llmstack-cli 0.1.0__py3-none-any.whl → 0.2.0__py3-none-any.whl

llmstack/__init__.py

@@ -1,3 +1,3 @@
 """LLMStack — One command. Full LLM stack. Zero config."""
 
-__version__ = "0.1.0"
+__version__ = "0.2.0"

llmstack/cli/app.py

@@ -80,6 +80,24 @@ def logs(
     _logs(service=service, follow=follow, tail=tail)
 
 
+@app.command()
+def chat(
+    model: str = typer.Option(None, "--model", "-m", help="Model name to chat with"),
+) -> None:
+    """Interactive chat with your local LLM."""
+    from llmstack.cli.commands.chat import chat as _chat
+    _chat(model=model)
+
+
+@app.command()
+def export(
+    output: str = typer.Option("docker-compose.yml", "--output", "-o", help="Output file path"),
+) -> None:
+    """Export llmstack.yaml as a standalone docker-compose.yml."""
+    from llmstack.cli.commands.export import export as _export
+    _export(output=output)
+
+
 @app.command()
 def doctor() -> None:
     """Check system requirements and diagnose issues."""

llmstack/cli/commands/chat.py

@@ -0,0 +1,104 @@
+"""llmstack chat — interactive terminal chat with your local LLM."""
+
+from __future__ import annotations
+
+import json
+import sys
+
+import httpx
+
+from llmstack.cli.console import console
+from llmstack.config.loader import load_config
+
+
+def chat(model: str | None = None) -> None:
+    """Start an interactive chat session with the running LLM."""
+    config = load_config()
+    base_url = f"http://localhost:{config.gateway.port}"
+    model_name = model or config.models.chat.name
+
+    # Get API key
+    api_key = config.gateway.api_keys[0] if config.gateway.api_keys else ""
+    headers = {"Content-Type": "application/json"}
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
+
+    # Verify gateway is reachable
+    try:
+        resp = httpx.get(f"{base_url}/healthz", headers=headers, timeout=5)
+        if resp.status_code != 200:
+            console.print("[error]Gateway is not healthy. Run 'llmstack up' first.[/]")
+            sys.exit(1)
+    except httpx.ConnectError:
+        console.print("[error]Cannot connect to gateway. Run 'llmstack up' first.[/]")
+        sys.exit(1)
+
+    console.print(f"\n[bold]LLMStack Chat[/] — model: [cyan]{model_name}[/]")
+    console.print("[dim]Type 'exit' or Ctrl+C to quit. '/clear' to reset conversation.[/]\n")
+
+    messages: list[dict[str, str]] = []
+
+    while True:
+        try:
+            user_input = console.input("[bold green]You:[/] ")
+        except (KeyboardInterrupt, EOFError):
+            console.print("\n[dim]Goodbye![/]")
+            break
+
+        text = user_input.strip()
+        if not text:
+            continue
+        if text.lower() in ("exit", "quit"):
+            console.print("[dim]Goodbye![/]")
+            break
+        if text == "/clear":
+            messages.clear()
+            console.print("[dim]Conversation cleared.[/]\n")
+            continue
+
+        messages.append({"role": "user", "content": text})
+
+        # Stream response
+        console.print("[bold cyan]Assistant:[/] ", end="")
+        assistant_text = ""
+
+        try:
+            with httpx.stream(
+                "POST",
+                f"{base_url}/v1/chat/completions",
+                headers=headers,
+                json={"model": model_name, "messages": messages, "stream": True},
+                timeout=httpx.Timeout(300, connect=10),
+            ) as resp:
+                if resp.status_code != 200:
+                    console.print(f"[error]Error {resp.status_code}[/]")
+                    messages.pop()
+                    continue
+
+                for line in resp.iter_lines():
+                    if not line.startswith("data: "):
+                        continue
+                    data = line[6:]
+                    if data.strip() == "[DONE]":
+                        break
+                    try:
+                        chunk = json.loads(data)
+                        delta = chunk["choices"][0].get("delta", {})
+                        token = delta.get("content", "")
+                        if token:
+                            print(token, end="", flush=True)
+                            assistant_text += token
+                    except (json.JSONDecodeError, KeyError, IndexError):
+                        continue
+
+        except httpx.ConnectError:
+            console.print("[error]Connection lost.[/]")
+            messages.pop()
+            continue
+        except KeyboardInterrupt:
+            pass
+
+        print()  # newline after streamed response
+        if assistant_text:
+            messages.append({"role": "assistant", "content": assistant_text})
+        print()

