diff --git a/.env.example b/.env.example
index 20c8f98..61f1e7d 100644
--- a/.env.example
+++ b/.env.example
@@ -1,3 +1,9 @@
 # OpenAI-compatible API configuration
 LLM_BASE_URL=http://localhost:11434/v1
 LLM_MODEL=llama
+
+# API key for the LLM provider (leave empty for local servers like Ollama)
+LLM_API_KEY=
+
+# Logfire observability (set your token to enable)
+LOGFIRE_TOKEN=
diff --git a/Makefile b/Makefile
index 572bf09..d91de4f 100644
--- a/Makefile
+++ b/Makefile
@@ -1,5 +1,11 @@
 .PHONY: install run dev clean setup \
-        frontend-install frontend-dev frontend-build
+        frontend-install frontend-dev frontend-build \
+        compose-up compose-down compose-build compose-logs compose-rebuild
+
+# ── Container runtime detection (Docker / Podman) ─────────────────────────
+# Uses Docker if available, otherwise falls back to Podman.
+DOCKER := $(shell command -v docker 2>/dev/null || command -v podman 2>/dev/null)
+DOCKER_COMPOSE := $(shell command -v docker-compose 2>/dev/null || (command -v docker 2>/dev/null && echo "docker compose") || (command -v podman 2>/dev/null && echo "podman compose"))
 
 # ── Backend Setup ──────────────────────────────────────────────────────────
 
@@ -32,6 +38,23 @@ frontend-dev:                      ## Start the Vite dev server
 frontend-build:                    ## Build frontend for production
 	cd frontend && bun run build
 
+# ── Containers (Docker / Podman) ───────────────────────────────────────────
+
+compose-up:                        ## Start all containers in detached mode
+	$(DOCKER_COMPOSE) up -d
+
+compose-down:                      ## Stop and remove containers
+	$(DOCKER_COMPOSE) down
+
+compose-build:                     ## Build (or rebuild) all container images
+	$(DOCKER_COMPOSE) build
+
+compose-rebuild:                   ## Rebuild images & restart containers
+	$(DOCKER_COMPOSE) up -d --build
+
+compose-logs:                      ## Follow container logs
+	$(DOCKER_COMPOSE) logs -f
+
 # ── Utilities ──────────────────────────────────────────────────────────────
 
 clean:                             ## Remove caches and build artifacts
diff --git a/backend/core/agent.py b/backend/core/agent.py
index 43ebc0b..5d076a8 100644
--- a/backend/core/agent.py
+++ b/backend/core/agent.py
@@ -32,6 +32,7 @@ class AgentService:
     def __init__(self, settings: Settings) -> None:
         self._settings = settings
         self._agent: Agent | None = None
+        self._model: OpenAIResponsesModel | None = None
 
     # ── Lifecycle ──────────────────────────────────────────────────────────
 
@@ -40,16 +41,23 @@ class AgentService:
         if self._agent is not None:
             return
 
-        logfire.configure()
-        logfire.instrument_pydantic_ai()
+        if self._settings.logfire_token:
+            try:
+                logfire.configure(token=self._settings.logfire_token)
+                logfire.instrument_pydantic_ai()
+            except Exception:
+                pass  # Logfire is optional — don't crash if it fails
 
-        model = OpenAIResponsesModel(
+        self._model = OpenAIResponsesModel(
             self._settings.llm_model,
-            provider=OpenAIProvider(base_url=self._settings.llm_base_url),
+            provider=OpenAIProvider(
+                base_url=self._settings.llm_base_url,
+                api_key=self._settings.llm_api_key or "no-key-required",
+            ),
         )
 
         self._agent = Agent(
-            model,
+            self._model,
             capabilities=self._build_capabilities(),
         )
 
@@ -69,8 +77,7 @@ class AgentService:
 
     # ── Internal helpers ───────────────────────────────────────────────────
 
-    @staticmethod
-    def _build_capabilities() -> list:
+    def _build_capabilities(self) -> list:
         """Assemble the full list of agent capabilities."""
         return [
             CodeMode(),
@@ -89,7 +96,8 @@ class AgentService:
                         description="Deep research on a topic",
                         instructions="You are a thorough research assistant.",
                     ),
-                ]
+                ],
+                default_model=self._model,
             ),
             TodoCapability(enable_subtasks=True),
             CostTracking(budget_usd=5.0),
diff --git a/backend/core/config.py b/backend/core/config.py
index f09f0b6..60647b8 100644
--- a/backend/core/config.py
+++ b/backend/core/config.py
@@ -22,6 +22,8 @@ class Settings(BaseSettings):
 
     llm_base_url: str = "http://localhost:11434/v1"
     llm_model: str = "llama"
+    llm_api_key: str = ""
+    logfire_token: str = ""
 
     # ── Factories ──────────────────────────────────────────────────────────
 
diff --git a/frontend/nginx.conf b/frontend/nginx.conf
index 443b4c5..6360c67 100644
--- a/frontend/nginx.conf
+++ b/frontend/nginx.conf
@@ -5,14 +5,22 @@ server {
     root /usr/share/nginx/html;
     index index.html;
 
+    # Podman/Docker DNS — forces runtime resolution so nginx picks up
+    # container IP changes after restarts.
+    resolver 10.89.6.1 valid=10s;
+
     # ── API proxy ─────────────────────────────────────────────────
+    # Use a variable so nginx resolves DNS at runtime, not at startup.
+    set $backend_upstream http://backend:8000;
     location /api/ {
-        proxy_pass http://backend:8000;
+        proxy_pass $backend_upstream;
         proxy_set_header Host $host;
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
         proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_read_timeout 120s;
+        proxy_connect_timeout 30s;
+        proxy_read_timeout 200s;
+        proxy_send_timeout 200s;
     }
 
     # ── Static files ──────────────────────────────────────────────