fix: Fix docker compose files across the board

Signed-off-by: Diwank Tomer <diwank@julep.ai>

.env.example

@@ -1,54 +1,69 @@
-AGENTS_API_KEY=myauthkey
-AGENTS_API_KEY_HEADER_NAME=Authorization
-AGENTS_API_URL=http://agents-api:8080
-COZO_AUTH_TOKEN=myauthkey
-COZO_HOST=http://memory-store:9070
-COZO_PORT=9070
-COZO_ROCKSDB_DIR=cozo.db
-DTYPE=float16
-EMBEDDING_SERVICE_BASE=http://text-embeddings-inference
-EMBEDDING_SERVICE_URL=${EMBEDDING_SERVICE_BASE}/embed
-GATEWAY_PORT=80
-GPU_MEMORY_UTILIZATION=0.90
-
-HF_TOKEN=
-HUGGING_FACE_HUB_TOKEN=
-JWT_SHARED_KEY=
-
-MAX_MODEL_LEN=8192
-MAX_NUM_SEQS=1
-MNT_DIR=/data
-
+# Security
+# --------
+JWT_SHARED_KEY=<your_jwt_shared_key>
+AGENTS_API_KEY=<your_agents_api_key>
+COZO_AUTH_TOKEN=<your_cozo_auth_token>
+TEMPORAL_POSTGRES_PASSWORD=<your_temporal_postgres_password>
+LITELLM_POSTGRES_PASSWORD=<your_litellm_postgres_password>
+LITELLM_MASTER_KEY=<your_litellm_master_key>
+LITELLM_REDIS_PASSWORD=<your_litellm_redis_password>
 SKIP_CHECK_DEVELOPER_HEADERS=true
-SUMMARIZATION_TOKENS_THRESHOLD=2048
-TEMPERATURE_SCALING_FACTOR=0.9
-TEMPERATURE_SCALING_POWER=0.9
-TEMPORAL_ENDPOINT=temporal:7233
-TEMPORAL_NAMESPACE=default
-TEMPORAL_WORKER_URL=temporal:7233
-TP_SIZE=1
-TRUNCATE_EMBED_TEXT=true
-TRAEFIK_LOG_LEVEL=DEBUG
-WORKER_URL=temporal:7233
-
-AGENTS_API_DEBUG=false
-OPENAI_API_KEY=
-ANTHROPIC_API_KEY=
-GROQ_API_KEY=
-CLOUDFLARE_API_KEY=
-CLOUDFLARE_ACCOUNT_ID=
-NVIDIA_NIM_API_KEY=
-GITHUB_API_KEY=
-VOYAGE_API_KEY=
-GOOGLE_APPLICATION_CREDENTIALS=
-
-LITELLM_URL=http://litellm:4000
-POSTGRES_DB=litellm
-POSTGRES_USER=llmproxy
-POSTGRES_PASSWORD=
-LITELLM_DATABASE_URL=postgresql://${POSTGRES_USER}:${POSTGRES_PASSWORD}@litellm-db:5432/${POSTGRES_DB}
-LITELLM_MASTER_KEY=
-LITELLM_REDIS_HOST=litellm-redis
-LITELLM_REDIS_PORT=6379
-LITELLM_REDIS_PASSWORD=
-REDIS_ARGS="--requirepass ${LITELLM_REDIS_PASSWORD}"
+EMBEDDING_SERVICE_BASE=http://text-embeddings-inference-<gpu|cpu>  # Use the 'gpu' profile to run on GPU
+
+# Memory Store
+# -----------
+
+# COZO_HOST=http://memory-store:9070
+# COZO_PORT=9070
+# COZO_ROCKSDB_DIR=cozo.db
+# COZO_BACKUP_DIR=backup
+# COZO_MNT_DIR=/data
+
+# Gateway
+# ------
+
+# GATEWAY_PORT=80
+# TRAEFIK_LOG_LEVEL=INFO
+
+# Agents API
+# ---------
+
+# AGENTS_API_KEY_HEADER_NAME=Authorization
+# AGENTS_API_URL=http://agents-api:8080
+# TRUNCATE_EMBED_TEXT=true
+# WORKER_URL=temporal:7233
+# AGENTS_API_DEBUG=false
+# EMBEDDING_MODEL_ID=Alibaba-NLP/gte-large-en-v1.5
+# NUM_GPUS=1
+
+# Temporal
+# --------
+
+# TEMPORAL_ENDPOINT=temporal:7233
+# TEMPORAL_NAMESPACE=default
+# TEMPORAL_WORKER_URL=temporal:7233
+# TEMPORAL_POSTGRES_DB=temporal
+# TEMPORAL_POSTGRES_USER=temporal
+
+# LiteLLM
+# -------
+
+# LITELLM_URL=http://litellm:4000
+# LITELLM_POSTGRES_DB=litellm
+# LITELLM_POSTGRES_USER=llmproxy
+# LITELLM_REDIS_HOST=litellm-redis
+# LITELLM_REDIS_PORT=6379
+
+# LLM Providers
+# --------------
+
+# OPENAI_API_KEY=<your_openai_api_key>
+# HUGGING_FACE_HUB_TOKEN=<your_hugging_face_hub_token>
+# ANTHROPIC_API_KEY=<your_anthropic_api_key>
+# GROQ_API_KEY=<your_groq_api_key>
+# CLOUDFLARE_API_KEY=<your_cloudflare_api_key>
+# CLOUDFLARE_ACCOUNT_ID=<your_cloudflare_account_id>
+# NVIDIA_NIM_API_KEY=<your_nvidia_nim_api_key>
+# GITHUB_API_KEY=<your_github_api_key>
+# VOYAGE_API_KEY=<your_voyage_api_key>
+# GOOGLE_APPLICATION_CREDENTIALS=.keys/julep-vertexai-svc.json