llmstack/cli/commands/export.py

@@ -0,0 +1,190 @@
+"""llmstack export — generate a standalone docker-compose.yml from llmstack.yaml."""
+
+from __future__ import annotations
+
+import yaml
+
+from llmstack.cli.console import console
+from llmstack.config.loader import load_config
+from llmstack.config.schema import StackConfig
+from llmstack.core.hardware import detect_hardware
+from llmstack.core.resolver import resolve_inference_backend, resolve_embedding_backend
+
+
+def _build_compose(config: StackConfig) -> dict:
+    """Convert StackConfig into a docker-compose.yml dict."""
+    hw = detect_hardware()
+    inference_backend = resolve_inference_backend(config.models.chat, hw)
+    embed_backend = resolve_embedding_backend(config.models.embeddings, hw)
+
+    services: dict = {}
+    volumes: dict = {}
+    network_name = config.docker.network
+
+    # 1. Qdrant
+    services["qdrant"] = {
+        "image": "qdrant/qdrant:latest",
+        "container_name": "llmstack-qdrant",
+        "ports": [f"{config.services.vectors.port}:6333", f"{config.services.vectors.port + 1}:6334"],
+        "volumes": ["qdrant_data:/qdrant/storage"],
+        "restart": "unless-stopped",
+        "networks": [network_name],
+    }
+    volumes["qdrant_data"] = None
+
+    # 2. Redis
+    services["redis"] = {
+        "image": "redis:7-alpine",
+        "container_name": "llmstack-redis",
+        "ports": [f"{config.services.cache.port}:6379"],
+        "command": f"redis-server --maxmemory {config.services.cache.max_memory} --maxmemory-policy allkeys-lru",
+        "restart": "unless-stopped",
+        "networks": [network_name],
+    }
+
+    # 3. Inference
+    if inference_backend == "vllm":
+        model = config.models.chat
+        cmd = f"--model {model.name} --host 0.0.0.0 --port 8000 --max-model-len {model.context_length}"
+        if model.quantization:
+            cmd += f" --quantization {model.quantization}"
+        services["vllm"] = {
+            "image": "vllm/vllm-openai:latest",
+            "container_name": "llmstack-vllm",
+            "ports": ["8001:8000"],
+            "command": cmd,
+            "volumes": ["vllm_cache:/root/.cache/huggingface"],
+            "shm_size": "4g",
+            "deploy": {"resources": {"reservations": {"devices": [{"driver": "nvidia", "count": "all", "capabilities": ["gpu"]}]}}},
+            "restart": "unless-stopped",
+            "networks": [network_name],
+        }
+        volumes["vllm_cache"] = None
+        inference_url = "http://vllm:8000/v1"
+    else:
+        svc: dict = {
+            "image": "ollama/ollama:latest",
+            "container_name": "llmstack-ollama",
+            "ports": ["11434:11434"],
+            "volumes": ["ollama_data:/root/.ollama"],
+            "restart": "unless-stopped",
+            "networks": [network_name],
+        }
+        if hw.gpu_vendor == "nvidia":
+            svc["deploy"] = {"resources": {"reservations": {"devices": [{"driver": "nvidia", "count": "all", "capabilities": ["gpu"]}]}}}
+        services["ollama"] = svc
+        volumes["ollama_data"] = None
+        inference_url = "http://ollama:11434/v1"
+
+    # 4. Embeddings
+    embeddings_url = inference_url  # fallback: use Ollama for embeddings
+    if embed_backend == "tei":
+        tei_image = "ghcr.io/huggingface/text-embeddings-inference:cpu-latest"
+        tei_svc: dict = {
+            "image": tei_image,
+            "container_name": "llmstack-tei",
+            "ports": ["8002:80"],
+            "command": f"--model-id {config.models.embeddings.name} --port 80",
+            "volumes": ["tei_cache:/data"],
+            "restart": "unless-stopped",
+            "networks": [network_name],
+        }
+        if hw.gpu_vendor == "nvidia":
+            tei_svc["image"] = "ghcr.io/huggingface/text-embeddings-inference:latest"
+            tei_svc["deploy"] = {"resources": {"reservations": {"devices": [{"driver": "nvidia", "count": "all", "capabilities": ["gpu"]}]}}}
+        services["tei"] = tei_svc
+        volumes["tei_cache"] = None
+        embeddings_url = "http://tei:80/v1"
+
+    # 5. Gateway
+    qdrant_url = f"http://qdrant:{config.services.vectors.port}"
+    redis_url = f"redis://redis:{config.services.cache.port}"
+    api_keys = ",".join(config.gateway.api_keys) if config.gateway.api_keys else ""
+
+    services["gateway"] = {
+        "build": {
+            "context": ".",
+            "dockerfile": "src/llmstack/gateway/Dockerfile",
+        },
+        "container_name": "llmstack-gateway",
+        "ports": [f"{config.gateway.port}:8000"],
+        "environment": {
+            "LLMSTACK_INFERENCE_URL": inference_url,
+            "LLMSTACK_EMBEDDINGS_URL": embeddings_url,
+            "LLMSTACK_QDRANT_URL": qdrant_url,
+            "LLMSTACK_REDIS_URL": redis_url,
+            "LLMSTACK_API_KEYS": api_keys,
+            "LLMSTACK_CORS_ORIGINS": ",".join(config.gateway.cors),
+            "LLMSTACK_REQUEST_TIMEOUT": str(config.gateway.request_timeout),
+            "LLMSTACK_RATE_LIMIT": config.gateway.rate_limit,
+        },
+        "depends_on": ["qdrant", "redis"],
+        "restart": "unless-stopped",
+        "networks": [network_name],
+    }
+    # Add inference dependency
+    if inference_backend == "vllm":
+        services["gateway"]["depends_on"].append("vllm")
+    else:
+        services["gateway"]["depends_on"].append("ollama")
+    if embed_backend == "tei":
+        services["gateway"]["depends_on"].append("tei")
+
+    # 6. Observability
+    if config.observe.metrics:
+        services["prometheus"] = {
+            "image": "prom/prometheus:latest",
+            "container_name": "llmstack-prometheus",
+            "ports": ["9090:9090"],
+            "command": [
+                "--config.file=/etc/prometheus/prometheus.yml",
+                f"--storage.tsdb.retention.time={config.observe.retention}",
+                "--web.enable-lifecycle",
+            ],
+            "volumes": [
+                "./prometheus.yml:/etc/prometheus/prometheus.yml:ro",
+                "prometheus_data:/prometheus",
+            ],
+            "restart": "unless-stopped",
+            "networks": [network_name],
+        }
+        volumes["prometheus_data"] = None
+
+        services["grafana"] = {
+            "image": "grafana/grafana:latest",
+            "container_name": "llmstack-grafana",
+            "ports": [f"{config.observe.dashboard_port}:3000"],
+            "environment": {
+                "GF_SECURITY_ADMIN_USER": "admin",
+                "GF_SECURITY_ADMIN_PASSWORD": "llmstack",
+                "GF_AUTH_ANONYMOUS_ENABLED": "true",
+                "GF_AUTH_ANONYMOUS_ORG_ROLE": "Viewer",
+            },
+            "restart": "unless-stopped",
+            "networks": [network_name],
+        }
+
+    compose = {
+        "version": "3.8",
+        "services": services,
+        "volumes": volumes,
+        "networks": {network_name: {"driver": "bridge"}},
+    }
+
+    return compose
+
+
+def export(output: str = "docker-compose.yml") -> None:
+    """Generate a docker-compose.yml from the current llmstack.yaml."""
+    config = load_config()
+    compose = _build_compose(config)
+
+    # Clean up None volume values for yaml output
+    compose["volumes"] = {k: {} for k in compose["volumes"]}
+
+    with open(output, "w") as f:
+        yaml.dump(compose, f, default_flow_style=False, sort_keys=False)
+
+    service_count = len(compose["services"])
+    console.print(f"\n[success]Exported {service_count} services to {output}[/]")
+    console.print(f"Run with: [bold]docker compose -f {output} up -d[/]\n")

