Env vars and docker files review (#70)

* refactor: update .env.sample

* chore: unused setting

* feat: update docker files

* feat: add docker build + push step

* chore: format file

* chore: reorder items in .env.sample

* feat: update docker files

* feat: add `:latest` docker tag in CI

* chore: default port

.env.sample

@@ -1,36 +1,39 @@
-## Secrets that must be available as environment variables
-# Used by LLM API services (api keys or access tokens for instance)
-# and by other third party services that need specific env variables
-BREVIA_ENV_SECRETS='{
-    "OPENAI_API_KEY": "#########",
-    "COHERE_API_KEY": "#####"
-}'
-
-## QA vars
-# QA completion LLM
-QA_COMPLETION_LLM='{
-    "_type": "openai-chat",
-    "model_name": "gpt-4o-mini",
-    "temperature": 0,
-    "max_tokens": 200,
-    "verbose": true
-}'
-
-# QA followup LLM
-QA_FOLLOWUP_LLM='{
-    "_type": "openai-chat",
-    "model_name": "gpt-4o-mini",
-    "temperature": 0,
-    "max_tokens": 200,
-    "verbose": true
-}'
-
-QA_FOLLOWUP_SIM_THRESHOLD=0.735
+### Secrets that must be available as environment variables
+# You can define here secrets used by LLM API services (api keys or access tokens for instance)
+# and by other third party services that need specific env variables.
+# If you don't set these variables, the services that depend on them may not work properly.
+# By using BREVIA_ENV_SECRETS you make sure that these secrets will be available as environment variables.
+# BREVIA_ENV_SECRETS='{"OPENAI_API_KEY": "#########", "COHERE_API_KEY": "#########"}'
+
+## OpenAI, Cohere, and other API secrets
+# If these vars are not available as environment variables you may want to add them to BREVIA_ENV_SECRETS (see above)
+# OPENAI_API_KEY=#########
+# COHERE_API_KEY=#########
+
+## Tracing - Langsmith: if you have an account you need the env vars listed below, edit API_KEY and PROJECT vars
+# If these vars are not available as environment variables you may want to add them to BREVIA_ENV_SECRETS (see above)
+# LANGCHAIN_TRACING_V2=true
+# LANGCHAIN_ENDPOINT="https://api.smith.langchain.com"
+# LANGCHAIN_API_KEY="########"
+# LANGCHAIN_PROJECT="My Project"
+
+### QA vars
+# Completion LLM configuration
+# QA_COMPLETION_LLM='{"_type": "openai-chat", "model_name": "gpt-4o-mini", "temperature": 0, "max_tokens": 200, "verbose": true}'
+
+# Followup LLM configuration
+# QA_FOLLOWUP_LLM='{"_type": "openai-chat", "model_name": "gpt-4o-mini", "temperature": 0, "max_tokens": 200, "verbose": true}'
+
+# Followup similarity threshold
+# QA_FOLLOWUP_SIM_THRESHOLD=0.735
+
 # Chat history - uncomment to disable session conversation, avoiding chat history loading
 # QA_NO_CHAT_HISTORY=True
+
 # Uncomment to use a custom QA retriever with custom arguments
-# QA_RETRIEVER='{"retriever": "my_project.CustomSplitter", "some_var": "some_value"}'
+# QA_RETRIEVER='{"retriever": "my_project.CustomRetriever", "some_var": "some_value"}'
 
+### Security
 # Access tokens secret - if missing no access token validity is checked
 # Generate it with: openssl rand -hex 32
 # TOKENS_SECRET=##########################
@@ -41,68 +44,57 @@ QA_FOLLOWUP_SIM_THRESHOLD=0.735
 # Special token used by `GET /status` only
 # STATUS_TOKEN=#############
 
-## Search
+### Search
 # default number of documents to return in a vector search
-SEARCH_DOCS_NUM=4
+# SEARCH_DOCS_NUM=4
 
-## Embedding
-EMBEDDINGS='{"_type": "openai-embeddings"}'
+### Embedding
+# Embeddings engine configuration
+# EMBEDDINGS='{"_type": "openai-embeddings"}'
 
-## Index creation
-TEXT_CHUNK_SIZE=2000
-TEXT_CHUNK_OVERLAP=100
+### Index creation
+# TEXT_CHUNK_SIZE=2000
+# TEXT_CHUNK_OVERLAP=100
 # Uncomment to use a custom splitter with custom arguments
 # TEXT_SPLITTER='{"splitter": "my_project.CustomSplitter", "some_var": "some_value"}'
 
