Fix bugs

.dockerignore

@@ -0,0 +1,10 @@
+.git
+.venv
+__pycache__
+*.pyc
+.pytest_cache
+.ruff_cache
+.mypy_cache
+.DS_Store
+logs
+notebooks

.gitignore

@@ -111,7 +111,9 @@ venv.bak
 data/sample/
 data/output/
 logs/
+
+# Deployment
 temp/
 
 # W&B files
-wandb/
+wandb/

Caddyfile

@@ -0,0 +1,9 @@
+{$SERVER_NAME} {
+    encode zstd gzip
+
+    request_body {
+        max_size 50MB
+    }
+
+    reverse_proxy metadata-demo:8501
+}

Dockerfile

@@ -0,0 +1,19 @@
+FROM python:3.11-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PIP_NO_CACHE_DIR=1
+
+WORKDIR /app
+
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends make \
+    && rm -rf /var/lib/apt/lists/* \
+    && pip install --no-cache-dir uv
+
+COPY . .
+RUN make docker-install-demo
+
+EXPOSE 8501
+
+CMD ["make", "docker-run-demo"]

demo/workflows/metadata_generation.py

@@ -6,9 +6,15 @@
 from pathlib import Path
 from typing import Any, Callable
 
+import numpy as np
 import pandas as pd
 
-from src.config import LLM_PROVIDER, PLANNING_TEMPERATURE, get_model_name
+from src.config import (
+    DEFAULT_TOPOLOGY,
+    LLM_PROVIDER,
+    PLANNING_TEMPERATURE,
+    get_model_name,
+)
 from src.context import create_context
 from src.orchestrator import Orchestrator
 from src.standards import METADATA_STANDARDS, load_metadata_standard
@@ -69,7 +75,7 @@ def generate_metadata(
     file_name: str,
     file_bytes: bytes,
     standard_name: str,
-    topology_name: str = "default",
+    topology_name: str = DEFAULT_TOPOLOGY,
     progress_callback: Callable[[str, Any | None], None] | None = None,
 ) -> dict[str, Any]:
     """Run metadata extraction for an uploaded file and return JSON-friendly output.
@@ -103,7 +109,7 @@ def generate_metadata(
     try:
         dataset_name = Path(file_name).stem
         metadata_standard = load_metadata_standard(standard_name)
-        context = create_context(temp_path, name=dataset_name)
+        context = create_context({"data": temp_path}, name=dataset_name)
         _publish_progress(progress_callback, "context_created", context.to_dict())
 
         orchestrator = Orchestrator(
@@ -213,7 +219,17 @@ def _to_displayable(value: Any) -> Any:
         A JSON-friendly representation of the value.
     """
     if hasattr(value, "model_dump"):
-        return value.model_dump(mode="json")
-    if isinstance(value, (dict, list, str, int, float, bool)) or value is None:
+        return _to_displayable(value.model_dump(mode="python"))
+    if isinstance(value, dict):
+        return {str(key): _to_displayable(item) for key, item in value.items()}
+    if isinstance(value, (list, tuple, set)):
+        return [_to_displayable(item) for item in value]
+    if isinstance(value, np.ndarray):
+        return _to_displayable(value.tolist())
+    if isinstance(value, np.generic):
+        return _to_displayable(value.item())
+    if isinstance(value, pd.Timestamp):
+        return value.isoformat()
+    if isinstance(value, (str, int, float, bool)) or value is None:
         return value
     return json.loads(json.dumps(value, default=str))

deploy.sh

@@ -0,0 +1,3 @@
+git clone --single-branch --branch deploy https://github.com/naliATsurf/metadata_agent.git
+cd metadata_agent
+sudo docker compose up --build

docker-compose.yml

@@ -0,0 +1,27 @@
+services:
+  metadata-demo:
+    build: .
+    restart: unless-stopped
+    env_file:
+      - .env
+    expose:
+      - "8501"
+
+  caddy:
+    image: caddy:2-alpine
+    restart: unless-stopped
+    depends_on:
+      - metadata-demo
+    environment:
+      SERVER_NAME: ${SERVER_NAME:?Set SERVER_NAME in .env to your public domain}
+    ports:
+      - "80:80"
+      - "443:443"
+    volumes:
+      - ./Caddyfile:/etc/caddy/Caddyfile:ro
+      - caddy_data:/data
+      - caddy_config:/config
+
+volumes:
+  caddy_data:
+  caddy_config:

docs/demo.md

@@ -1,16 +1,163 @@
 # Streamlit Demo
+
 ## Run
+
 ```bash
 uv run --group demo streamlit run demo_app.py
 ```
-or 
+
+or
+
 ```bash
 make demo
 ```
 
