fix: docker image build

.github/badges/coverage.json

@@ -1 +1 @@
-{"schemaVersion":1,"label":"coverage","message":"71.21%","color":"red"}
+{"schemaVersion":1,"label":"coverage","message":"71.11%","color":"red"}

.gitignore

@@ -219,4 +219,6 @@ __marimo__/
 
 # OpenGateRAG
 config.yml
-.DS_Store
+.DS_Store
+compose.yml
+uv.lock

compose.example.yml

@@ -1,4 +1,4 @@
-name: opengatellm
+name: opengaterag
 
 services:
   rag:

config.example.yml

@@ -2,7 +2,7 @@
 dependencies:
   opengatellm:
     url: ${OPENGATELLM_URL:-http://localhost:8000}
-    model_name: ${OPENGATELLM_MODEL_NAME:-openweight-embedding}
+    model_name: ${OPENGATELLM_MODEL_NAME:-openweight-embeddings}
     model_vector_size: ${OPENGATELLM_MODEL_VECTOR_SIZE:-1024}
     # openapi_url: /openapi.json
 
@@ -42,4 +42,3 @@ settings:
   # swagger_redoc_url: /redoc
 
   # monitoring_sentry_enabled: True
-  # monitoring_prometheus_enabled: True

opengaterag/api/Dockerfile

@@ -13,15 +13,15 @@ RUN apt-get update && apt-get install -y \
     python3-dev \
     && rm -rf /var/lib/apt/lists/*
 
-WORKDIR /
+WORKDIR /app
 
-COPY ./opengaterag/api/ opengaterag/api/api
+COPY ./pyproject.toml /app/pyproject.toml
+COPY ./opengaterag /app/opengaterag
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv venv
+    uv venv /.venv
 RUN --mount=type=cache,target=/root/.cache/uv \
-    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
-    uv pip install .
+    uv pip install --python /.venv/bin/python .
 
 # Runtime stage
 FROM python:3.12-slim AS runtime
@@ -38,15 +38,13 @@ RUN apt-get update && apt-get install -y \
 
 # Set a non-root user
 USER opengaterag
-WORKDIR /
+WORKDIR /app
 
 # Copy application from builder
-COPY --from=builder --chown=opengaterag:opengaterag opengaterag/api /api
 COPY --from=builder --chown=opengaterag:opengaterag /.venv /.venv
 COPY --chown=opengaterag:opengaterag ./scripts/ /scripts/
 
 ENV PATH="/.venv/bin:${PATH}"
-ENV PYTHONPATH="/api"
 ENV PROMETHEUS_MULTIPROC_DIR="/tmp/prometheus_multiproc"
 ENV CONFIG_FILE="/config.yml"
 

opengaterag/api/utils/configuration.py

@@ -4,9 +4,9 @@
 import os
 from pathlib import Path
 import re
-from typing import Any, Literal, get_args, get_origin
+from typing import Annotated, Any, Literal, get_args, get_origin
 
-from pydantic import BaseModel, ConfigDict, Field, constr, field_validator, model_validator
+from pydantic import BaseModel, ConfigDict, Field, StringConstraints, field_validator, model_validator
 from pydantic import ValidationError as PydanticValidationError
 from pydantic_settings import BaseSettings
 import yaml
@@ -104,11 +104,11 @@ class ElasticsearchDependency(ConfigBaseModel):
     Other arguments declared below are used to configure the Elasticsearch index.
     """
 
