github copilot integration

Author: eleonorecharles

README.md

@@ -15,7 +15,7 @@
 
 *The Jupyter AI Agents are equipped with tools like 'execute', 'insert_cell', and more, to transform your Jupyter Notebooks into an intelligent, interactive workspace!*
 
-![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agents/ai-agents-prompt-demo-terminal.gif)
+![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agent/ai-agent-prompt-demo-terminal.gif)
 
 ```
 Jupyter AI Agents <-----------> JupyterLab
@@ -73,6 +73,8 @@ Start JupyterLab, setting a `port` and a `token` to be reused by the agent, and
 jupyter lab --port 8888 --IdentityProvider.token MY_TOKEN
 ```
 
+Jupyter AI Agents supports multiple AI model providers (more information on [here](https://jupyter-ai-agents.datalayer.tech/docs/models/)). Here is an example with the Azure OpenAI provider.
+
 Read the [Azure Documentation](https://learn.microsoft.com/en-us/azure/ai-services/openai) to get the needed credentials and make sure you define them in the following `.env` file.
 
 ```bash
@@ -90,23 +92,25 @@ To use the Jupyter AI Agents, an easy way is to launch a CLI (update the Azure d
 jupyter-ai-agents prompt \
   --url http://localhost:8888 \
   --token MY_TOKEN \
-  --azure-ai-deployment-name gpt-40-mini \
+  --model-provider azure-openai \
+  --model-name gpt-4o-mini \
   --path test.ipynb \
   --input "Create a matplotlib example"
 ```
 
-![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agents/ai-agents-prompt-demo-terminal.gif)
+![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agent/ai-agent-prompt-demo-terminal.gif)
 
 ```bash
 # Explain Error agent example.
 jupyter-ai-agents explain-error \
   --url http://localhost:8888 \
   --token MY_TOKEN \
-  --azure-ai-deployment-name gpt-40-mini \
+  --model-provider azure-openai \
+  --model-name gpt-4o-mini \
   --path test.ipynb
 ```
 
-![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agents/ai-agents-explainerror-demo-terminal.gif)
+![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agent/ai-agent-explainerror-demo-terminal.gif)
 
 ## Uninstall
 

docs/docs/agents/explain_error/index.mdx

@@ -11,14 +11,16 @@ The Explain Error Agent explains an error encountered in a notebook. It leverage
 jupyter-ai-agents explain-error \
   --url http://localhost:8888 \
   --token MY_TOKEN \
-  --azure-ai-deployment-name gpt-40-mini \
+  --model-provider azure-openai \
+  --model-name gpt-4o-mini \
   --path test.ipynb
 ```
 
 ## Parameters
 
 - `--url`: JupyterLab URL.
 - `--token`: JupyterLab token.
-- `--azure-ai-deployment-name`: Azure AI model deployment name.
+- `--model-provider`: `azure-openai` or `github-copilot`.
+- `--model-name`: Azure AI model deployment name or Github Copilot model to use.
 - `--path`: Notebook to modify path.
 - `--current-cell-index`: Optional flag to provide the index of the cell where the error is encountered. If not provided, the error considered is the first one in the notebook.

docs/docs/agents/prompt/index.mdx

@@ -1,9 +1,10 @@
 # Prompt Agent
 
-The Prompt Agent generates code cells based on natural language input. It leverages AI models to generate code, add new cells, and modify the notebook content. This is a great Agent to support the following use cases:
+The Prompt Agent generates code cells based on natural language input. It leverages AI models to generate code, markdown, add new cells, and modify the notebook content. This is a great Agent to support the following use cases:
 
 - Code Generation: Generate cells code based on natural language input.
-- Notebook Modification: Add new cells based on existing one.
+- Markdown Generation: Generate markdown cells based on natural language input.
+- Notebook Modification: Insert new cells based on existing one to your notebook.
 
 ![Jupyter AI Agents](https://assets.datalayer.tech/jupyter-ai-agents/ai-agent-prompt-demo-terminal.gif)
 
@@ -13,7 +14,8 @@ To use the Jupyter AI Agents, an easy way is to launch a CLI (update the Azure d
 jupyter-ai-agents prompt \
   --url http://localhost:8888 \
   --token MY_TOKEN \
-  --azure-ai-deployment-name gpt-40-mini \
+  - model-provider azure-openai \
+  --model-name gpt-4o-mini \
   --path test.ipynb \
   --input "Create a matplotlib example"
 ```
@@ -23,7 +25,8 @@ jupyter-ai-agents prompt \
 The Prompt Agent can be configured with the following parameters:
 - `--url`: JupyterLab URL.
 - `--token`: JupyterLab token.
-- `--azure-ai-deployment-name`: Azure AI model deployment name.
+- `--model-provider`: `azure-openai` or `github-copilot`.
+- `--model-name`: Azure AI model deployment name or Github Copilot model to use.
 - `--path`: Notebook to modify path.
 - `--input`: Natural language input.
 - `--full-context`: Optional flag to provide the full notebook context i.e. notebook content to the AI model (default: False).

