Implement 10 platform features: evaluation UX, admin tools, AI summaries, applicant portal

Batch 1 - Quick Wins:
- F1: Evaluation progress indicator with touch tracking in sticky status bar
- F2: Export filtering results as CSV with dynamic AI column flattening
- F3: Observer access to analytics dashboards (8 procedures changed to observerProcedure)

Batch 2 - Jury Experience:
- F4: Countdown timer component with urgency colors + email reminder service with cron endpoint
- F5: Conflict of interest declaration system (dialog, admin management, review workflow)

Batch 3 - Admin & AI Enhancements:
- F6: Bulk status update UI with selection checkboxes, floating toolbar, status history recording
- F7: AI-powered evaluation summary with anonymized data, OpenAI integration, scoring patterns
- F8: Smart assignment improvements (geo diversity penalty, round familiarity bonus, COI blocking)

Batch 4 - Form Flexibility & Applicant Portal:
- F9: Evaluation form flexibility (text, boolean, section_header types, conditional visibility)
- F10: Applicant portal (status timeline, per-round documents, mentor messaging)

Schema: 5 new models (ReminderLog, ConflictOfInterest, EvaluationSummary, ProjectStatusHistory, MentorMessage), ProjectFile extended with roundId + isLate.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/architecture/api.md

@@ -119,6 +119,9 @@ export const hasRole = (...roles: UserRole[]) =>
 export const protectedProcedure = t.procedure.use(isAuthenticated)
 export const adminProcedure = t.procedure.use(hasRole('SUPER_ADMIN', 'PROGRAM_ADMIN'))
 export const juryProcedure = t.procedure.use(hasRole('JURY_MEMBER'))
+export const observerProcedure = t.procedure.use(hasRole('SUPER_ADMIN', 'PROGRAM_ADMIN', 'OBSERVER'))
+export const mentorProcedure = t.procedure.use(hasRole('MENTOR'))
+export const applicantProcedure = t.procedure.use(hasRole('APPLICANT'))
 ```
 
 ## Router Structure
@@ -136,7 +139,9 @@ src/server/routers/
 ├── export.ts         # Export operations
 ├── audit.ts          # Audit log access
 ├── settings.ts       # Platform settings (admin)
-└── gracePeriod.ts    # Grace period management
+├── gracePeriod.ts    # Grace period management
+├── analytics.ts      # Reports & analytics (admin + observer)
+└── mentor.ts         # Mentor messaging endpoints
 ```
 
 ### Root Router
@@ -1024,6 +1029,164 @@ export const gracePeriodRouter = router({
 })
 ```
 
+### Conflict of Interest Endpoints (Evaluation Router)
+
+```typescript
+// Added to src/server/routers/evaluation.ts
+
+// Declare COI for an assignment (jury member)
+declareCOI: protectedProcedure
+  .input(z.object({
+    assignmentId: z.string(),
+    hasConflict: z.boolean(),
+    conflictType: z.string().optional(),
+    description: z.string().optional(),
+  }))
+  .mutation(async ({ ctx, input }) => {
+    // Creates/updates ConflictOfInterest record
+    // Blocks evaluation access if hasConflict = true until reviewed
+  }),
+
+// Get COI status for an assignment
+getCOIStatus: protectedProcedure
+  .input(z.object({ assignmentId: z.string() }))
+  .query(async ({ ctx, input }) => {
+    // Returns COI declaration status for the assignment
+  }),
+
+// List all COI declarations for a round (admin only)
+listCOIByRound: adminProcedure
+  .input(z.object({ roundId: z.string() }))
+  .query(async ({ ctx, input }) => {
+    // Returns all COI declarations with user and project info
+  }),
+
+// Review a COI declaration (admin only)
+reviewCOI: adminProcedure
+  .input(z.object({
+    id: z.string(),
+    reviewNotes: z.string().optional(),
+  }))
+  .mutation(async ({ ctx, input }) => {
+    // Marks COI as reviewed, logs audit event
+  }),
+```
+
+### AI Evaluation Summary Endpoints (Evaluation Router)
+
+```typescript
+// Added to src/server/routers/evaluation.ts
+
+// Generate AI summary for a project's evaluations
+generateSummary: adminProcedure
+  .input(z.object({
+    projectId: z.string(),
+    roundId: z.string(),
+  }))
+  .mutation(async ({ ctx, input }) => {
+    // Anonymizes evaluation data, sends to GPT
+    // Generates strengths/weaknesses, themes, scoring patterns
+    // Stores EvaluationSummary record
+  }),
+
+// Get existing summary for a project
+getSummary: adminProcedure
+  .input(z.object({
+    projectId: z.string(),
+    roundId: z.string(),
+  }))
+  .query(async ({ ctx, input }) => {
+    // Returns EvaluationSummary with parsed summaryJson
+  }),
+
+// Generate summaries for all projects in a round
+generateBulkSummaries: adminProcedure
+  .input(z.object({
+    roundId: z.string(),
+  }))
+  .mutation(async ({ ctx, input }) => {
+    // Iterates through projects, generates summaries in batch
+    // Returns count of generated summaries
+  }),
+```
+
+### Evaluation Reminders Endpoint
+
+```typescript
+// Added to src/server/routers/evaluation.ts
+
+// Trigger reminder emails for incomplete assignments (admin)
+triggerReminders: adminProcedure
+  .input(z.object({
+    roundId: z.string(),
+  }))
+  .mutation(async ({ ctx, input }) => {
+    // Calls EvaluationRemindersService
+    // Finds incomplete assignments, sends email reminders
+    // Logs to ReminderLog table
+  }),
+```
+
+### Cron API Route
+
+```typescript
+// src/app/api/cron/reminders/route.ts
+
+// POST /api/cron/reminders
+// Protected by CRON_SECRET header validation
+// Automatically finds rounds with approaching deadlines
+// Sends reminder emails to jurors with incomplete evaluations
+// Designed to be called by external cron scheduler
+```
+
+### Export Router - Filtering Results
+
+```typescript
+// Added to src/server/routers/export.ts
+
+// Export filtering results as CSV
+filteringResults: adminProcedure
+  .input(z.object({
+    roundId: z.string(),
+  }))
+  .query(async ({ ctx, input }) => {
+    // Queries projects with evaluations and AI screening data
+    // Dynamically flattens aiScreeningJson columns
+    // Returns CSV-ready data
+  }),
+```
+
+### Evaluation Form Schema - Extended Criterion Types
+
+The evaluation form `criteriaJson` now supports extended criterion types:
+
+```typescript
+type CriterionType = 'numeric' | 'text' | 'boolean' | 'section_header'
+
+type Criterion = {
+  id: string
+  label: string
+  type: CriterionType  // Defaults to 'numeric' for backward compatibility
+  scale?: string       // Only for 'numeric' type
+  weight?: number      // Only for 'numeric' type
+  required?: boolean
+  description?: string
+  // Conditional visibility
+  visibleWhen?: {
+    criterionId: string
+    value: unknown
+  }
+  // Section grouping
+  section?: string
+}
+```
+
+Field components per type:
+- **numeric**: Standard slider/number input with scale
+- **text**: Free-text textarea field (`TextCriterionField`)
+- **boolean**: Yes/No toggle (`BooleanCriterionField`)
+- **section_header**: Non-input visual divider for form organization (`SectionHeader`)
+
 ## Authentication Flow
 
 ### Magic Link Implementation