Qdrant

.docker/app/Dockerfile

@@ -0,0 +1,25 @@
+# Use an official Python image
+FROM python:3.11
+
+# Set environment variables
+ENV PYTHONUNBUFFERED=1
+ENV POETRY_HOME="/opt/poetry"
+ENV PATH="$POETRY_HOME/bin:$PATH"
+
+# Install Poetry
+RUN curl -sSL https://install.python-poetry.org | python3 -
+
+# Set up the working directory
+WORKDIR /app
+ENV PYTHONPATH="/app"
+
+# Copy project files
+COPY pyproject.toml poetry.lock ./
+RUN poetry install --no-root --no-interaction --no-ansi
+
+# Copy the entire application
+COPY config.yml main.py ./
+COPY src ./src
+
+# Run the application
+ENTRYPOINT ["poetry", "run", "python", "main.py"]

.docker/vector_storage_loader/add_snapshots.sh

@@ -0,0 +1,99 @@
+#!/bin/bash
+set -e  # Exit on error
+
+# Install curl if not present
+apk add --no-cache curl || echo "Warning: Unable to install curl, assuming it's already present"
+
+echo "Waiting for vector storage to be ready..."
+BASEURL="http://${VECTOR_STORAGE_HOST:-localhost}:6333"
+echo "Using BASEURL: $BASEURL"
+
+# Wait for Qdrant to be ready
+until curl -sSf "$BASEURL/healthz" > /dev/null; do
+    echo "Waiting for Qdrant health check..."
+    sleep 5
+done
+echo "Vector storage is ready!"
+
+# Check and handle PROJECT_COLLECTION
+if [ -n "$PROJECT_COLLECTION" ]; then
+    echo "Verifying PROJECT_COLLECTION: $PROJECT_COLLECTION"
+
+    # Validate vector size and distance if provided
+    VECTOR_SIZE="${PROJECT_COLLECTION_VECTOR_SIZE:-1536}"
+    VECTOR_DISTANCE="${PROJECT_COLLECTION_VECTOR_DISTANCE:-Cosine}"
+
+    # Validate vector size is a positive integer
+    if ! [[ "$VECTOR_SIZE" =~ ^[0-9]+$ ]] || [ "$VECTOR_SIZE" -le 0 ]; then
+        echo "Error: PROJECT_COLLECTION_VECTOR_SIZE must be a positive integer, got: $VECTOR_SIZE"
+        VECTOR_SIZE=1536
+        echo "Falling back to default size: $VECTOR_SIZE"
+    fi
+
+    # Validate vector distance
+    case "$VECTOR_DISTANCE" in
+        "Cosine"|"Euclid"|"Dot")
+            echo "Using vector distance: $VECTOR_DISTANCE"
+            ;;
+        *)
+            echo "Error: PROJECT_COLLECTION_VECTOR_DISTANCE must be 'Cosine', 'Euclid', or 'Dot', got: $VECTOR_DISTANCE"
+            VECTOR_DISTANCE="Cosine"
+            echo "Falling back to default distance: $VECTOR_DISTANCE"
+            ;;
+    esac
+
+    # Check if collection exists
+    EXISTS_RESPONSE=$(curl -s -X GET "$BASEURL/collections/$PROJECT_COLLECTION/exists")
+    echo "Exists response: $EXISTS_RESPONSE"
+
+    if echo "$EXISTS_RESPONSE" | grep -q '"exists":true'; then
+        echo "Collection '$PROJECT_COLLECTION' already exists. Continuing with existing collection."
+    else
+        echo "Collection '$PROJECT_COLLECTION' does not exist. Creating empty collection..."
+        CREATE_RESPONSE=$(curl -s -X PUT "$BASEURL/collections/$PROJECT_COLLECTION" \
+            -H "Content-Type: application/json" \
+            -d "{\"vectors\": {\"size\": $VECTOR_SIZE, \"distance\": \"$VECTOR_DISTANCE\"}}")
+        echo "Create response: $CREATE_RESPONSE"
+        if echo "$CREATE_RESPONSE" | grep -q '"status":"ok"'; then
+            echo "Empty collection '$PROJECT_COLLECTION' created successfully with size $VECTOR_SIZE and distance $VECTOR_DISTANCE."
+        else
+            echo "Error creating collection. Continuing anyway..."
+        fi
+    fi
+else
+    echo "PROJECT_COLLECTION not specified. Continuing with snapshot restoration only."
+fi
+
+# Set the snapshot directory path
+SNAPSHOT_DIR="/snapshots"
+echo "Looking for snapshots in: $SNAPSHOT_DIR"
+
+# Loop over snapshot files
+for snapshot in "$SNAPSHOT_DIR"/*.snapshot; do
+    if [ -e "$snapshot" ]; then
+        COLLECTION_NAME=$(basename "$snapshot" .snapshot)
+        echo "Processing snapshot for collection: $COLLECTION_NAME"
+
+        EXISTS_RESPONSE=$(curl -s -X GET "$BASEURL/collections/$COLLECTION_NAME/exists")
+        echo "Exists response: $EXISTS_RESPONSE"
+
+        if echo "$EXISTS_RESPONSE" | grep -q '"exists":true'; then
+            echo "Collection '$COLLECTION_NAME' already exists. Skipping upload."
+        else
+            echo "Restoring snapshot for collection '$COLLECTION_NAME'..."
+            UPLOAD_RESPONSE=$(curl -s -X POST "$BASEURL/collections/$COLLECTION_NAME/snapshots/upload" \
+                -F "snapshot=@$snapshot")
+            echo "Upload response: $UPLOAD_RESPONSE"
+            if echo "$UPLOAD_RESPONSE" | grep -q '"status":"ok"'; then
+                echo "Snapshot restored successfully."
+            else
+                echo "Error uploading snapshot."
+            fi
+        fi
+    else
+        echo "No snapshots found in $SNAPSHOT_DIR"
+        break
+    fi
+done
+
+echo "Snapshots restoration finished successfully."

.gitignore

@@ -199,4 +199,15 @@ pyrightconfig.json
 
 .assets/
 !.assets/.gitkeep
-.localstorage
+.localstorage
+pandoc*
+
+/.docker/vector_storage/storage/
+/.docker/memory_storage/db/
+/.docker/vector_storage_loader/
+/.docker/ollama/
+!/.docker/ollama/.gitkeep
+!/.docker/vector_storage_loader/storage/.gitkeep
+!/.docker/memory_storage/db/.gitkeep
+
+

.python-version

@@ -0,0 +1 @@
+3.11

config.yml

@@ -16,18 +16,23 @@ core:
 
 ai:
   openai:
+    embedding_model_name: ${AI_OPENAI_EMBEDDING_MODEL:text-embedding-3-small}
     model_name: ${AI_OPENAI_MODEL:gpt-3.5-turbo}
     api_key:  ${AI_OPENAI_APIKEY}
+  ollama:
+    base_url: ${AI_OLLAMA_BASE_URL:http://localhost:11434}
 
 assistant:
   k: ${ASSISTANT_K:100}
   tokens_limit: ${ASSISTANT_TOKENS_LIMIT:2000}
-  score_threshold: ${ASSISTANT_SCORE_THRESHOLD:0.9}
+  score_threshold: ${ASSISTANT_SCORE_THRESHOLD:0.01}
   distance_threshold: ${DISTANCE_THRESHOLD:null}
 
 storage:
     vector:
-      url: ${STORAGE_VECTOR_URL:postgresql://root:example@localhost:5432/postgres}
+      url: ${STORAGE_VECTOR_URL:http://localhost:6333}
+      collection_name: ${STORAGE_VECTOR_COLLECTION_NAME:my_collection}
+      timeout: ${STORAGE_VECTOR_TIMEOUT:60}
       pre_delete_collection: false
     memory:
       url: ${STORAGE_MEMORY_URL:mongodb://root:example@localhost:27017}

docker-compose.yml

@@ -1,26 +1,87 @@
-version: '3'
-
 services:
-  vector_storage:
-    image: ankane/pgvector
+  vector-storage:
+    image: qdrant/qdrant:latest
     container_name: vector_storage
-    restart: always
+    restart: unless-stopped
     ports:
-      - 5432:5432
-    environment:
-      - POSTGRES_USER=root
-      - POSTGRES_PASSWORD=example
-      - POSTGRES_DB=postgres
+      - "6333:6333"
+      - "6334:6334"
+    volumes:
+      - ./.docker/vector_storage/storage:/qdrant/storage
 
+  vector-storage-loader:
+    image: alpine:latest
+    profiles:
+      - loader
+    container_name: vector_storage_loader
     volumes:
-      - .docker/postgres/docker-entrypoint-initdb.d:/docker-entrypoint-initdb.d
+      - ./.docker/vector_storage_loader/snapshots:/snapshots  # Mount snapshot folder
+      - ./.docker/vector_storage_loader/add_snapshots.sh:/app/add_snapshots.sh:ro  # Mount script
+    environment:
+      VECTOR_STORAGE_HOST: vector_storage
+      PROJECT_COLLECTION: jabref
+      PROJECT_COLLECTION_VECTOR_SIZE: 768 # nomic dimension, for openai use 1536
+      PROJECT_COLLECTION_VECTOR_DISTANCE: Cosine
+    depends_on:
+      - vector-storage
+    entrypoint: ["/bin/sh", "/app/add_snapshots.sh"]
 
-  memory_storage:
+  memory-storage:
     image: mongo
     container_name: memory_storage
-    restart: always
+    restart: unless-stopped
     ports:
-      - 27017:27017
+      - "27017:27017"
     environment:
       MONGO_INITDB_ROOT_USERNAME: root
       MONGO_INITDB_ROOT_PASSWORD: example
+    volumes:
+      - ./.docker/memory_storage/db:/data/db
+
+  ollama:
+    image: ollama/ollama:latest
+    container_name: docgpt_ollama
+    restart: unless-stopped
+    ports:
+      - "11434:11434"
+    volumes:
+      - ./.docker/ollama:/root/.ollama  # Persist Ollama models and data
+    entrypoint: ["/bin/sh", "-c"]
+    environment:
+      OLLAMA_NUM_PARALLEL: 16
+    command:
+      - |
+        ollama serve &  # Start the Ollama server in the background
+        sleep 5 &&      # Give the server a moment to start
+        ollama pull nomic-embed-text &&  # Pull the nomic-embed-text model
+        ollama pull llama2 && # Pull the llama2 model
+        wait            # Keep the container running
+
+  app:
+    container_name: app
+    profiles:
+      - self
+    depends_on:
+      - memory-storage
+      - vector-storage
+      - ollama
+    restart: unless-stopped
+    build:
+      context: .
+      dockerfile: ./.docker/app/Dockerfile
+    env_file: .env
+    ports:
+      - "8000:8000"
+    environment:
+      API_PORT: 8000
+      STORAGE_VECTOR_COLLECTION_NAME: ${STORAGE_VECTOR_COLLECTION_NAME:-mozilla}
+      STORAGE_VECTOR_URL: http://vector_storage:6333
+      STORAGE_MEMORY_URL: mongodb://root:example@memory_storage:27017
+      AI_OLLAMA_BASE_URL: http://docgpt_ollama:11434
+    volumes:
+      - ./.assets:/app/.assets
+
+
+networks:
+  default:
+    name: docgpt-network