feat: add demo endpoints with drift processing for driftos-core

- Separated LLM routes into /llm and /demo prefixes (matching embed-enterprise)
- Created /demo/stream endpoint for simple streaming (uses server-side Groq API key)
- Created /demo/chat endpoint with full drift processing:
  - Creates temporary conversation in database
  - Processes messages through LLM-based drift orchestrator
  - Streams response and returns final state (branches, messages, actions)
  - Cleans up demo conversation after completion
- Updated auth plugin to allow /demo routes publicly (rate limited)
- Maintains same API surface as embed-enterprise for consistent testing

Note: Core uses LLM-based routing (classify-route) vs embed's semantic similarity,
but returns similar drift decisions (BRANCH/STAY/ROUTE) and state structure.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: scotty595

src/app.ts

@@ -17,6 +17,7 @@ import driftRoutes from './routes/drift/index';
 import factsRoutes from './routes/facts/index';
 import contextRoutes from './routes/context/index';
 import llmRoutes from './routes/llm/index';
+import demoRoutes from './routes/demo/index';
 import conversationsRoutes from './routes/conversations/index';
 
 export async function buildApp() {
@@ -61,6 +62,7 @@ export async function buildApp() {
       await fastify.register(factsRoutes, { prefix: '/facts' });
       await fastify.register(contextRoutes, { prefix: '/context' });
       await fastify.register(llmRoutes, { prefix: '/llm' });
+      await fastify.register(demoRoutes, { prefix: '/demo' });
       await fastify.register(conversationsRoutes, { prefix: '/conversations' });
     },
     { prefix: `${app.config.API_PREFIX}/${app.config.API_VERSION}` }

src/plugins/auth.ts

@@ -4,8 +4,8 @@ import type { FastifyRequest, FastifyReply } from 'fastify';
 
 // Routes that don't require authentication
 const PUBLIC_ROUTES = [
-  '/api/v1/llm/demo/stream', // Demo endpoint - rate limited by IP
-  '/api/v1/llm/demo/chat', // Demo chat with drift - rate limited by IP
+  '/api/v1/demo/stream', // Demo streaming endpoint - rate limited by IP
+  '/api/v1/demo/chat', // Demo chat with drift - rate limited by IP
   '/api/v1/health', // Health checks for infrastructure
   '/health', // Root health check
   '/', // Root route