Merge pull request #6 from mathysgrapotte/modal

main.py

@@ -2,25 +2,42 @@ from smolagents import CodeAgent, LiteLLMModel
 from smolagents.tools import ToolCollection
 import gradio as gr
 import requests
-from tools.meta_yml_tools import get_meta_yml_file, extract_tools_from_meta_json, extract_information_from_meta_json, extract_module_name_description, get_biotools_response
+import modal
+import sys
+import subprocess
+import time
+from .tools.meta_yml_tools import fetch_meta_yml,get_meta_yml_file, extract_tools_from_meta_json, extract_information_from_meta_json, extract_module_name_description,get_biotools_response
 
-def main(module_name): 
-    meta_yml = get_meta_yml_file(module_name=module_name)
-    module_info = extract_module_name_description(meta_file=meta_yml)
-    module_tools = extract_tools_from_meta_json(meta_file=meta_yml)
-    # TODO: agent to choose the right tool
-    # Only call the agent if there is more than one tool, otherwise get the first name
-    first_prompt = f"""
-        The module {module_info[0]} with desciption '{module_info[1]}' contains a series of tools. 
-        Find the tool that best describes the module. Return only one tool. Return the name. 
-        This is the list of tools:
-        {"\n\t".join(f"{tool[0]}: {tool[1]}" for tool in module_tools)}
-    """
-    tool_name = "fastqc" # this would be the answer of the first agent
-    meta_info = extract_information_from_meta_json(meta_file=meta_yml, tool_name=tool_name)
+# Define the custom image
+ollama_image = (
+    modal.Image.debian_slim()
+    .apt_install("curl", "gnupg", "software-properties-common", "procps")
+    .run_commands("curl -fsSL https://ollama.com/install.sh | sh")
+    .run_commands(
+        "bash -c 'ollama serve >/dev/null 2>&1 & "
+        "PID=$!; "
+        "sleep 10 && "
+        "ollama pull devstral:latest && "
+        "ollama pull qwen3:0.6b && "
+        "kill $PID'"
+    )
+    .pip_install(
+        "fastmcp>=2.6.1",
+        "gradio[mcp]>=5.0.0",
+        "huggingface_hub[mcp]>=0.32.2",
+        "mcp>=1.9.2",
+        "smolagents[litellm,mcp]>=1.17.0",
+        "textblob>=0.19.0",
+    )
+)
+
+# Initialize the Modal app with the custom image
+app = modal.App("agent-ontology", image=ollama_image)
 
 def chat_with_agent(message, history):
-    """Initialize MCP client for each request to avoid connection issues"""
+    """ Function to handle chat messages and interact with the agent.
+    This function creates a new MCP connection for each request, allowing the agent to use tools from the MCP server.
+    """
     try:
         with ToolCollection.from_mcp(
             {"url": "https://notredameslab-nf-ontology.hf.space/gradio_api/mcp/sse", "transport": "sse"},
@@ -50,41 +67,68 @@ def chat_with_agent(message, history):
             
     except Exception as e:
         return f"❌ Error: {e}\nType: {type(e).__name__}"
+    
+def run_multi_agent(module_name): 
+    meta_yml = get_meta_yml_file(module_name=module_name)
+    module_info = extract_module_name_description(meta_file=meta_yml)
+    module_tools = extract_tools_from_meta_json(meta_file=meta_yml)
+    # TODO: agent to choose the right tool
+    # Only call the agent if there is more than one tool, otherwise get the first name
+    first_prompt = f"""
+        The module {module_info[0]} with desciption '{module_info[1]}' contains a series of tools. 
+        Find the tool that best describes the module. Return only one tool. Return the name. 
+        This is the list of tools:
+        {"\n\t".join(f"{tool[0]}: {tool[1]}" for tool in module_tools)}
+    """
+    tool_name = "fastqc" # this would be the answer of the first agent
+    meta_info = extract_information_from_meta_json(meta_file=meta_yml, tool_name=tool_name)
+    return(meta_info)
 
-# TODO: placeholder function
-def fetch_meta_yml(module_name):
-    # Adjust the URL or path to your actual source of nf-core modules
-    base_url = f"https://raw.githubusercontent.com/nf-core/modules/refs/heads/master/modules/nf-core/{module_name}/meta.yml"
-    try:
-        response = requests.get(base_url)
-        response.raise_for_status()
-        content = response.text
-
-        # Save for download
-        with open("meta.yml", "w") as f:
-            f.write(content)
-
-        return content, "meta.yml"
-    except Exception as e:
-        return f"Error: Could not retrieve meta.yml for module '{module_name}'\n{e}", None
-
-if __name__ == "__main__":
+def run_interface():
+    """ Function to run the agent with a Gradio interface.
+    This function sets up the Gradio interface and launches it.
+    """
+    # create the Gradio interface
     with gr.Blocks() as demo:
         gr.Markdown("### 🔍 Update an nf-core module `meta.yml` file by adding EDAM ontology terms.")
 