-## Summarize
+### Summarize
 # Summarization LLM
-SUMMARIZE_LLM='{
-    "_type": "openai-chat",
-    "model_name": "gpt-4o-mini",
-    "temperature": 0,
-    "max_tokens": 2000
-}'
-SUMM_TOKEN_SPLITTER=4000
-SUMM_TOKEN_OVERLAP=500
-SUMM_DEFAULT_CHAIN=stuff
-
-# General
-VERBOSE_MODE=True
-
-# OPENAPI
+# SUMMARIZE_LLM='{"_type": "openai-chat", "model_name": "gpt-4o-mini", "temperature": 0, "max_tokens": 2000}'
+# SUMM_TOKEN_SPLITTER=4000
+# SUMM_TOKEN_OVERLAP=500
+# SUMM_DEFAULT_CHAIN=stuff
+
+### General
+# VERBOSE_MODE=True
+
+### OPENAPI
 # Uncomment to block openapi urls - /docs, /redoc
 # BLOCK_OPENAPI_URLS=True
 
-## Feature flags
-#
-# Uncomment to detect question language with Spacy
-# FEATURE_QA_LANG_DETECT=True
-
-## DB vars
-PGVECTOR_DRIVER=psycopg2
+### Database connection
+# PGVECTOR_DRIVER=psycopg2
 # host var is used on local development
 # but ignored by docker compose
-PGVECTOR_HOST=localhost
-PGVECTOR_PORT=5432
-PGVECTOR_DATABASE=brevia
-PGVECTOR_USER=postgres
-PGVECTOR_PASSWORD=postgres
-PGVECTOR_POOL_SIZE=10
-# used by docker compose only
-PGVECTOR_VOLUME_NAME=pgvector-data
-
-## PG ADMIN vars
-PGADMIN_DEFAULT_EMAIL=[email protected]
-PGADMIN_DEFAULT_PASSWORD=admin
-PGADMIN_PORT=4000
-# used by docker compose only
-PGADMIN_VOLUME_NAME=pgadmin-data
-
-## Tracing - Langsmith
-# if you have an account uncomment lines below and edit API_KEY and PROJECT vars
-# NB: You may want to add these vars to BREVIA_ENV_SECRETS in a local development environment
-# LANGCHAIN_TRACING_V2=true
-# LANGCHAIN_ENDPOINT="https://api.smith.langchain.com"
-# LANGCHAIN_API_KEY="########"
-# LANGCHAIN_PROJECT="My Project"
+# PGVECTOR_HOST=localhost
+# PGVECTOR_PORT=5432
+# PGVECTOR_DATABASE=brevia
+# PGVECTOR_USER=postgres
+# PGVECTOR_PASSWORD=postgres
+# PGVECTOR_POOL_SIZE=10
+
+### Docker compose
+## Postgres pgvector
+# PGVECTOR_VOLUME_NAME=pgdata
+## PG ADMIN
+# [email protected]
+# PGADMIN_DEFAULT_PASSWORD=admin
+# PGADMIN_PORT=4000
+# PGADMIN_VOLUME_NAME=pgadmindata
+## BREVIA API
+# BREVIA_API_PORT=8000
+# BREVIA_API_VOLUME_NAME=breviaapidata
+## BREVIA APP
+# BREVIA_APP_PORT=3000
+# BREVIA_APP_VOLUME_NAME=breviaappdata
 

.github/workflows/release.yml

@@ -14,6 +14,9 @@ on:
           - patch
           - minor
           - major
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}
 
 jobs:
   release_job:
@@ -55,3 +58,22 @@ jobs:
         uses: JRubics/[email protected]
         with:
           pypi_token: ${{ secrets.PYPI_TOKEN }}
+
+      - name: Log in to the Container registry
+        if: ${{ !contains(needs.release_job.outputs.version, 'undefined') }}
+        uses: docker/login-action@65b78e6e13532edd9afa3aa52ac7964289d1a9c1
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Build and push Docker image
+        if: ${{ !contains(needs.release_job.outputs.version, 'undefined') }}
+        id: push
+        uses: docker/build-push-action@f2a1d5e99d037542a71f64918e516c093c6f3fc4
+        with:
+          context: .
+          push: true
+          tags: |
+            ${{ env.REGISTRY }}/${{ github.repository }}:${{ needs.release_job.outputs.version }}
+            ${{ env.REGISTRY }}/${{ github.repository }}:latest

Dockerfile

@@ -1,19 +1,24 @@
 # syntax=docker/dockerfile:1
-FROM python:3.10-alpine
+FROM python:3.12-alpine
+
+RUN apk add g++
 
 # Extra python env
 ENV PYTHONDONTWRITEBYTECODE=0
 ENV PYTHONUNBUFFERED=1
 ENV PIP_DISABLE_PIP_VERSION_CHECK=1
+ENV POETRY_VIRTUALENVS_CREATE=false
 