llmstack/core/stack.py

@@ -7,6 +7,7 @@ import secrets
 from rich.console import Console
 from rich.table import Table
 
+from llmstack.config.loader import save_config
 from llmstack.config.schema import StackConfig
 from llmstack.core.hardware import detect_hardware
 from llmstack.core.health import wait_healthy
@@ -100,11 +101,21 @@ class Stack:
         self._services = self._build_services()
         self.docker.ensure_network()
 
+        # Build local images if needed
+        for svc in self._services:
+            if hasattr(svc, "build_info") and svc.build_info():
+                info = svc.build_info()
+                console.print(f"  [cyan]Building {svc.name} image...[/]", end="")
+                self.docker.build_image(**info)
+                console.print(" [green]done[/]")
+
         # Generate API key if needed
         if self.config.gateway.auth == "api_key" and not self.config.gateway.api_keys:
             key = f"sk-llmstack-{secrets.token_urlsafe(24)}"
             self.config.gateway.api_keys = [key]
-            console.print(f"\n[bold green]Generated API key:[/] {key}\n")
+            save_config(self.config)
+            console.print(f"\n[bold green]Generated API key:[/] {key}")
+            console.print("[dim]Saved to llmstack.yaml[/]\n")
 
         for svc in self._services:
             console.print(f"  [cyan]Starting {svc.name}...[/]", end="")

llmstack/docker/manager.py

@@ -35,6 +35,11 @@ class DockerManager:
         except NotFound:
             self.client.networks.create(self.network_name, driver="bridge")
 
+    def build_image(self, path: str, dockerfile: str, tag: str) -> str:
+        """Build a Docker image from a local Dockerfile. Returns the image tag."""
+        self.client.images.build(path=path, dockerfile=dockerfile, tag=tag, rm=True)
+        return tag
+
     def run_service(self, service: ServiceBase) -> Container:
         """Start a container for a service. Removes any existing container with the same name."""
         spec = service.container_spec()
@@ -53,13 +58,24 @@ class DockerManager:
             self.LABEL_SERVICE: service.name,
         }
 