docs/docs/models/github_copilot/_category_.yaml

@@ -0,0 +1,2 @@
+label: "GitHub Copilot"
+position: 2

docs/docs/models/github_copilot/index.mdx

@@ -0,0 +1,17 @@
+# GitHub Copilot
+
+Jupyter AI Agents supports models from [GitHub Copilot](https://copilot.github.com/).
+
+Read the [Langchain GitHubCopilot README](https://github.com/datalayer/langchain-github-copilot) to get the `GITHUB` token and make sur you define it in the following environment variable.
+
+```bash
+export GITHUB_TOKEN="..."
+```
+
+GitHub Copilot supports different models that you can specify using the `--model-name` parameter. The available models as of 2024-02-07 are 
+- gpt-4o
+- o1
+- o3-mini
+
+Make sure you have a GitHubCopilot subscription and have enabled the model you want to use in your GitHubCopilot settings.
+

docs/docs/models/index.mdx

@@ -2,4 +2,6 @@ import DocCardList from '@theme/DocCardList';
 
 # Models
 
+Jupyter AI Agents can be used with multiple AI models providers.
+
 <DocCardList />

jupyter_ai_agents/__version__.py

@@ -4,4 +4,4 @@
 
 """Jupyter AI Agents."""
 
-__version__ = "0.2.0"
+__version__ = "0.3.0"

jupyter_ai_agents/agents/explain_error.py

@@ -5,13 +5,11 @@
 import logging
 
 from langchain.agents import tool
-from langchain.agents import AgentExecutor
 
-from jupyter_ydoc import YNotebook
 from jupyter_nbmodel_client import NbModelClient
 from jupyter_kernel_client import KernelClient
 
-from jupyter_ai_agents.providers.azure_openai import create_azure_open_ai_agents
+from jupyter_ai_agents.agents.utils import create_ai_agent
 from jupyter_ai_agents.tools import insert_execute_code_cell_tool
 from jupyter_ai_agents.utils import retrieve_cells_content_until_first_error, retrieve_cells_content_error
 
@@ -25,7 +23,7 @@
 """
 
 
-def explain_error(notebook: NbModelClient, kernel: KernelClient, azure_deployment_name: str, current_cell_index: int) -> list:
+def explain_error(notebook: NbModelClient, kernel: KernelClient, model_provider: str, model_name: str, current_cell_index: int) -> list:
     """Explain and correct an error in a notebook based on the prior cells."""
 
     @tool
@@ -61,7 +59,6 @@ def insert_execute_code_cell(cell_index: int, cell_content: str) -> str:
     logger.debug("Prompt with content", system_prompt_final)
     logger.debug("Input", input)
 
-    agent = create_azure_open_ai_agents(azure_deployment_name, system_prompt_final, tools)
-    agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
-    
-    return list(agent_executor.stream({"input": input}))
+    agent = create_ai_agent(model_provider, model_name, system_prompt_final, tools)
+        
+    return list(agent.stream({"input": input}))

jupyter_ai_agents/agents/prompt.py

@@ -3,12 +3,11 @@
 # BSD 3-Clause License
 
 from langchain.agents import tool
-from langchain.agents import AgentExecutor
 
 from jupyter_nbmodel_client import NbModelClient
 from jupyter_kernel_client import KernelClient
 
-from jupyter_ai_agents.providers.azure_openai import create_azure_open_ai_agents
+from jupyter_ai_agents.agents.utils import create_ai_agent
 from jupyter_ai_agents.tools import insert_execute_code_cell_tool, insert_markdown_cell_tool
 from jupyter_ai_agents.utils import retrieve_cells_content
 
@@ -19,7 +18,7 @@
 Ensure updates to cell indexing when new cells are inserted. Maintain the logical flow of execution by adjusting cell index as needed.
 """
 
-def prompt(notebook: NbModelClient, kernel: KernelClient, input: str, azure_deployment_name: str, full_context: bool, current_cell_index: int) -> list:
+def prompt(notebook: NbModelClient, kernel: KernelClient, input: str, model_provider: str, model_name: str, full_context: bool, current_cell_index: int) -> list:
     """From a given instruction, code and markdown cells are added to a notebook."""
 
