project scaffolding (monorepo structure)
-- [ ] Docker Compose (FastAPI + Next.js + PostgreSQL)
-- [ ] Database models + Alembic migrations
-- [ ] Basic CRUD API for applications (kanban data)
-- [ ] LLM abstraction layer + Bedrock provider
-- [ ] Health check endpoints
-- [ ] Frontend: Layout, sidebar, routing
-
-### Phase 2: Agent Core (Day 3-5)
-- [ ] LangGraph agent graph definition
-- [ ] Agent state schema
-- [ ] Tools: cv_parser, job_scraper, keyword_extractor, semantic_scorer
-- [ ] Nodes: parse_job, parse_cv, analyze_match
-- [ ] SSE streaming endpoint
-- [ ] Frontend: Agent stream viewer component
-
-### Phase 3: Full Agent Pipeline (Day 6-8)
-- [ ] Nodes: identify_gaps, generate_cover_letter, reflect, compile_report
-- [ ] Reflection loop with quality gate
-- [ ] Multi-model routing (Haiku for parse, Sonnet for analysis)
-- [ ] Analysis results stored in DB
-- [ ] Frontend: Analysis report page (score card, skill match, gaps)
-
-### Phase 4: Application Tracking + Polish (Day 9-10)
-- [ ] Kanban board frontend (drag-and-drop)
-- [ ] Cover letter editor
-- [ ] Link analysis → application → cover letter
-- [ ] Dashboard with stats
-- [ ] CI/CD pipeline
+### Phase 1: Foundation ✅
+- [x] Project scaffolding (monorepo structure)
+- [x] Docker Compose (FastAPI + Next.js + PostgreSQL)
+- [x] Database models + Alembic migrations
+- [x] Basic CRUD API for applications (kanban data)
+- [x] LLM abstraction layer + Bedrock provider
+- [x] Health check endpoints
+- [x] Frontend: Layout, sidebar, routing
+
+### Phase 2: Agent Core ✅
+- [x] LangGraph agent graph definition
+- [x] Agent state schema
+- [x] Tools: cv_parser, job_scraper, keyword_extractor, semantic_scorer
+- [x] Nodes: parse_job, parse_cv, analyze_match
+- [x] SSE streaming endpoint
+- [x] Frontend: Agent stream viewer component
+
+### Phase 3: Full Agent Pipeline ✅
+- [x] Nodes: identify_gaps, generate_cover_letter, reflect, compile_report
+- [x] Reflection loop with quality gate
+- [x] Multi-model routing (Haiku for parse, Sonnet for analysis)
+- [x] Analysis results stored in DB
+- [x] Frontend: Analysis report page (score card, skill match, gaps)
+
+### Phase 4: Application Tracking + Polish ✅
+- [x] Kanban board frontend (drag-and-drop)
+- [x] Cover letter editor
+- [x] Link analysis → application → cover letter (auto-creates Application on analysis completion)
+- [x] Dashboard with live stats (dynamic API-driven)
+- [x] CI/CD pipeline (GitHub Actions: lint, test, build, docker)
 - [ ] README + demo screenshots
 
 ### Phase 5: Future Enhancements (Backlog)
@@ -607,6 +610,7 @@ docker compose exec backend pytest -v  # Tests in container
 - [ ] CV version management
 - [ ] Export reports as PDF
 - [ ] Email notifications for application deadlines
+- [ ] Real-time SSE streaming to frontend during agent execution (currently polls result)
 
 ---
 
@@ -643,3 +647,35 @@ docker compose exec backend pytest -v  # Tests in container
 6. **Docker Compose** is the source of truth for local development. Both frontend and backend should work with `docker compose up`.
 
 7. **PostgreSQL JSONB** — use for analysis results, parsed job/CV data. Define Pydantic schemas for validation before storing.
+
+8. **AWS Bedrock inference profiles** — New Claude models on Bedrock require inference profile IDs (prefixed with region, e.g. `eu.anthropic.claude-haiku-4-5-20251001-v1:0`). On-demand model IDs will return `ValidationException`.
+
+9. **Analysis → Application auto-creation** — `analysis_service.py` automatically creates an `Application` record (status `wishlist`) after each successful analysis, linking it to the analysis and cover letter. No manual application creation needed.
+
+---
+
+## Key Changes Log
+
+### March 2026 — Full Pipeline Wiring + Frontend Redesign
+
+**Backend:**
+- Wired `analysis_service.py` to actually invoke `agent_graph.ainvoke()` (was previously a TODO stub)
+- Fixed `bedrock.py` system prompt handling — `_get_system_prompt()` existed but was never called in `invoke()`/`stream()`
+- Added `dashboard.py` route with `/dashboard/stats` and `/dashboard/recent-analyses` endpoints
+- Auto-creation of `Application` records after each analysis (links analysis + cover letter to kanban)
+- Generated and applied initial Alembic migration for all 4 tables
+- Fixed 18+ ruff lint issues across the codebase
+- Corrected model IDs to use Bedrock inference profile format
+
+**Frontend:**
+- Replaced `next.config.ts` with `next.config.mjs` (TS config not supported in Node 20 Alpine + Next.js 14.2)
+- Complete visual redesign: indigo/purple gradient palette, glass morphism, hover-lift cards
+- Added CSS animations: fadeIn, slideUp, slideIn, pulse-gentle, shimmer loading skeletons
+- Dashboard now fetches real stats via API (analyses count, applications count, avg score, recent analyses)
+- ScoreCard: SVG circular gauge with animated gradient stroke
+- Analysis report: vertical card stack with gradient accent stripes, strengths/weaknesses grid
+- Agent stream: vertical timeline with gradient connecting line, rich step descriptions
+- Sidebar: gradient logo, active nav indicator, "Powered by AI" footer
+
+**CI/CD:**
+- Fixed all GitHub Actions jobs (backend-lint, frontend-lint, frontend-build)