-    index_name: constr(strip_whitespace=True, min_length=1) = Field(default="opengaterag", description="Name of the Elasticsearch index.", examples=["my_index"])  # fmt: off
-    index_language: ElasticsearchIndexLanguage = Field(default=ElasticsearchIndexLanguage.ENGLISH, description="Language of the Elasticsearch index.", examples=[ElasticsearchIndexLanguage.ENGLISH.value])  # fmt: off
-    number_of_shards: int = Field(default=12, ge=1, le=75, description="Number of shards for the Elasticsearch index.", examples=[4])  # fmt: off
-    number_of_replicas: int = Field(default=1, ge=0, description="Number of replicas for the Elasticsearch index.", examples=[1])  # fmt: off
-    refresh_interval: constr(strip_whitespace=True, pattern=r"^(-1|\d+(ms|s|m|h|d))$") = Field(default="1s", description="Refresh interval for the Elasticsearch index", examples=["2s"])  # fmt: off
+    index_name: Annotated[str, StringConstraints(strip_whitespace=True, min_length=1)] = Field(default="opengaterag", description="Name of the Elasticsearch index.", examples=["my_index"])  # fmt: off
+    index_language: Annotated[ElasticsearchIndexLanguage, Field(default=ElasticsearchIndexLanguage.ENGLISH, description="Language of the Elasticsearch index.", examples=[ElasticsearchIndexLanguage.ENGLISH.value])]  # fmt: off
+    number_of_shards: Annotated[int, Field(default=12, ge=1, le=75, description="Number of shards for the Elasticsearch index.", examples=[4])]
+    number_of_replicas: Annotated[int, Field(default=1, ge=0, description="Number of replicas for the Elasticsearch index.", examples=[1])]
+    refresh_interval: Annotated[str, StringConstraints(strip_whitespace=True, pattern=r"^(-1|\d+(ms|s|m|h|d))$")] = Field(default="1s", description="Refresh interval for the Elasticsearch index", examples=["2s"])  # fmt: off
 
 
 @custom_validation_error()
@@ -117,10 +117,17 @@ class OpengateLLMDependency(ConfigBaseModel):
     OpengateLLM is a required dependency of OpenGateLLM. It is used to connect to the OpengateLLM API.
     """
 
-    url: constr(strip_whitespace=True, min_length=1) = Field(..., description="URL of the OpengateLLM API.", examples=["https://opengatellm.com"])  # fmt: off
-    model_name: str = Field(..., description="Model of the vector store to be used to embed the text in the vector store.", examples=["text-embedding-3-small"])  # fmt: off
-    model_vector_size: int = Field(ge=1, description="Size of the vector to be used to embed the text in the vector store.", examples=[1536])  # fmt: off
-    openapi_url: str = Field(default="/openapi.json", description="Path, relative to `url`, where OpengateLLM exposes its OpenAPI schema. Used to build the unified API documentation.", examples=["/openapi.json"])  # fmt: off
+    url: Annotated[str, StringConstraints(strip_whitespace=True, min_length=1)] = Field(..., description="URL of the OpengateLLM API.", examples=["https://opengatellm.com"])  # fmt: off
+    model_name: Annotated[str, StringConstraints(strip_whitespace=True, min_length=1)] = Field(..., description="Model of the vector store to be used to embed the text in the vector store.", examples=["text-embedding-3-small"])  # fmt: off
+    model_vector_size: Annotated[int, Field(ge=1, description="Size of the vector to be used to embed the text in the vector store.", examples=[1536])]  # fmt: off
+    public_url: Annotated[str, StringConstraints(strip_whitespace=True, min_length=1)] | None = Field(default=None, description="Public URL of the OpengateLLM API to run swagger UI without CORS issues. If not provided, the public URL will be the same as the `url`.")  # fmt: off
+    openapi_url: Annotated[str, Field(default="/openapi.json", description="Path, relative to `url`, where OpengateLLM exposes its OpenAPI schema. Used to build the unified API documentation.", examples=["/openapi.json"])]  # fmt: off
+
+    @field_validator("public_url", mode="after")
+    def set_public_url(cls, public_url: str | None) -> str:
+        if public_url is None:
+            return cls.url
+        return public_url
 
 
 @custom_validation_error()
@@ -130,15 +137,7 @@ class PostgresDependency(ConfigBaseModel):
     Only the `url` argument is required. The connection URL must use the asynchronous scheme, `postgresql+asyncpg://`. If you provide a standard `postgresql://` URL, it will be automatically converted to use asyncpg.
     """
 
-    url: constr(strip_whitespace=True, min_length=1) = Field(..., pattern=r"^postgresql", description="PostgreSQL connection url.", examples=["postgresql+asyncpg://postgres:changeme@localhost:5432/postgres"])  # fmt: off
-
-    @field_validator("url", mode="after")
-    def force_async(cls, url):
-        if url.startswith("postgresql://"):
-            logging.warning(msg="PostgreSQL connection must be async, force asyncpg connection.")
-            url = url.replace("postgresql://", "postgresql+asyncpg://")
-
-        return url
+    url: Annotated[str, StringConstraints(strip_whitespace=True, min_length=1, pattern=r"^postgresql\+asyncpg://")] = Field(..., description="PostgreSQL connection url.", examples=["postgresql+asyncpg://postgres:changeme@localhost:5432/postgres"])  # fmt: off
 