 RUN pip install poetry
 
 WORKDIR /python-docker
 
 COPY . /python-docker/
 
-RUN poetry install
+RUN poetry install --no-interaction --no-ansi
+
+RUN apk del g++
 
 EXPOSE 8000
 
-CMD ["poetry", "run", "uvicorn", "--host", "0.0.0.0", "--port", "8000", "main:app"]
+CMD poetry run db_upgrade && poetry run uvicorn --host 0.0.0.0 --port 8000 main:app

brevia/settings.py

@@ -85,7 +85,6 @@ class Settings(BaseSettings):
     qa_no_chat_history: bool = False  # don't load chat history
     qa_followup_sim_threshold: float = 0.735  # similitude threshold in followup
     qa_retriever: Json[dict[str, Any]] = '{}'  # custom retriever settings
-    feature_qa_lang_detect: bool = False
 
     # Summarization
     summ_default_chain: str = 'stuff'

docker-compose.yml

@@ -1,27 +1,55 @@
 version: '3.9'
 services:
-    pgdatabase:
-        image: pgvector/pgvector:0.7.4-pg16
-        environment:
-            - POSTGRES_USER=${PGVECTOR_USER}
-            - POSTGRES_PASSWORD=${PGVECTOR_PASSWORD}
-            - POSTGRES_DB=${PGVECTOR_DATABASE}
-        volumes:
-            - pgdata:/var/lib/postgresql/data
-        ports:
-            - "${PGVECTOR_PORT}:5432"
-    pgadmin:
-        image: dpage/pgadmin4
-        profiles: ["admin"]
-        environment:
-            - PGADMIN_DEFAULT_EMAIL=${PGADMIN_DEFAULT_EMAIL}
-            - PGADMIN_DEFAULT_PASSWORD=${PGADMIN_DEFAULT_PASSWORD}
-        ports:
-            - "${PGADMIN_PORT}:80"
-        volumes:
-          - pgadmindata:/var/lib/pgadmin
+  pgdatabase:
+    image: pgvector/pgvector:0.7.4-pg16
+    environment:
+      - POSTGRES_USER=${PGVECTOR_USER:-postgres}
+      - POSTGRES_PASSWORD=${PGVECTOR_PASSWORD:-postgres}
+      - POSTGRES_DB=${PGVECTOR_DATABASE:-brevia}
+    volumes:
+      - pgdata:/var/lib/postgresql/data
+    ports:
+      - 5433:${PGVECTOR_PORT:-5432}
+  pgadmin:
+    image: dpage/pgadmin4
+    profiles: ["admin"]
+    environment:
+      - PGADMIN_DEFAULT_EMAIL=${PGADMIN_DEFAULT_EMAIL:[email protected]}
+      - PGADMIN_DEFAULT_PASSWORD=${PGADMIN_DEFAULT_PASSWORD:-admin}
+    ports:
+      - "${PGADMIN_PORT:-4000}:80"
+    volumes:
+      - pgadmindata:/var/lib/pgadmin
+  breviaapi:
+    image: ghcr.io/brevia-ai/brevia:latest
+    profiles: ["api"]
+    environment:
+      - PGVECTOR_HOST=pgdatabase
+      - PGVECTOR_PORT=${PGVECTOR_PORT:-5432}
+      - PGVECTOR_DATABASE=${PGVECTOR_DATABASE:-brevia}
+      - PGVECTOR_USER=${PGVECTOR_USER:-postgres}
+      - PGVECTOR_PASSWORD=${PGVECTOR_PASSWORD:-postgres}
+      - BREVIA_ENV_SECRETS=${BREVIA_ENV_SECRETS}
+    ports:
+      - "${BREVIA_API_PORT:-8000}:8000"
+    volumes:
+      - breviaapidata:/var/lib/breviaapi
+  breviaapp:
+    image: ghcr.io/brevia-ai/brevia-app:latest
+    profiles: ["admin", "api"]
+    environment:
+      - NUXT_API_BASE_URL=http://breviaapi:${BREVIA_API_PORT:-8000}
+    ports:
+      - "${BREVIA_APP_PORT:-3000}:3000"
+    volumes:
+      - breviaappdata:/var/lib/breviaapp
+
 volumes:
   pgdata:
-    name: ${PGVECTOR_VOLUME_NAME}
+    name: ${PGVECTOR_VOLUME_NAME:-pgdata}
   pgadmindata:
-    name: ${PGADMIN_VOLUME_NAME}
+    name: ${PGADMIN_VOLUME_NAME:-pgadmindata}
+  breviaappdata:
+    name: ${BREVIA_APP_VOLUME_NAME:-breviaapp}
+  breviaapidata:
+    name: ${BREVIA_API_VOLUME_NAME:-breviaapi}