backend/app/api/routes/dashboard.py

@@ -0,0 +1,53 @@
+from fastapi import APIRouter, Depends
+from sqlalchemy import func, select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.api.deps import get_session
+from app.models.agent_run import AgentRun
+from app.models.application import Application
+from app.models.cover_letter import CoverLetter
+from app.models.job_analysis import JobAnalysis
+from app.schemas.job_analysis import JobAnalysisResponse
+
+router = APIRouter()
+
+
+@router.get("/dashboard/stats")
+async def get_dashboard_stats(session: AsyncSession = Depends(get_session)):
+    """Get summary statistics for the dashboard."""
+    analyses_count = await session.scalar(select(func.count(JobAnalysis.id)))
+    applications_count = await session.scalar(select(func.count(Application.id)))
+    cover_letters_count = await session.scalar(select(func.count(CoverLetter.id)))
+    cvs_count = await session.scalar(
+        select(func.count(func.distinct(JobAnalysis.cv_filename))).where(
+            JobAnalysis.cv_filename.is_not(None)
+        )
+    )
+    avg_score = await session.scalar(
+        select(func.avg(JobAnalysis.overall_score)).where(
+            JobAnalysis.overall_score.is_not(None)
+        )
+    )
+
+    return {
+        "analyses_count": analyses_count or 0,
+        "applications_count": applications_count or 0,
+        "cover_letters_count": cover_letters_count or 0,
+        "cvs_parsed_count": cvs_count or 0,
+        "avg_score": round(avg_score, 1) if avg_score else 0,
+    }
+
+
+@router.get("/dashboard/recent-analyses", response_model=list[JobAnalysisResponse])
+async def get_recent_analyses(
+    limit: int = 10,
+    session: AsyncSession = Depends(get_session),
+):
+    """Get recent analyses for the dashboard."""
+    result = await session.execute(
+        select(JobAnalysis)
+        .where(JobAnalysis.overall_score.is_not(None))
+        .order_by(JobAnalysis.created_at.desc())
+        .limit(limit)
+    )
+    return result.scalars().all()

backend/app/main.py

@@ -3,7 +3,7 @@
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 
-from app.api.routes import agent, analyze, applications, cover_letter, cv_review, health
+from app.api.routes import agent, analyze, applications, cover_letter, cv_review, dashboard, health
 from app.config import settings
 
 
@@ -38,3 +38,4 @@ async def lifespan(app: FastAPI):
 app.include_router(cover_letter.router, prefix="/api/v1", tags=["cover-letter"])
 app.include_router(cv_review.router, prefix="/api/v1", tags=["cv-review"])
 app.include_router(agent.router, prefix="/api/v1", tags=["agent"])
+app.include_router(dashboard.router, prefix="/api/v1", tags=["dashboard"])

backend/app/services/analysis_service.py

@@ -6,6 +6,7 @@
 from app.agent.graph import agent_graph
 from app.agent.tools.cv_parser import parse_cv as parse_cv_file
 from app.models.agent_run import AgentRun
+from app.models.application import Application
 from app.models.cover_letter import CoverLetter
 from app.models.job_analysis import JobAnalysis
 
@@ -69,13 +70,28 @@ async def run_analysis(
 
             # Create cover letter record if generated
             cover_letter_text = result.get("cover_letter")
+            cover_letter_record = None
             if cover_letter_text:
-                cover_letter = CoverLetter(
+                cover_letter_record = CoverLetter(
                     analysis_id=analysis.id,
                     content=cover_letter_text,
                     tone=result.get("cover_letter_tone", "professional"),
                 )
-                self.session.add(cover_letter)
+                self.session.add(cover_letter_record)
+                await self.session.flush()
+
+            # Auto-create application for kanban tracking
+            application = Application(
+                company_name=analysis.company_name or "Unknown Company",
+                position_title=analysis.job_title or "Untitled Position",
+                job_url=job_url,
+                status="wishlist",
+                priority="medium",
+                match_score=analysis.overall_score,
+                analysis_id=analysis.id,
+                cover_letter_id=cover_letter_record.id if cover_letter_record else None,
+            )
+            self.session.add(application)
 
             # Update agent run as completed
             agent_run.status = "completed"