llama-cpp-runner/Dockerfile
2025-04-18 18:39:29 -04:00

117 lines
3.6 KiB
Docker

FROM python:3.11-slim
WORKDIR /app
# Install only essential packages and clean up in one layer to reduce image size
RUN apt-get update && apt-get install -y --no-install-recommends \
curl \
wget \
git \
build-essential \
&& apt-get clean \
&& rm -rf /var/lib/apt/lists/*
# Copy only necessary files
COPY pyproject.toml README.md LICENSE /app/
COPY src/ /app/src/
# Install the package in development mode and required dependencies
RUN pip install --no-cache-dir -e . && pip install --no-cache-dir requests fastapi uvicorn
# Create volume mount points
VOLUME /models
VOLUME /cache
# Create proxy server script directly in the Dockerfile
RUN echo 'import os\n\
import uvicorn\n\
from fastapi import FastAPI, Request\n\
from fastapi.responses import StreamingResponse, JSONResponse\n\
from llama_cpp_runner.main import LlamaCpp\n\
\n\
app = FastAPI(title="LlamaCpp Proxy")\n\
\n\
# Initialize the LlamaCpp class\n\
models_dir = os.environ.get("MODELS_DIR", "/models")\n\
cache_dir = os.environ.get("CACHE_DIR", "/cache")\n\
verbose = os.environ.get("VERBOSE", "true").lower() == "true"\n\
timeout = int(os.environ.get("TIMEOUT_MINUTES", "30"))\n\
\n\
print(f"Models directory: {models_dir}")\n\
print(f"Cache directory: {cache_dir}")\n\
\n\
# Create the LlamaCpp instance\n\
llama_runner = LlamaCpp(\n\
models_dir=models_dir,\n\
cache_dir=cache_dir, \n\
verbose=verbose, \n\
timeout_minutes=timeout\n\
)\n\
\n\
@app.get("/")\n\
def read_root():\n\
"""Get server status and list of available models."""\n\
return {"status": "running", "models": llama_runner.list_models()}\n\
\n\
@app.post("/v1/chat/completions")\n\
async def chat_completions(request: Request):\n\
"""Forward chat completion requests to the LlamaCpp server."""\n\
try:\n\
body = await request.json()\n\
\n\
if "model" not in body:\n\
return JSONResponse(\n\
status_code=400,\n\
content={"error": "Model not specified in request"}\n\
)\n\
\n\
try:\n\
result = llama_runner.chat_completion(body)\n\
\n\
# Handle streaming responses\n\
if body.get("stream", False):\n\
async def generate():\n\
for line in result:\n\
if line:\n\
yield f"data: {line}\\n\\n"\n\
yield "data: [DONE]\\n\\n"\n\
\n\
return StreamingResponse(generate(), media_type="text/event-stream")\n\
else:\n\
return result\n\
except Exception as e:\n\
return JSONResponse(\n\
status_code=500,\n\
content={"error": str(e)}\n\
)\n\
except Exception as e:\n\
return JSONResponse(\n\
status_code=400,\n\
content={"error": f"Invalid request: {str(e)}"}\n\
)\n\
\n\
@app.get("/models")\n\
def list_models():\n\
"""List all available models."""\n\
return {"models": llama_runner.list_models()}\n\
\n\
if __name__ == "__main__":\n\
print("Starting LlamaCpp Proxy Server on port 3636")\n\
models = llama_runner.list_models()\n\
print(f"Available models: {models}")\n\
if not models:\n\
print("WARNING: No models found in the models directory.")\n\
uvicorn.run(app, host="0.0.0.0", port=3636)' > /app/proxy_server.py
# Expose the proxy server port
EXPOSE 3636
# Set environment variables
ENV PYTHONUNBUFFERED=1
ENV MODELS_DIR=/models
ENV CACHE_DIR=/cache
ENV VERBOSE=true
ENV TIMEOUT_MINUTES=30
# Command to run when the container starts
CMD ["python", "/app/proxy_server.py"]