agents-api/docker-compose.yml

@@ -1,16 +1,49 @@
 name: julep-agents-api
 
+# Base for embedding service
+x--text-embeddings-inference: &text-embeddings-inference
+  container_name: text-embeddings-inference
+  environment:
+    - MODEL_ID=${EMBEDDING_MODEL_ID:-Alibaba-NLP/gte-large-en-v1.5}
+
+  image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5
+  ports:
+    - "8082:80"
+  volumes:
+    - ~/.cache/huggingface/hub:/data
+
+# Shared environment variables
+x-shared-environment: &shared-environment
+  AGENTS_API_KEY: ${AGENTS_API_KEY}
+  LITELLM_MASTER_KEY: ${LITELLM_MASTER_KEY}
+  COZO_AUTH_TOKEN: ${COZO_AUTH_TOKEN}
+  SKIP_CHECK_DEVELOPER_HEADERS: ${SKIP_CHECK_DEVELOPER_HEADERS:-True}
+  AGENTS_API_KEY_HEADER_NAME: ${AGENTS_API_KEY_HEADER_NAME:-Authorization}
+  AGENTS_API_URL: ${AGENTS_API_URL:-http://agents-api:8080}
+  TRUNCATE_EMBED_TEXT: ${TRUNCATE_EMBED_TEXT:-False}
+  WORKER_URL: ${WORKER_URL:-temporal:7233}
+  DEBUG: ${AGENTS_API_DEBUG:-False}
+  EMBEDDING_SERVICE_BASE: ${EMBEDDING_SERVICE_BASE:-http://text-embeddings-inference}
+  EMBEDDING_MODEL_ID: ${EMBEDDING_MODEL_ID:-Alibaba-NLP/gte-large-en-v1.5}
+  LITELLM_URL: ${LITELLM_URL:-http://litellm:4000}
+  COZO_HOST: ${COZO_HOST:-http://memory-store:9070}
+  SUMMARIZATION_MODEL_NAME: ${SUMMARIZATION_MODEL_NAME:-gpt-4-turbo}
+  TEMPORAL_WORKER_URL: ${TEMPORAL_WORKER_URL:-temporal:7233}
+  TEMPORAL_NAMESPACE: ${TEMPORAL_NAMESPACE:-default}
+  TEMPORAL_ENDPOINT: ${TEMPORAL_ENDPOINT:-temporal:7233}
+  TEMPORAL_TASK_QUEUE: ${TEMPORAL_TASK_QUEUE:-julep-task-queue}
+
 services:
   agents-api:
     image: julepai/agents-api:${TAG:-dev}
-    env_file:
-      - ../.env
     container_name: agents-api
     depends_on:
       memory-store:
         condition: service_started
       worker:
         condition: service_started
+    environment:
+      <<: *shared-environment
     build:
       context: .
       dockerfile: Dockerfile
@@ -31,15 +64,12 @@ services:
 
