feat: unified provider resolver with protocol-based dispatch

Extract provider resolution logic into two new modules:
- provider-catalog.ts: vendor presets, default model catalogs, protocol definitions
- provider-resolver.ts: unified resolution for all consumers (Claude SDK, AI SDK, model listing)

DB migration:
- Add protocol, headers_json, env_overrides_json, role_models_json columns to api_providers
- Create provider_models table for per-provider model catalog

Consumer refactoring:
- claude-client.ts: replace inline env-building with toClaudeCodeEnv()
- text-generator.ts: replace inline config with toAiSdkConfig()
- conversation-engine.ts, onboarding, checkin, plan routes: use shared resolveProvider()
- /api/providers/models: read from catalog + provider_models table

UI improvements:
- Merge confusing "Custom API (Anthropic-compatible)" into "Anthropic Third-party API"
- Add 3-field model name mapping (sonnet/opus/haiku) in advanced options
- ProviderForm.tsx: support new fields (protocol, env_overrides, role_models)

Bug fixes:
- Protect auth credentials from being deleted by legacy extra_env empty-string overrides
- Fix env var names: ANTHROPIC_DEFAULT_SONNET → ANTHROPIC_DEFAULT_SONNET_MODEL (SDK convention)

Tests: 229/229 passed. Verified with aiberm (no mapping) and pipellm (with mapping).

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

src/app/api/workspace/checkin/route.ts

@@ -1,7 +1,8 @@
 import { NextRequest, NextResponse } from 'next/server';
 import fs from 'fs';
 import path from 'path';
-import { getSetting, getDefaultProviderId, getSession } from '@/lib/db';
+import { getSetting, getSession } from '@/lib/db';
+import { resolveProvider } from '@/lib/provider-resolver';
 import { loadState, saveState, writeDailyMemory } from '@/lib/assistant-workspace';
 import { getLocalDateString } from '@/lib/utils';
 import { generateTextFromProvider } from '@/lib/text-generator';
@@ -42,9 +43,10 @@ export async function POST(request: NextRequest) {
       return NextResponse.json({ error: 'Invalid answers format' }, { status: 400 });
     }
 
-    // Validate that the calling session belongs to this workspace
+    // Look up the calling session for workspace validation AND provider/model context
+    let session: ReturnType<typeof getSession> | undefined;
     if (sessionId) {
-      const session = getSession(sessionId);
+      session = getSession(sessionId) ?? undefined;
       if (session && session.working_directory !== workspacePath) {
         return NextResponse.json({ error: 'Session does not belong to current workspace' }, { status: 403 });
       }
@@ -66,8 +68,14 @@ export async function POST(request: NextRequest) {
     try { existingUser = fs.readFileSync(userPath, 'utf-8'); } catch { /* new file */ }
 
     try {
-      const providerId = getDefaultProviderId() || '';
-      const model = getSetting('default_model') || 'claude-sonnet-4-20250514';
+      // Resolve using session's provider/model so check-in uses the same provider as chat
+      const resolved = resolveProvider({
+        sessionProviderId: session?.provider_id || undefined,
+        sessionModel: session?.model || undefined,
+      });
+      // Preserve 'env' semantics (see onboarding route for rationale)
+      const providerId = resolved.provider?.id || 'env';
+      const model = resolved.upstreamModel || resolved.model || getSetting('default_model') || 'claude-sonnet-4-20250514';
 
       // Generate daily memory entry (episodic, not destructive)
       const dailyMemoryPrompt = `You maintain daily memory entries for an AI assistant. Given the user's daily check-in answers, generate a daily memory entry for ${today}.

src/app/api/workspace/onboarding/route.ts

@@ -1,7 +1,8 @@
 import { NextRequest, NextResponse } from 'next/server';
 import fs from 'fs';
 import path from 'path';
-import { getSetting, getDefaultProviderId, getSession } from '@/lib/db';
+import { getSetting, getSession } from '@/lib/db';
+import { resolveProvider } from '@/lib/provider-resolver';
 import { loadState, saveState, ensureDailyDir, generateRootDocs } from '@/lib/assistant-workspace';
 import { getLocalDateString } from '@/lib/utils';
 import { generateTextFromProvider } from '@/lib/text-generator';
@@ -62,9 +63,10 @@ export async function POST(request: NextRequest) {
       return NextResponse.json({ error: 'Invalid answers format' }, { status: 400 });
     }
 
-    // Validate that the calling session belongs to this workspace
+    // Look up the calling session for workspace validation AND provider/model context
+    let session: ReturnType<typeof getSession> | undefined;
     if (sessionId) {
-      const session = getSession(sessionId);
+      session = getSession(sessionId) ?? undefined;
       if (session && session.working_directory !== workspacePath) {
         return NextResponse.json({ error: 'Session does not belong to current workspace' }, { status: 403 });
       }
@@ -82,8 +84,15 @@ export async function POST(request: NextRequest) {
     let memoryContent: string;
 
     try {
-      const providerId = getDefaultProviderId() || '';
-      const model = getSetting('default_model') || 'claude-sonnet-4-20250514';
+      // Resolve using session's provider/model so onboarding uses the same provider as chat
+      const resolved = resolveProvider({
+        sessionProviderId: session?.provider_id || undefined,
+        sessionModel: session?.model || undefined,
+      });
+      // Preserve 'env' semantics: when provider=undefined, pass 'env' so text-generator
+      // uses shell env credentials instead of falling back to a random DB provider
+      const providerId = resolved.provider?.id || 'env';
+      const model = resolved.upstreamModel || resolved.model || getSetting('default_model') || 'claude-sonnet-4-20250514';
 
       const soulPrompt = `Based on the following user onboarding answers, generate a concise "soul.md" file that defines an AI assistant's personality, communication style, and behavioral rules. Write in second person ("You are..."). Keep it under 2000 characters. Use markdown headers and bullet points.\n\n${qaText}`;