-        container = self.client.containers.run(
-            detach=True,
-            name=name,
-            network=self.network_name,
-            labels=labels,
-            **spec,
-        )
+        try:
+            container = self.client.containers.run(
+                detach=True,
+                name=name,
+                network=self.network_name,
+                labels=labels,
+                **spec,
+            )
+        except APIError as exc:
+            msg = str(exc)
+            if "port is already allocated" in msg or "address already in use" in msg:
+                ports = spec.get("ports", {})
+                port = next(iter(ports.values()), "unknown") if ports else "unknown"
+                raise SystemExit(
+                    f"Port {port} is already in use. "
+                    f"Stop the conflicting service or change the port in llmstack.yaml."
+                ) from exc
+            raise
         return container
 
     def stop_service(self, service_name: str) -> None:

llmstack/gateway/Dockerfile

@@ -8,7 +8,9 @@ RUN pip install --no-cache-dir \
     httpx>=0.27 \
     starlette>=0.40
 
-COPY . /app/llmstack/gateway/
+COPY src/llmstack/__init__.py /app/llmstack/__init__.py
+COPY src/llmstack/gateway/ /app/llmstack/gateway/
+
 ENV PYTHONPATH=/app
 
 EXPOSE 8000

llmstack/services/gateway/service.py

@@ -2,11 +2,14 @@
 
 from __future__ import annotations
 
+from pathlib import Path
 from typing import Any
 
 from llmstack.config.schema import GatewayConfig
 from llmstack.services.base import ServiceBase
 
+IMAGE_TAG = "llmstack-gateway:local"
+
 
 class GatewayService(ServiceBase):
     name = "gateway"
@@ -26,9 +29,21 @@ class GatewayService(ServiceBase):
         self.qdrant_url = qdrant_url
         self.redis_url = redis_url
 
+    def build_info(self) -> dict[str, str] | None:
+        """Return build context for the gateway Docker image."""
+        import llmstack
+        pkg_dir = Path(llmstack.__file__).resolve().parent
+        project_root = pkg_dir.parent.parent  # src/llmstack -> src -> project root
+        dockerfile = str(pkg_dir / "gateway" / "Dockerfile")
+        return {
+            "path": str(project_root),
+            "dockerfile": dockerfile,
+            "tag": IMAGE_TAG,
+        }
+
     def container_spec(self) -> dict[str, Any]:
         return {
-            "image": "ghcr.io/mara-werils/llmstack-gateway:latest",
+            "image": IMAGE_TAG,
             "name": "llmstack-gateway",
             "ports": {"8000/tcp": self.config.port},
             "environment": {

llmstack/services/observe/prometheus.py

@@ -3,31 +3,38 @@
 from __future__ import annotations
 
 import json
+import os
+from pathlib import Path
 from typing import Any
 
+import yaml
+
 from llmstack.config.schema import ObserveConfig
 from llmstack.services.base import ServiceBase
 
+CONFIG_DIR = Path.home() / ".llmstack" / "config"
 
 # Prometheus config that scrapes the gateway /metrics endpoint
-PROMETHEUS_CONFIG = """
-global:
-  scrape_interval: 15s
-  evaluation_interval: 15s
-
-scrape_configs:
-  - job_name: 'llmstack-gateway'
-    metrics_path: '/metrics'
-    static_configs:
-      - targets: ['llmstack-gateway:8000']
-    scrape_interval: 5s
-
-  - job_name: 'qdrant'
-    metrics_path: '/metrics'
-    static_configs:
-      - targets: ['llmstack-qdrant:6333']
-    scrape_interval: 15s
-"""
+PROMETHEUS_CONFIG = {
+    "global": {
+        "scrape_interval": "15s",
+        "evaluation_interval": "15s",
+    },
+    "scrape_configs": [
+        {
+            "job_name": "llmstack-gateway",
+            "metrics_path": "/metrics",
+            "static_configs": [{"targets": ["llmstack-gateway:8000"]}],
+            "scrape_interval": "5s",
+        },
+        {
+            "job_name": "qdrant",
+            "metrics_path": "/metrics",
+            "static_configs": [{"targets": ["llmstack-qdrant:6333"]}],
+            "scrape_interval": "15s",
+        },
+    ],
+}
 
 # Grafana datasource provisioning
 GRAFANA_DATASOURCE = {
@@ -47,7 +54,7 @@ GRAFANA_DASHBOARD_PROVIDER = {
     "providers": [{
         "name": "LLMStack",
         "type": "file",
-        "options": {"path": "/var/lib/grafana/dashboards"},
+        "options": {"path": "/opt/grafana/dashboards"},
     }],
 }
 
@@ -97,6 +104,12 @@ GRAFANA_DASHBOARD = {
 }
 
 
+def _write_file(path: Path, content: str) -> None:
+    """Write content to file, creating parent dirs as needed."""
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(content)
+
+
 class PrometheusService(ServiceBase):
     name = "prometheus"
     category = "observe"
@@ -105,7 +118,14 @@ class PrometheusService(ServiceBase):
         self.config = config
         self.host_port = 9090
 
+    def _prepare_config(self) -> str:
+        """Write prometheus.yml to ~/.llmstack/config/prometheus/ and return the dir path."""
+        config_dir = CONFIG_DIR / "prometheus"
+        _write_file(config_dir / "prometheus.yml", yaml.dump(PROMETHEUS_CONFIG, default_flow_style=False))
+        return str(config_dir)
+
     def container_spec(self) -> dict[str, Any]:
+        config_dir = self._prepare_config()
         return {
             "image": "prom/prometheus:latest",
             "name": "llmstack-prometheus",
@@ -116,7 +136,7 @@ class PrometheusService(ServiceBase):
                 "--web.enable-lifecycle",
             ],
             "volumes": {
-                "llmstack_prometheus_config": {"bind": "/etc/prometheus", "mode": "rw"},
+                config_dir: {"bind": "/etc/prometheus", "mode": "ro"},
                 "llmstack_prometheus_data": {"bind": "/prometheus", "mode": "rw"},
             },
             "environment": {},
@@ -127,7 +147,7 @@ class PrometheusService(ServiceBase):
 
     def get_config_yaml(self) -> str:
         """Return the prometheus.yml content."""
-        return PROMETHEUS_CONFIG
+        return yaml.dump(PROMETHEUS_CONFIG, default_flow_style=False)
 
 
 class GrafanaService(ServiceBase):
@@ -138,7 +158,32 @@ class GrafanaService(ServiceBase):
         self.config = config
         self.host_port = config.dashboard_port
 
+    def _prepare_provisioning(self) -> str:
+        """Write Grafana provisioning files to ~/.llmstack/config/grafana/ and return the dir."""
+        base = CONFIG_DIR / "grafana"
+
+        # Datasource provisioning
+        _write_file(
+            base / "provisioning" / "datasources" / "datasource.yml",
+            yaml.dump(GRAFANA_DATASOURCE, default_flow_style=False),
+        )
+
+        # Dashboard provider provisioning
+        _write_file(
+            base / "provisioning" / "dashboards" / "provider.yml",
+            yaml.dump(GRAFANA_DASHBOARD_PROVIDER, default_flow_style=False),
+        )
+
+        # Dashboard JSON
+        _write_file(
+            base / "dashboards" / "llmstack.json",
+            json.dumps(GRAFANA_DASHBOARD, indent=2),
+        )
+
+        return str(base)
+
     def container_spec(self) -> dict[str, Any]:
+        base = self._prepare_provisioning()
         return {
             "image": "grafana/grafana:latest",
             "name": "llmstack-grafana",
@@ -148,10 +193,18 @@ class GrafanaService(ServiceBase):
                 "GF_SECURITY_ADMIN_PASSWORD": "llmstack",
                 "GF_AUTH_ANONYMOUS_ENABLED": "true",
                 "GF_AUTH_ANONYMOUS_ORG_ROLE": "Viewer",
-                "GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH": "/var/lib/grafana/dashboards/llmstack.json",
+                "GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH": "/opt/grafana/dashboards/llmstack.json",
             },
             "volumes": {
-                "llmstack_grafana_data": {"bind": "/var/lib/grafana", "mode": "rw"},
+                os.path.join(base, "provisioning", "datasources"): {
+                    "bind": "/etc/grafana/provisioning/datasources", "mode": "ro",
+                },
+                os.path.join(base, "provisioning", "dashboards"): {
+                    "bind": "/etc/grafana/provisioning/dashboards", "mode": "ro",
+                },
+                os.path.join(base, "dashboards"): {
+                    "bind": "/opt/grafana/dashboards", "mode": "ro",
+                },
             },
         }
 

{llmstack_cli-0.1.0.dist-info → llmstack_cli-0.2.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llmstack-cli
-Version: 0.1.0
+Version: 0.2.0
 Summary: One command. Full LLM stack. Zero config.
 Author: mara-werils
 License-Expression: Apache-2.0
@@ -44,21 +44,43 @@ Description-Content-Type: text/markdown
   <a href="https://github.com/mara-werils/llmstack/actions/workflows/ci.yml"><img src="https://github.com/mara-werils/llmstack/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
   <a href="https://github.com/mara-werils/llmstack/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-green" alt="License"></a>
   <a href="https://www.python.org/"><img src="https://img.shields.io/badge/python-3.11+-blue" alt="Python"></a>
+  <a href="https://github.com/mara-werils/llmstack/stargazers"><img src="https://img.shields.io/github/stars/mara-werils/llmstack?style=social" alt="Stars"></a>
 </p>
 
 ---
 
-**llmstack** spins up a production-grade LLM stack locally with a single command. It auto-detects your hardware, picks the optimal inference backend, and wires everything together.
+<p align="center">
+  <img src="demo.gif" alt="llmstack demo" width="800">
+</p>
+
+## Quick Start
 
 ```bash
 pip install llmstack-cli
-llmstack init
+llmstack init --preset rag
 llmstack up
 ```
 
-That's it. You now have a full LLM API running locally.
+That's it. You now have **7 services** running: inference, embeddings, vector DB, cache, API gateway, Prometheus, and Grafana.
+
+```bash
+# Test it immediately
+curl http://localhost:8000/v1/chat/completions \
+  -H "Authorization: Bearer YOUR_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{"model":"llama3.2","messages":[{"role":"user","content":"Hello!"}]}'
+```
+
+Works with **any OpenAI-compatible client**: LangChain, LlamaIndex, Vercel AI SDK, openai-python.
+
+## Who is this for?
 
-## Architecture
+- **AI app developers** who want local inference without Docker boilerplate
+- **Teams** who need an OpenAI-compatible API backed by local models
+- **Hobbyists** running LLMs locally who want vector search, caching, and monitoring out of the box
+- **Anyone** tired of writing 200+ lines of docker-compose.yml every time
+
+## What you get
 
 ```
                          llmstack up
@@ -84,8 +106,6 @@ That's it. You now have a full LLM API running locally.
                                                    :8080
 ```
 
-## What you get
-
 | Layer | Service | Default | Port |
 |-------|---------|---------|------|
 | Inference | Ollama / vLLM (auto) | llama3.2 | 11434 |
@@ -97,7 +117,7 @@ That's it. You now have a full LLM API running locally.
 
 ## How it works
 
-```
+```bash
 llmstack init       # Detects hardware, generates llmstack.yaml
                     # Picks optimal backend: vLLM for NVIDIA 16GB+, Ollama otherwise
 
@@ -109,27 +129,6 @@ llmstack logs ollama # Stream inference logs
 llmstack down       # Stops everything
 ```
 
-### Use the API
-
-```bash
-curl http://localhost:8000/v1/chat/completions \
-  -H "Authorization: Bearer YOUR_KEY" \
-  -H "Content-Type: application/json" \
-  -d '{"model":"llama3.2","messages":[{"role":"user","content":"Hello!"}]}'
-```
-
-Works with **any OpenAI-compatible client**: LangChain, LlamaIndex, Vercel AI SDK, openai-python.
-
-```python
-from openai import OpenAI
-
-client = OpenAI(base_url="http://localhost:8000/v1", api_key="YOUR_KEY")
-response = client.chat.completions.create(
-    model="llama3.2",
-    messages=[{"role": "user", "content": "Explain quantum computing"}]
-)
-```
-
 ## Auto hardware detection
 
 | Your hardware | Backend | Why |
@@ -181,6 +180,50 @@ observe:
   dashboard_port: 8080
 ```
 
+## Interactive Chat
+
+```bash
+llmstack chat
+```
+
+```
+LLMStack Chat — model: llama3.2
+Type 'exit' or Ctrl+C to quit. '/clear' to reset conversation.
+
+You: What is quantum computing?
+Assistant: Quantum computing uses quantum mechanical phenomena like
+superposition and entanglement to process information...
+
+You: /clear
+Conversation cleared.
+```
+
+Streaming responses, conversation history, works with any model in your stack.
+
+## Export to Docker Compose
+
+Don't want to install llmstack? Generate a standalone `docker-compose.yml`:
+
+```bash
+llmstack export
+# Exported 7 services to docker-compose.yml
+# Run with: docker compose up -d
+```
+
+Share the generated file with your team — no llmstack dependency required.
+
+## Use the API
+
+```python
+from openai import OpenAI
+
+client = OpenAI(base_url="http://localhost:8000/v1", api_key="YOUR_KEY")
+response = client.chat.completions.create(
+    model="llama3.2",
+    messages=[{"role": "user", "content": "Explain quantum computing"}]
+)
+```
+
 ## CLI
 
 | Command | Description |
@@ -189,6 +232,8 @@ observe:
 | `llmstack up [--attach]` | Start all services |
 | `llmstack down [--volumes]` | Stop and clean up |
 | `llmstack status` | Health check all services |
+| `llmstack chat [--model]` | Interactive terminal chat |
+| `llmstack export [--output]` | Generate docker-compose.yml |
 | `llmstack logs <service>` | Stream service logs |
 | `llmstack doctor` | Diagnose system issues |
 
@@ -204,6 +249,34 @@ When `observe.metrics: true`, llmstack boots Prometheus + Grafana with a pre-bui
 
