feat: integrate Ollama embedding service with mxbai-embed-large model support

.env.example

@@ -48,7 +48,11 @@ MAGIC_LINK_SECRET=27192e6432564f4788d55c15131bd5ac
 NEO4J_AUTH=neo4j/27192e6432564f4788d55c15131bd5ac
 OLLAMA_URL=http://ollama:11434
 
-EMBEDDING_MODEL=text-embedding-3-small
+# Embedding model configuration
+# For self-hosted with Ollama, use: mxbai-embed-large (recommended, 1024 dimensions)
+# For OpenAI, use: text-embedding-3-small
+EMBEDDING_MODEL=mxbai-embed-large
+EMBEDDING_MODEL_SIZE=1024
 MODEL=gpt-4.1-2025-04-14
 
 ## AWS Bedrock ##

apps/webapp/app/env.server.ts

@@ -64,8 +64,9 @@ const EnvironmentSchema = z.object({
   NEO4J_USERNAME: z.string(),
   NEO4J_PASSWORD: z.string(),
 
-  //OpenAI
+  //Model API Key
   OPENAI_API_KEY: z.string(),
+  ANTHROPIC_API_KEY: z.string(),
 
   EMAIL_TRANSPORT: z.string().optional(),
   FROM_EMAIL: z.string().optional(),

apps/webapp/app/lib/model.server.ts

@@ -99,6 +99,9 @@ export async function makeModelCall(
     case "claude-3-7-sonnet-20250219":
     case "claude-3-opus-20240229":
     case "claude-3-5-haiku-20241022":
+    case "claude-sonnet-4-5":
+    case "claude-haiku-4-5":
+    case "claude-opus-4-1":
       modelInstance = anthropic(model, { ...options });
       break;
 
@@ -206,6 +209,7 @@ export async function getEmbedding(text: string) {
     });
     return embedding;
   }
+  console.log("Using Ollama embedding url: ", ollamaUrl);
 
   const ollama = createOllama({
     baseURL: ollamaUrl,

hosting/docker/.env

@@ -43,7 +43,8 @@ NEO4J_PASSWORD=27192e6432564f4788d55c15131bd5ac
 NEO4J_AUTH=neo4j/27192e6432564f4788d55c15131bd5ac
 
 OPENAI_API_KEY=
-OLLAMA_URL=
+ANTHROPIC_API_KEY=
+OLLAMA_URL=http://core-ollama:11434
 
 EMBEDDING_MODEL=text-embedding-3-small
 MODEL=gpt-4.1-2025-04-14

hosting/docker/docker-compose.yaml

@@ -51,6 +51,8 @@ services:
         condition: service_started
       neo4j:
         condition: service_healthy
+      ollama:
+        condition: service_healthy
     networks:
       - core
 
@@ -108,6 +110,25 @@ services:
       retries: 10
       start_period: 20s
 
+  ollama:
+    container_name: core-ollama
+    image: ollama/ollama:0.12.6
+    ports:
+      - "11434:11434"
+    volumes:
+      - ollama_data:/root/.ollama
+      - ./scripts/ollama-init.sh:/usr/local/bin/ollama-init.sh:ro
+    networks:
+      - core
+    entrypoint: ["/bin/bash", "/usr/local/bin/ollama-init.sh"]
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
+      interval: 30s
+      timeout: 10s
+      retries: 5
+      start_period: 90s
+    restart: unless-stopped
+
   webapp:
     container_name: trigger-webapp
     image: ghcr.io/triggerdotdev/trigger.dev@sha256:a19c438f348ac05c939f39ed455ed27b4f189f720b4c9810aef8e71fdc731211
@@ -351,6 +372,7 @@ networks:
 volumes:
   postgres_data:
   neo4j_data:
+  ollama_data:
   shared:
   clickhouse:
   minio:

hosting/docker/scripts/ollama-init.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+set -e
+
+echo "Starting Ollama server..."
+ollama serve &
+OLLAMA_PID=$!
+
+echo "Waiting for Ollama server to be ready..."
+sleep 5
+
+echo "Pulling mxbai-embed-large model..."
+ollama pull mxbai-embed-large
+
+echo "Model pulled successfully!"
+echo "Ollama is ready to accept requests."
+
+# Keep the Ollama server running
+wait $OLLAMA_PID