     @tool
@@ -54,7 +53,6 @@ def insert_markdown_cell(cell_index: int, cell_content: str) -> str:
     else:
         system_prompt_final = system_prompt_enriched
 
-    agent = create_azure_open_ai_agents(azure_deployment_name, system_prompt_final, tools)
-    agent_executor = AgentExecutor(name="NotebookPromptAgent", agent=agent, tools=tools, verbose=True)
+    agent = create_ai_agent(model_provider, model_name, system_prompt_final, tools)
 
-    return list(agent_executor.stream({"input": input}))
+    return list(agent.stream({"input": input}))

jupyter_ai_agents/agents/utils.py

@@ -0,0 +1,16 @@
+# Copyright (c) 2023-2024 Datalayer, Inc.
+#
+# BSD 3-Clause License
+
+from jupyter_ai_agents.providers.azure_openai import create_azure_openai_agent
+from jupyter_ai_agents.providers.github_copilot import create_github_copilot_agent
+
+def create_ai_agent(model_provider: str, model_name: str, system_prompt_final: str, tools: list):
+    """Create an AI agent based on the model provider."""
+    if model_provider == "azure-openai": 
+        agent = create_azure_openai_agent(model_name, system_prompt_final, tools)
+    elif model_provider == "github-copilot":
+        agent = create_github_copilot_agent(model_name, system_prompt_final, tools)
+    else:
+        raise ValueError(f"Model provider {model_provider} is not supported.")
+    return agent

jupyter_ai_agents/app.py

@@ -53,7 +53,7 @@ def initialize(self, *args, **kwargs):
         super(PromptAgentApp, self).initialize(*args, **kwargs)
 
     def ask(self):
-        reply = prompt(self.notebook, self.kernel, super().input, super().azure_ai_deployment_name, self.full_context, self.current_cell_index)
+        reply = prompt(self.notebook, self.kernel, super().input, super().model_provider, super().model_name, self.full_context, self.current_cell_index)
         logger.debug("Reply", reply)
 
     def start(self):
@@ -78,7 +78,7 @@ def initialize(self, *args, **kwargs):
         super(ExplainErrorAgentApp, self).initialize(*args, **kwargs)
 
     def ask(self):
-        reply = explain_error(self.notebook, self.kernel, super().azure_ai_deployment_name, self.current_cell_index)
+        reply = explain_error(self.notebook, self.kernel, super().model_provider, super().model_name, self.current_cell_index)
         logger.debug("Reply", reply)
 
     def start(self):

jupyter_ai_agents/base.py

@@ -40,10 +40,11 @@
         "path": "JupyterAIAgentBaseApp.path",
         "agent": "JupyterAIAgentBaseApp.agent_name",
         "input": "JupyterAIAgentBaseApp.input",
+        "model-provider": "JupyterAIAgentBaseApp.model_provider",
         "openai-api-version": "JupyterAIAgentBaseApp.openai_api_version",
         "azure-openai-version": "JupyterAIAgentBaseApp.azure_openai_version",
         "azure-openai-api-key": "JupyterAIAgentBaseApp.azure_openai_api_key",
-        "azure-ai-deployment-name": "JupyterAIAgentBaseApp.azure_ai_deployment_name",
+        "model-name": "JupyterAIAgentBaseApp.model_name",
         "current-cell-index": "JupyterAIAgentBaseApp.current_cell_index",
     }
 )
@@ -73,9 +74,8 @@ class JupyterAIAgentBaseApp(JupyterApp):
     path = Unicode(
         "",
         config=True,
-        help="Jupyter Notebok path."
+        help="Jupyter Notebook path."
     )
-
     agent_name = Unicode(
         "prompt",
         config=True,
@@ -86,7 +86,11 @@ class JupyterAIAgentBaseApp(JupyterApp):
         config=True,
         help="Input."
     )
-
+    model_provider = Unicode(
+        "github-copilot",
+        config=True,
+        help="Model provider can be 'azure-openai' or 'github-copilot'."
+    )
     openai_api_version = Unicode(
         os.environ.get("OPENAI_API_VERSION"),
         help="""OpenAI version.""",
@@ -102,9 +106,18 @@ class JupyterAIAgentBaseApp(JupyterApp):
         help="""Azure OpenAI key.""",
         config=True,
     )
-    azure_ai_deployment_name = Unicode(
-        "",
-        help="""Azure AI deployment name.""",
+    github_token = Unicode(
+        os.environ.get("GITHUB_TOKEN"),
+        help="""Github token.""",
+        config=True,
+    )
+    model_name = Unicode(
+        "gpt-4o",
+        help=(
+            "The 'Azure AI deployment' name for 'azure-openai' model provider."
+            "For 'github-copilot' model provider, gpt-4o, o1, or o3-mini (as of 2024-02-07) "
+            "- check your GithubCopilot settings to make sure the model you want to use is enabled."""
+            ),
         config=True,
     )
     current_cell_index = Integer(

jupyter_ai_agents/providers/azure_openai.py

@@ -2,18 +2,15 @@
 #
 # BSD 3-Clause License
 
-from langchain.agents import tool
 from langchain_openai import AzureChatOpenAI
+from langchain.agents import AgentExecutor, create_openai_tools_agent
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
-from langchain.agents.format_scratchpad.openai_tools import format_to_openai_tool_messages
-from langchain.agents.output_parsers.openai_tools import OpenAIToolsAgentOutputParser
 
 
-def create_azure_open_ai_agents(azure_deployment_name: str, system_prompt: str, tools: list) -> dict:
-    """Create an agent from a set of tools and an Azure deployment name."""
-
-    llm = AzureChatOpenAI(azure_deployment=azure_deployment_name)
-    llm_with_tools = llm.bind_tools(tools)
+def create_azure_openai_agent(model_name: str, system_prompt: str, tools: list) -> dict:
+    """Create an agent from a set of tools and an Azure deployment"""
+    
+    llm = AzureChatOpenAI(azure_deployment=model_name)
 
     prompt = ChatPromptTemplate.from_messages(
     [
@@ -22,15 +19,9 @@ def create_azure_open_ai_agents(azure_deployment_name: str, system_prompt: str,
         MessagesPlaceholder(variable_name="agent_scratchpad"),
     ])
 
-    agent = (
-    {
-        "input": lambda x: x["input"],
-        "agent_scratchpad": lambda x: format_to_openai_tool_messages(
-            x["intermediate_steps"]
-        ),
-    }
-    | prompt
-    | llm_with_tools
-    | OpenAIToolsAgentOutputParser())
+    # Create the agent using LangChain's built-in tool handling
+    agent = create_openai_tools_agent(llm, tools, prompt)
+    
+    agent_executor = AgentExecutor(name="NotebookPromptAgent", agent=agent, tools=tools, verbose=True, handle_parsing_errors=True)
 
-    return agent
+    return agent_executor

jupyter_ai_agents/providers/github_copilot.py

@@ -0,0 +1,27 @@
+# Copyright (c) 2023-2024 Datalayer, Inc.
+#
+# BSD 3-Clause License
+
+from langchain_github_copilot import ChatGithubCopilot
+from langchain.agents import AgentExecutor, create_openai_tools_agent
+from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
+
+
+def create_github_copilot_agent(model_name: str, system_prompt: str, tools: list) -> dict:
+    """Create an agent from a set of tools and a Github Copilot model"""
+    
+    llm = ChatGithubCopilot(model_name=model_name)
+
+    prompt = ChatPromptTemplate.from_messages(
+    [
+        ("system", system_prompt),
+        ("user", "{input}"),
+        MessagesPlaceholder(variable_name="agent_scratchpad"),
+    ])
+
+    # Create the agent using LangChain's built-in tool handling
+    agent = create_openai_tools_agent(llm, tools, prompt)
+    
+    agent_executor = AgentExecutor(name="NotebookPromptAgent", agent=agent, tools=tools, verbose=True, handle_parsing_errors=True)
+
+    return agent_executor

jupyter_ai_agents/tools.py

@@ -2,10 +2,6 @@
 #
 # BSD 3-Clause License
 
-from dotenv import load_dotenv, find_dotenv
-
-from langchain.agents import tool
-
 from jupyter_nbmodel_client import NbModelClient
 from jupyter_kernel_client import KernelClient
 

pyproject.toml

@@ -23,6 +23,7 @@ classifiers = [
 dependencies = [
   "langchain",
   "langchain-openai",
+  "langchain-github-copilot",
   "jupyter_kernel_client",
   "jupyter_nbmodel_client",
   "python-dotenv",