   worker:
     image: julepai/worker:${TAG:-dev}
-    env_file:
-      - ../.env
-
+    environment:
+      <<: *shared-environment
     build:
       context: .
       dockerfile: Dockerfile.worker
     depends_on:
-      text-embeddings-inference:
-        condition: service_started
       temporal:
         condition: service_started
 
@@ -55,32 +85,32 @@ services:
         - action: rebuild
           path: Dockerfile.worker
 
-  text-embeddings-inference:
-    container_name: text-embeddings-inference
+  text-embeddings-inference-cpu:
+    <<: *text-embeddings-inference
+    profiles:
+      - ''  # Acts as a default profile. See: https://stackoverflow.com/questions/75758174/how-to-make-profile-default-for-docker-compose
+
+  text-embeddings-inference-gpu:
+    <<: *text-embeddings-inference
+    profiles:
+      - gpu
+    image: ghcr.io/huggingface/text-embeddings-inference:1.5
     environment:
       - DTYPE=float16
-      - MODEL_ID=Alibaba-NLP/gte-large-en-v1.5
+      - MODEL_ID=${EMBEDDING_MODEL_ID:-Alibaba-NLP/gte-large-en-v1.5}
+      - NVIDIA_VISIBLE_DEVICES=all
 
-    image: ghcr.io/huggingface/text-embeddings-inference:1.5
-    ports:
-      - "8082:80"
-    volumes:
-      - ~/.cache/huggingface/hub:/data
     shm_size: "2gb"
     deploy:
       resources:
         reservations:
           devices:
             - driver: nvidia
-              count: all
+              count: ${NUM_GPUS:-1}
               capabilities: [gpu]
 
-
   cozo-migrate:
     image: julepai/cozo-migrate:${TAG:-dev}
-    env_file:
-      - ../.env
-
     container_name: cozo-migrate
     depends_on:
       memory-store:
@@ -89,6 +119,9 @@ services:
       context: .
       dockerfile: Dockerfile.migration
     restart: "no" # Make sure to double quote this
+    environment:
+      - COZO_HOST=${COZO_HOST:-http://cozo:9070}
+      - COZO_AUTH_TOKEN=${COZO_AUTH_TOKEN:-myauthkey}
 
     develop:
       watch:

gateway/docker-compose.yml

@@ -3,7 +3,6 @@ name: julep-gateway
 services:
   gateway:
     image: julepai/gateway:${TAG:-dev}
-    env_file: "../.env"
     environment:
       - GATEWAY_PORT=80
       - JWT_SHARED_KEY=${JWT_SHARED_KEY}
@@ -14,7 +13,6 @@ services:
       - AGENTS_API_KEY=${AGENTS_API_KEY}
       - AGENTS_API_KEY_HEADER_NAME=${AGENTS_API_KEY_HEADER_NAME}
       - TRAEFIK_LOG_LEVEL=${TRAEFIK_LOG_LEVEL}
-
     container_name: gateway
     depends_on:
       agents-api: