Merge pull request #78 from datakind/bugfix/pgvector-assistant

Bugfix/pgvector assistant

.github/workflows/e2e_tests.yml

@@ -81,8 +81,8 @@ jobs:
             # TODO this should be enhanced to use a buildx bake to leverage layer caching for faster builds, or push to repo and simply have a pull for the run
             # TODO docker-compose files should be refactored to use scopes instead of different versions for each environment
             echo "Starting docker containers for dbs and server ..."
-            docker-compose pull
-            docker-compose up -d --build datadb recipedb server chat
+            docker compose pull
+            docker compose up -d --build datadb recipedb server chat
 
             # TODO: For some reason, maybe buildkit, in Github docker compose builds the image differently, and it doesn't work. Individual image build works.
             docker build --build-arg OPENAI_API_KEY=$ASSISTANTS_API_KEY \

.github/workflows/get_memory_test.yml

@@ -77,8 +77,8 @@ jobs:
             ls data/datadb
             
             echo "Starting docker containers for dbs and server ..."
-            docker-compose pull
-            docker-compose up -d --build datadb recipedb server datadb
+            docker compose pull
+            docker compose up -d --build datadb recipedb server datadb
 
             sleep 10
 

deployment/deploy_azure.py

@@ -14,6 +14,7 @@
 import docker
 from dotenv import load_dotenv
 
+# client = docker.from_env()
 load_dotenv()
 
 container_registry = os.getenv("AZURE_CONTAINER_REGISTRY")
@@ -25,8 +26,11 @@
 
 if sys.platform == "darwin":
     print("Running on Mac")
+    print(f"container_registry: {container_registry}")
+    print(f"repo: {repo}")
     client = docker.DockerClient(
-        base_url="unix:///Users/matthewharris/.docker/run/docker.sock "
+        # base_url="unix:///Users/matthewharris/.docker/run/docker.sock "
+        base_url="unix:///Users/t.o./.docker/run/docker.sock "
     )
 else:
     client = docker.from_env()
@@ -72,8 +76,8 @@ def deploy():
         sys.exit()
 
     tags = {
-        "data-recipes-ai-server": [f"{container_registry}/{repo}", "server"],
-        "data-recipes-ai-chat": [f"{container_registry}/{repo}", "chat"],
+        "data-recipes-ai-server:latest": [f"{container_registry}/{repo}", "server"],
+        "data-recipes-ai-chat:latest": [f"{container_registry}/{repo}", "chat"],
     }
 
     run_cmd("az login")
@@ -87,13 +91,15 @@ def deploy():
         f"DOCKER_DEFAULT_PLATFORM={azure_platform} && docker compose -f {docker_compose_file} build"
     )
 
+    # run_cmd("docker compose build")
+
     for image in tags.keys():
         print(f"Tagging {image} image ... with tag {tags[image][0]}:{tags[image][1]}")
         client.images.get(image).tag(tags[image][0], tags[image][1])
         print(f"Pushing {image} image ... to {tags[image][0]}:{tags[image][1]}")
         client.images.push(tags[image][0], tags[image][1])
 
-    sys.exit()
+    # sys.exit()
 
     run_cmd(f"docker compose -f {docker_compose_file} down")
     run_cmd(f"docker compose -f {docker_compose_file} pull")

requirements.txt

@@ -18,7 +18,7 @@ langchain_openai==0.1.7
 literalai==0.0.604
 matplotlib==3.9.0
 pandas==2.2.2
-pgvector==0.2.4
+pgvector==0.2.5
 plotly==5.22.0
 psycopg2_binary==2.9.9
 pyarrow==16.1.0

server/robocorp/actions_plugins/recipe-server/package.yaml

@@ -21,8 +21,7 @@ dependencies:
   - robocorp-actions=0.0.7
   - langchain=0.1.16 
   - langchain_community=0.0.34
-  - langchain_openai=0.1.3
-  - psycopg2-binary=2.9.3 
+  - langchain_openai=0.1.3 
   - pgvector=0.2.5
   - python-dotenv=0.19.1
   - pillow=10.3.0
@@ -32,3 +31,4 @@ dependencies:
   - seaborn=0.13.2
   - geopandas=0.10.2
   - hdx_python_api=6.2.4
+  - psycopg2-binary=2.9.3

server/robocorp/package.yaml

@@ -23,6 +23,6 @@ dependencies:
   - langchain=0.1.16
   - langchain_community=0.0.34
   - langchain_openai=0.1.3
-  - psycopg2-binary=2.9.3
   - pgvector=0.2.5
+  - psycopg2-binary=2.9.3
 

utils/recipes.py

@@ -96,6 +96,7 @@ def initialize_vector_db():
             collection_name=COLLECTION_NAME,
             connection_string=CONNECTION_STRING,
             embedding_function=embedding_model,
+            create_extension=False,
         )
 
     return db