+## Deploy as a web server with Docker Compose
+
+This repository includes a Docker Compose setup for running the Streamlit demo
+behind Caddy with automatic HTTPS:
+
+- `metadata-demo`: builds the app image from `Dockerfile` and runs Streamlit on
+  port `8501` inside the Compose network.
+- `caddy`: exposes ports `80` and `443`, gets and renews TLS certificates, and
+  proxies browser traffic to the Streamlit container.
+
+### 1. Prepare the server
+
+Install Docker Engine and the Compose plugin on the server, then clone the
+repository:
+
+```bash
+git clone <repository-url>
+cd metadata_agent
+```
+
+The server must allow inbound traffic on ports `80` and `443`. Caddy uses port
+`80` for the HTTP-to-HTTPS redirect and Let's Encrypt validation, then serves
+the app over HTTPS on port `443`.
+
+### 2. Create `.env`
+
+Docker Compose loads `.env` for both Compose variables and application
+configuration. Create it in the repository root:
+
+```bash
+LLM_PROVIDER=surf
+SURF_API_KEY=your_surf_api_key
+SURF_API_BASE=
+LLM_MODEL=
+DEFAULT_TOPOLOGY=single
+SERVER_NAME=metadata.example.org
+```
+
+Provider choices:
+
+```bash
+# Google
+LLM_PROVIDER=google
+GOOGLE_API_KEY=your_google_api_key
+
+# OpenAI
+LLM_PROVIDER=openai
+OPENAI_API_KEY=your_openai_api_key
+
+# SURF/OpenAI-compatible endpoint
+LLM_PROVIDER=surf
+SURF_API_KEY=your_surf_api_key
+SURF_API_BASE=
+```
+
+`DEFAULT_TOPOLOGY=single` is a good web-demo default because it avoids the extra
+multi-player debate calls used by heavier topologies.
+
+Set `SERVER_NAME` to the public domain that resolves to your server. Caddy needs
+a real hostname for public Let's Encrypt certificates.
+
+### 3. Build and start the service
+
+```bash
+docker compose up -d --build
+```
+
+Open the app:
+
+```text
+https://metadata.example.org/
+```
+
+For local testing on the server:
+
+```bash
+curl -I https://metadata.example.org/
+```
+
+### 4. Watch logs
+
+Use the repository helper:
+
+```bash
+make docker-logs
+```
+
+or plain Compose:
+
+```bash
+docker compose logs -f metadata-demo
+docker compose logs -f caddy
+```
+
+The Streamlit container should show `streamlit run demo_app.py`. Caddy should
+show certificate provisioning logs the first time it starts, then proxy requests
+to `metadata-demo:8501`.
+
+### 5. Update the deployment
+
+Pull the latest code, rebuild the image, and restart:
+
+```bash
+git pull
+docker compose up -d --build
+```
+
+To stop the service:
+
+```bash
+docker compose down
+```
+
+The Caddy certificate cache is stored in named Docker volumes. Use plain
+`docker compose down` for normal restarts so certificates are preserved.
+
+### 6. Troubleshooting
+
+If the page does not load, check:
+
+```bash
+docker compose ps
+docker compose logs metadata-demo
+docker compose logs caddy
+```
+
+Common causes:
+
+- Missing API key in `.env`: the app starts, but metadata generation fails when
+  the selected provider is called.
+- Port `80` or `443` already in use: stop the existing service or change the
+  published ports in `docker-compose.yml`.
+- `SERVER_NAME` is missing or is not a public DNS name: set it in `.env` to the
+  domain that points at the server.
+- Domain not routed to the server: test with:
+  ```bash
+  curl -H 'Host: metadata.example.org' http://SERVER_IP/
+  ```
+- Long metadata generation: keep `DEFAULT_TOPOLOGY=single` for the demo, then
+  use `fast`, `default`, or `thorough` only when you explicitly want more LLM
+  calls.
+
 ## Folder roles
 
 ### `demo_app.py`
+
 Streamlit entry point that configures and launches the demo app.
 
 ### `demo/`

examples/generation.py

@@ -113,6 +113,7 @@ def log_step_timing(step_name: str, start: float) -> float:
 
 # 5. Collect and persist the final metadata output produced by the plan.
 log_step_section(5, "Write metadata")
+
 metadata_output = result.final_workspace['metadata_output']
 output_dir = Path(example_config.OUTPUT_DIR)
 output_dir.mkdir(parents=True, exist_ok=True)

makefile

@@ -6,7 +6,7 @@ PYTHON-VERSION := 3.11
 PYTHON := $(VENV)/bin/python
 RUFF := $(VENV)/bin/ruff
 
-.PHONY: help uv-setup uv-clean activate install-docs docs docs-update docs-clean lint compile test ci
+.PHONY: help uv-setup uv-clean activate install-docs docs docs-update docs-clean install-demo demo docker-install-demo docker-run-demo docker-build docker-up docker-down docker-logs lint compile test ci
 
 help:
 	@printf '%s\n' \
@@ -18,6 +18,13 @@ help:
 		'  make docs       - build the documentation and open it in the browser' \
 		'  make docs-update - update the documentation' \
 		'  make docs-clean - clean the generated documentation files' \
+		'  make demo       - install demo dependencies and launch Streamlit locally' \
+		'  make docker-install-demo - install locked demo dependencies for Docker' \
+		'  make docker-run-demo - launch Streamlit for Docker' \
+		'  make docker-build - build the Docker image' \
+		'  make docker-up   - build and start the Docker Compose service' \
+		'  make docker-down - stop the Docker Compose service' \
+		'  make docker-logs - follow Docker Compose service logs' \
 		'  make lint       - run ruff linter on the codebase' \
 		'  make compile    - compile the source code to check for syntax errors' \
 		'  make test       - run unit tests' \
@@ -64,6 +71,24 @@ install-demo:
 demo: install-demo
 	$(UV) run --group demo streamlit run demo_app.py
 
+docker-install-demo:
+	$(UV) sync --locked --no-default-groups --group demo
+
+docker-run-demo:
+	$(UV) run --no-sync --group demo streamlit run demo_app.py --server.address=0.0.0.0 --server.port=8501
+
+docker-build:
+	docker compose build
+
+docker-up:
+	docker compose up -d --build
+
+docker-down:
+	docker compose down
+
+docker-logs:
+	docker compose logs -f metadata-demo caddy
+
 lint:
 	$(UV) run ruff check .
 
@@ -77,3 +102,7 @@ ci-install:
 	$(UV) sync --locked --no-default-groups
 
 ci: lint compile test
+
+tui:
+	$(UV) pip install -e .
+	metadata-agent --tui