 Access at `http://localhost:8080` (login: admin / llmstack)
 
+## Why not just Docker Compose?
+
+Here's what llmstack replaces:
+
+```yaml
+# Without llmstack: ~200 lines of docker-compose.yml
+# You have to configure each service, write health checks,
+# set up networking, manage GPU passthrough, create Prometheus
+# scrape configs, provision Grafana dashboards...
+
+# With llmstack:
+llmstack init && llmstack up
+```
+
+## Comparison
+
+| | llmstack | Ollama | LocalAI | AnythingLLM | LiteLLM |
+|---|---|---|---|---|---|
+| One-command full stack | **Yes** | No (inference only) | No | Partial | No (proxy only) |
+| Auto hardware detection | **Yes** | No | No | No | No |
+| OpenAI-compatible API | **Yes** | Yes | Yes | No | Yes |
+| Built-in vector DB | **Yes** | No | No | Bundled | No |
+| Built-in embeddings | **Yes** | No | No | Bundled | No |
+| Caching (Redis) | **Yes** | No | No | No | No |
+| Auth + rate limiting | **Yes** | No | No | Yes | Yes |
+| Observability dashboard | **Yes** | No | Partial | No | Partial |
+| Plugin ecosystem | **Yes** | No | No | No | No |
+
 ## Plugins
 
 Extend llmstack with new backends via pip:
@@ -215,21 +288,6 @@ pip install llmstack-cli-plugin-chromadb
 
 Create your own: implement `ServiceBase`, register via entry_points. See [CONTRIBUTING.md](CONTRIBUTING.md).
 
-## Why llmstack?
-
-| | llmstack | Ollama | Harbor | AnythingLLM | LiteLLM |
-|---|---|---|---|---|---|
-| One-command full stack | Yes | No (inference only) | Partial | Partial | No (proxy only) |
-| Auto hardware detection | Yes | No | No | No | No |
-| OpenAI-compatible API | Yes | Yes | Varies | No | Yes |
-| Built-in vector DB | Yes | No | Config needed | Bundled | No |
-| Built-in embeddings | Yes | No | No | Bundled | No |
-| Caching (Redis) | Yes | No | No | No | No |
-| Auth + rate limiting | Yes | No | No | Yes | Yes |
-| Observability dashboard | Yes | No | Partial | No | Partial |
-| Plugin ecosystem | Yes | No | No | No | No |
-| SSE streaming | Yes | Yes | Yes | Yes | Yes |
-
 ## Tech stack
 
 - **CLI**: [Typer](https://typer.tiangolo.com/) + [Rich](https://rich.readthedocs.io/)