-        with gr.Row():
-            module_input = gr.Textbox(label="nf-core Module Name", placeholder="e.g. fastqc")
+        # create the input textbox for the nf-core module name
+        module_input = gr.Textbox(label="nf-core Module Name", placeholder="e.g. fastqc")
 
+        # create the button to fetch the meta.yml file
         fetch_btn = gr.Button("Update meta.yml")
 
-        with gr.Row():
-            meta_output = gr.Textbox(label="meta.yml content", lines=20)
-            download_button = gr.File(label="Download meta.yml")
+        # create the output textbox for the meta.yml content and a download button
+        meta_output = gr.Textbox(label="meta.yml content", lines=20)
+        download_button = gr.File(label="Download meta.yml")
 
+        # set the function to run when the button is clicked
         fetch_btn.click(
-            fn=fetch_meta_yml, # TODO: change to final function
+            fn=run_multi_agent,  # TODO: change to final function
             inputs=module_input,
-            outputs=[meta_output, download_button]
+            outputs=[meta_output]
         )
+    
+    demo.launch(share=True)
 
-    demo.launch()
+@app.function(keep_warm=1, gpu="A10G", timeout=2400)
+def main_remote():
+    # spin up Ollama daemon in the background
+    server = subprocess.Popen(["ollama", "serve"])
+    time.sleep(6) # give it a moment to bind :11434
+    try:
+        run_interface()
+    finally:
+        server.terminate()
+
+def main_local():
+    run_interface()
+
+if __name__ == "__main__":
+    # check if it is modal running the script or python running the script
+    # if it is modal, run the remote function
+    # if it is python, run the local function
+    if "modal" in sys.modules:
+        main_remote().remote()
+    else:
+        main_local()

modal.py

@@ -1,89 +0,0 @@
-from smolagents import CodeAgent, LiteLLMModel
-from smolagents.tools import ToolCollection
-import gradio as gr
-import modal
-import subprocess
-import time
-
-app = modal.App("agent-ontology")
-
-OLLAMA_IMAGE = (
-    modal.Image.debian_slim()
-    # pkill/pgrep come from procps
-    .apt_install(
-        "curl", "gnupg", "software-properties-common",
-        "procps"               # ← adds pkill, pgrep, ps …
-    )
-    # install Ollama
-    .run_commands("curl -fsSL https://ollama.com/install.sh | sh")
-    # spin up daemon, pull the model, shut daemon down
-    .run_commands(
-        "bash -c 'ollama serve >/dev/null 2>&1 & "
-        "PID=$!; "
-        "sleep 10 && "
-        "ollama pull devstral:latest && "
-        "kill $PID'"
-    )
-    # python deps
-    .pip_install(
-        "fastmcp>=2.6.1",
-        "gradio[mcp]>=5.0.0",
-        "huggingface_hub[mcp]>=0.32.2",
-        "mcp>=1.9.2",
-        "smolagents[litellm,mcp]>=1.17.0",
-        "textblob>=0.19.0",
-    )
-)
-
-# Specify the dependencies in the Modal function
-@app.function(image=OLLAMA_IMAGE, gpu="A10G", timeout=2400)
-def run_agent():
-
-    def chat_with_agent(message, history):
-        """Initialize MCP client for each request to avoid connection issues"""
-        try:
-            with ToolCollection.from_mcp(
-                {"url": "https://notredameslab-nf-ontology.hf.space/gradio_api/mcp/sse", "transport": "sse"},
-                trust_remote_code=True  # Acknowledge that we trust this remote MCP server
-            ) as tool_collection:
-                
-                model = LiteLLMModel(
-                    model_id="ollama/devstral:latest",
-                    api_base="http://localhost:11434",
-                )
-
-                agent = CodeAgent(
-                    tools=tool_collection.tools,
-                    model=model,
-                    additional_authorized_imports=["inspect", "json"]
-                )
-
-                additional_instructions = """
-                ADDITIONAL IMPORTANT INSTRUCTIONS:
-                use the tool "final_answer" in the code block to provide the answer to the user. Prints are only for debugging purposes. So, to give your results concatenate everything you want to print in a single "final_answer" call as such : final_answer(f"your answer here").
-                """
-
-                agent.system_prompt += additional_instructions
-
-                result = agent.run(message)
-                return str(result)
-                
-        except Exception as e:
-            return f"❌ Error: {e}\nType: {type(e).__name__}"
-
-    demo = gr.ChatInterface(
-        fn=chat_with_agent,
-        type="messages",
-        examples=["can you extract input/output metadata from fastqc nf-core module ?"],
-        title="Agent with MCP Tools (Per-Request Connection)",
-        description="This version creates a new MCP connection for each request."
-    )
-    demo.launch(share=True)
-
-@app.local_entrypoint()
-def main():
-    """Run the Modal app locally."""
-    run_agent.remote()
-
-if __name__ == "__main__":
-    main()

tools/meta_yml_tools.py

@@ -2,6 +2,23 @@ import requests
 import json
 import yaml
 
+# TODO: placeholder function
+def fetch_meta_yml(module_name):
+    # Adjust the URL or path to your actual source of nf-core modules
+    base_url = f"https://raw.githubusercontent.com/nf-core/modules/refs/heads/master/modules/nf-core/{module_name}/meta.yml"
+    try:
+        response = requests.get(base_url)
+        response.raise_for_status()
+        content = response.text
+
+        # Save for download
+        with open("meta.yml", "w") as f:
+            f.write(content)
+
+        return content, "meta.yml"
+    except Exception as e:
+        return f"Error: Could not retrieve meta.yml for module '{module_name}'\n{e}", None
+
 def get_meta_yml_file(module_name: str) -> dict:
     """
     Access the nf-core/modules repository and return the meta.yml file of the given module.