First commit

Dockerfile

@@ -0,0 +1,13 @@
+FROM python:3.13-slim
+
+RUN useradd -m -u 1000 user
+WORKDIR /app
+
+COPY --chown=user ./requirements.txt requirements.txt
+RUN pip install --no-cache-dir --upgrade -r requirements.txt
+
+COPY --chown=user . /app
+
+EXPOSE 7860
+
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,7 +1,7 @@
 ---
 title: IriusRiskTestChallenge
 emoji: 🚀
-colorFrom: yellow
+colorFrom: green
 colorTo: indigo
 sdk: docker
 pinned: false
@@ -9,4 +9,76 @@ license: apache-2.0
 short_description: LLM backend for IriusRisk Tech challenge
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# IriusRisk test challenge
+
+This project implements a FastAPI API that uses LangChain and LangGraph to generate text with the SmolLM2-1.7B-Instruct model from HuggingFace.
+
+## Configuration
+
+### In HuggingFace Spaces
+
+This project is designed to run in HuggingFace Spaces. To configure it:
+
+1. Create a new Space in HuggingFace with SDK Docker
+
+### Local development
+
+For local development:
+
+1. Clone this repository
+3. Install the dependencies:
+   ```
+   pip install -r requirements.txt
+   ```
+
+## Local execution
+
+```bash
+uvicorn app:app --reload
+```
+
+The API will be available at `http://localhost:7860`.
+
+## Endpoints
+
+### GET `/`
+
+Welcome endpoint that returns a greeting message.
+
+### POST `/generate`
+
+Endpoint to generate text using the language model.
+
+**Request parameters:**
+```json
+{
+  "query": "Your question here",
+  "thread_id": "optional_thread_identifier"
+}
+```
+
+**Response:**
+```json
+{
+  "generated_text": "Generated text by the model",
+  "thread_id": "thread identifier"
+}
+```
+
+## Docker
+
+To run the application in a Docker container:
+
+```bash
+# Build the image
+docker build -t iriusrisk-test-challenge .
+
+# Run the container
+docker run -p 7860:7860 iriusrisk-test-challenge
+```
+
+## API documentation
+
+The interactive API documentation is available at:
+- Swagger UI: `http://localhost:7860/docs`
+- ReDoc: `http://localhost:7860/redoc`

app.py

@@ -0,0 +1,150 @@
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import torch
+
+from langchain_core.messages import HumanMessage, AIMessage
+from langgraph.checkpoint.memory import MemorySaver
+from langgraph.graph import START, MessagesState, StateGraph
+
+import os
+from dotenv import load_dotenv
+load_dotenv()
+
+# Initialize the model and tokenizer
+print("Loading model and tokenizer...")
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model_name = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
+
+try:
+    # Load the model in BF16 format for better performance and lower memory usage
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    
+    if device == "cuda":
+        print("Using GPU for the model...")
+        model = AutoModelForCausalLM.from_pretrained(
+            model_name,
+            torch_dtype=torch.bfloat16,
+            device_map="auto",
+            low_cpu_mem_usage=True
+        )
+    else:
+        print("Using CPU for the model...")
+        model = AutoModelForCausalLM.from_pretrained(
+            model_name,
+            device_map={"": device},
+            torch_dtype=torch.float32
+        )
+
+    print(f"Model loaded successfully on: {device}")
+except Exception as e:
+    print(f"Error loading the model: {str(e)}")
+    raise
+
+# Define the function that calls the model
+def call_model(state: MessagesState):
+    """
+    Call the model with the given messages
+
+    Args:
+        state: MessagesState
+
+    Returns:
+        dict: A dictionary containing the generated text and the thread ID
+    """
+    # Convert LangChain messages to chat format
+    messages = [
+        {"role": "system", "content": "You are a friendly Chatbot. Always reply in the language in which the user is writing to you."}
+    ]
+    
+    for msg in state["messages"]:
+        if isinstance(msg, HumanMessage):
+            messages.append({"role": "user", "content": msg.content})
+        elif isinstance(msg, AIMessage):
+            messages.append({"role": "assistant", "content": msg.content})
+    
+    # Prepare the input using the chat template
+    input_text = tokenizer.apply_chat_template(messages, tokenize=False)
+    inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
+    
+    # Generate response
+    outputs = model.generate(
+        inputs,
+        max_new_tokens=512,  # Increase the number of tokens for longer responses
+        temperature=0.7,
+        top_p=0.9,
+        do_sample=True,
+        pad_token_id=tokenizer.eos_token_id
+    )
+    
+    # Decode and clean the response
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # Extract only the assistant's response (after the last user message)
+    response = response.split("Assistant:")[-1].strip()
+    
+    # Convert the response to LangChain format
+    ai_message = AIMessage(content=response)
+    return {"messages": state["messages"] + [ai_message]}
+
+# Define the graph
+workflow = StateGraph(state_schema=MessagesState)
+
+# Define the node in the graph
+workflow.add_edge(START, "model")
+workflow.add_node("model", call_model)
+
+# Add memory
+memory = MemorySaver()
+graph_app = workflow.compile(checkpointer=memory)
+
+# Define the data model for the request
+class QueryRequest(BaseModel):
+    query: str
+    thread_id: str = "default"
+
+# Create the FastAPI application
+app = FastAPI(title="LangChain FastAPI", description="API to generate text using LangChain and LangGraph - Máximo Fernández Núñez IriusRisk test challenge")
+
+# Welcome endpoint
+@app.get("/")
+async def api_home():
+    """Welcome endpoint"""
+    return {"detail": "Welcome to Máximo Fernández Núñez IriusRisk test challenge"}
+
+# Generate endpoint
+@app.post("/generate")
+async def generate(request: QueryRequest):
+    """
+    Endpoint to generate text using the language model
+    
+    Args:
+        request: QueryRequest
+        query: str
+        thread_id: str = "default"
+
+    Returns:
+        dict: A dictionary containing the generated text and the thread ID
+    """
+    try:
+        # Configure the thread ID
+        config = {"configurable": {"thread_id": request.thread_id}}
+        
+        # Create the input message
+        input_messages = [HumanMessage(content=request.query)]
+        
+        # Invoke the graph
+        output = graph_app.invoke({"messages": input_messages}, config)
+        
+        # Get the model response
+        response = output["messages"][-1].content
+        
+        return {
+            "generated_text": response,
+            "thread_id": request.thread_id
+        }
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"Error generating text: {str(e)}")
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)

requirements.txt

@@ -0,0 +1,11 @@
+fastapi
+uvicorn
+requests
+pydantic>=2.0.0
+langchain>=0.1.0
+langchain-core>=0.1.10
+langgraph>=0.2.27
+python-dotenv>=1.0.0
+transformers>=4.36.0
+torch>=2.0.0
+accelerate>=0.26.0