{llmstack_cli-0.1.0.dist-info → llmstack_cli-0.2.0.dist-info}/RECORD

@@ -1,11 +1,13 @@
-llmstack/__init__.py,sha256=eeHMsyABNyNRnVgYQqyU_R4RZpbJ7Kz3HzPWsisW5zo,84
+llmstack/__init__.py,sha256=ixXALf2pEpG4MqBgKPZQGn27zrgKhzcpe22AoV7XXdY,84
 llmstack/__main__.py,sha256=XhknKG6tlsgyslTC57MpF3IGUu_cIo6G5rSIz4kxuyc,86
 llmstack/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-llmstack/cli/app.py,sha256=qcrUcYkJApyX2JYVCFsu2wXZx81b0ma5aBXRNVNukSw,2378
+llmstack/cli/app.py,sha256=eYm9U7BNTFsFNYjfE3COHBw7FX0znRNtce4A0Lbb4ic,2924
 llmstack/cli/console.py,sha256=qnwJ2g39BwXhA_WhGtJW5E1ZdbTau4npG9j9mBo2zX0,259
 llmstack/cli/commands/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+llmstack/cli/commands/chat.py,sha256=pdhVHqHj2hXMcxhcywYUgAkFLHN98x8jQyCu4I-zgYU,3586
 llmstack/cli/commands/doctor.py,sha256=71PM5uG-a8A4nNf4ywMXmVIL8w6W5MwLgLco9DxbnG8,2440
 llmstack/cli/commands/down.py,sha256=51XiP8n_Nq3oX4gbeLd_vCvXB1OP7dZEi3392eYgSJM,752
+llmstack/cli/commands/export.py,sha256=Xv6r-htPfHCyxqOFQlC0jlRNAD7nP0ocM3NU_aXscWU,7425
 llmstack/cli/commands/init.py,sha256=1olPvaPLapdSkuSBVIum4kzXh2BMtkRw-HNZKs1slKc,2311
 llmstack/cli/commands/logs.py,sha256=CGZO9kFiqc224LtpYH22uf0k5E26sCmAucrozjtb7BQ,840
 llmstack/cli/commands/status.py,sha256=ahnhKsw0C6-x2umOVUNsCnTU5XQS-nGIjKOFxEZVjdA,1382
@@ -21,10 +23,9 @@ llmstack/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 llmstack/core/hardware.py,sha256=dO2VRD2OZ4BnDETkHBRw1YeTERz6XhFByQpy8z7-jWc,3844
 llmstack/core/health.py,sha256=BxH-Y1DsoDFc0zNxYVTdYEmKTC1dIWiBxhCLDLbUyr0,706
 llmstack/core/resolver.py,sha256=MPUNLAoe0U_htLtMy789RbBir6xljWnw9p0AGOHYrAc,1633
-llmstack/core/stack.py,sha256=v9kPijvUm5IF-B608b0fUw71wDv14vg0K1ovffSuHf4,8310
+llmstack/core/stack.py,sha256=p0nPreFr8B7g1kKDJ1gONxxrjt9uLV488qW5NXb2tpk,8812
 llmstack/docker/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-llmstack/docker/manager.py,sha256=OzfY8H46dEw7acSe8sUglayo4hgxO0tj3SxEC47Xmmg,4728
-llmstack/gateway/Dockerfile,sha256=Mtiy9fcfK1DikZ6v-I4zBI_Op1pABVm1nYZ-LT8J9fg,308
+llmstack/docker/manager.py,sha256=dpKFNgkpKYOvrIPHeMGHwmVGrHLXu9rIArlquZToQqI,5528
 llmstack/gateway/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 llmstack/gateway/main.py,sha256=PYT48w4MuBNCqJ3YTq27OF_yC0dRRpnC61bpos1aveY,1483
 llmstack/gateway/proxy.py,sha256=WLErhSUICVK1XWu6Qt9v5gBWYtwalP3LKJE_D5Dj6C4,2077
@@ -47,16 +48,17 @@ llmstack/services/cache/redis.py,sha256=_6_JlFxFEvVutkmFwo-qzola6heODG_fxrfGVt0x
 llmstack/services/embeddings/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 llmstack/services/embeddings/tei.py,sha256=GNJL6PxQVUu7FQEzR0jj1N2jDCkhZoO1ezKLcEdo5EI,1502
 llmstack/services/gateway/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-llmstack/services/gateway/service.py,sha256=4PSED82DHUlkqP4FxL-lNxY5kO1E-ODTN-x6msEY3HE,1580
+llmstack/services/gateway/service.py,sha256=lDJ1f7JMkSUMsnfUH-PjvZQLG37vs-zEQZajpV48d0E,2086
 llmstack/services/inference/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 llmstack/services/inference/ollama.py,sha256=0A-4V3Ks3fiyQwnJkBhLtj9oIvKAYPK-QgsbcAsuqFI,1814
 llmstack/services/inference/vllm.py,sha256=fhi7xa2RitNG0FNnhypUYfjEemvylnnxOIksRfBzJcE,1607
 llmstack/services/observe/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-llmstack/services/observe/prometheus.py,sha256=6NG8yDY2JQqYCWtSldb6LEKbOqLdHiPB9u9G0c2dJpw,5343
+llmstack/services/observe/prometheus.py,sha256=9gyoU7uhrNpQCvnmIKcr5xLpIx8e9z05LB31ri2UryI,7418
 llmstack/services/vectordb/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 llmstack/services/vectordb/qdrant.py,sha256=zIOqf5km-4pItXOW8Y-qdxHLiU9lFfhDz33nBrTgPhM,900
-llmstack_cli-0.1.0.dist-info/METADATA,sha256=aWalxTiNMqv0lz5GjICnaOhTC1QgYXRR8Tq7Jyk4LQE,8128
-llmstack_cli-0.1.0.dist-info/WHEEL,sha256=QccIxa26bgl1E6uMy58deGWi-0aeIkkangHcxk2kWfw,87
-llmstack_cli-0.1.0.dist-info/entry_points.txt,sha256=i2BIacwqAqUaN1yAe-MaJZ22unHqAAUkTopk9M_iZPo,50
-llmstack_cli-0.1.0.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-llmstack_cli-0.1.0.dist-info/RECORD,,
+llmstack/gateway/Dockerfile,sha256=ZoId9IGitvOViaF4m6cXo-GGVwPYxPWZZHKu89Tnjak,385
+llmstack_cli-0.2.0.dist-info/METADATA,sha256=nfDKQ6CRTHENvR7ohwk9FgSZxAnWd1rrCgimY6S2KKs,9791
+llmstack_cli-0.2.0.dist-info/WHEEL,sha256=QccIxa26bgl1E6uMy58deGWi-0aeIkkangHcxk2kWfw,87
+llmstack_cli-0.2.0.dist-info/entry_points.txt,sha256=i2BIacwqAqUaN1yAe-MaJZ22unHqAAUkTopk9M_iZPo,50
+llmstack_cli-0.2.0.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+llmstack_cli-0.2.0.dist-info/RECORD,,