 
 @custom_validation_error()

opengaterag/api/utils/sql.py

@@ -8,15 +8,12 @@
 from sqlalchemy.orm import Mapped, declarative_base, mapped_column, relationship
 from sqlalchemy.types import JSON
 
+from opengaterag.api.schemas.collections import CollectionVisibility
+
 DEFAULT_TIMEOUT = 300
 COUNTRY_CODES = [country.alpha_3 for country in pycountry.countries] + ["WOR"]
 
 
-class CollectionVisibility(StrEnum):
-    PRIVATE = "private"
-    PUBLIC = "public"
-
-
 class LimitType(StrEnum):
     TPM = "tpm"
     TPD = "tpd"

opengaterag/docs/app.py

@@ -23,9 +23,10 @@ def create_docs_app(parent_app: FastAPI, configuration: Configuration) -> FastAP
     OpenGateRAG schema is served so the documentation never fully breaks.
     """
     settings = configuration.settings
-    opengatellm_url = configuration.dependencies.opengatellm.url.rstrip("/")
+    opengatellm_private_url = configuration.dependencies.opengatellm.url.rstrip("/")
+    opengatellm_public_url = (configuration.dependencies.opengatellm.public_url or opengatellm_private_url).rstrip("/")
     opengatellm_openapi_path = configuration.dependencies.opengatellm.openapi_url.lstrip("/")
-    opengatellm_openapi_url = f"{opengatellm_url}/{opengatellm_openapi_path}"
+    opengatellm_openapi_url = f"{opengatellm_private_url}/{opengatellm_openapi_path}"
 
     docs_app = FastAPI(docs_url=None, redoc_url=None, openapi_url=None)
     cache: dict[str, object] = {"schema": None, "expires_at": 0.0}
@@ -41,7 +42,7 @@ async def build_merged_schema() -> dict:
             merged = merge_openapi_schemas(
                 base=base_schema,
                 secondary=secondary_schema,
-                secondary_server_url=opengatellm_url,
+                secondary_server_url=opengatellm_public_url,
             )
         except httpx.HTTPError as error:
             logger.warning(f"Unable to fetch OpenGateLLM OpenAPI schema from {opengatellm_openapi_url}: {error}")

pyproject.toml

@@ -35,9 +35,17 @@ test = [
     "pytest-cov>=7.0.0",
 ]
 
+[build-system]
+requires = ["setuptools>=70", "wheel"]
+build-backend = "setuptools.build_meta"
+
 [tool.setuptools]
 py-modules = []
 
+[tool.setuptools.packages.find]
+where = ["."]
+include = ["opengaterag*"]
+
 [tool.ruff]
 line-length = 150
 target-version = "py312"

scripts/startup_api.sh

@@ -3,8 +3,19 @@ set -e
 
 GUNICORN_CMD_ARGS=${GUNICORN_CMD_ARGS:-""} # ex: --log-config app/log.conf
 
+# Allow running locally without env vars.
+PROMETHEUS_MULTIPROC_DIR=${PROMETHEUS_MULTIPROC_DIR:-/tmp/prometheus_multiproc}
 mkdir -p "$PROMETHEUS_MULTIPROC_DIR"
 # Empty prometheus dir if it exists, to prevent ghost metrics being ingested twice in case of docker restart
 rm -rf "${PROMETHEUS_MULTIPROC_DIR:?}"/*
 
-exec gunicorn opengaterag.api.main:app --worker-class uvicorn.workers.UvicornWorker --bind 0.0.0.0:8000 --config scripts/gunicorn.conf.py $GUNICORN_CMD_ARGS
+if [[ -f "scripts/gunicorn.conf.py" ]]; then
+  GUNICORN_CONFIG="scripts/gunicorn.conf.py"
+elif [[ -f "/scripts/gunicorn.conf.py" ]]; then
+  GUNICORN_CONFIG="/scripts/gunicorn.conf.py"
+else
+  echo "Error: gunicorn config not found (expected scripts/gunicorn.conf.py or /scripts/gunicorn.conf.py)" >&2
+  exit 1
+fi
+
+exec gunicorn opengaterag.api.main:app --worker-class uvicorn.workers.UvicornWorker --bind 0.0.0.0:8000 --config "$GUNICORN_CONFIG" $GUNICORN_CMD_ARGS