Add Composio integration to the third party tools

README.md

@@ -58,6 +58,7 @@ Disclaimer: Examples contributed by the community and partners do not represent
 | [CAMEL Graph RAG with Mistral Models](third_party/CAMEL_AI/camel_graph_rag.ipynb)                                     | multi-agent, tool, data gen     | CAMEL-AI.org|
 | [CAMEL Role-Playing Scraper](third_party/CAMEL_AI/camel_roleplaying_scraper.ipynb)                                    | multi-agent, tool, data gen     | CAMEL-AI.org|
 | [Chainlit - Mistral reasoning.ipynb](third_party/Chainlit/Chainlit_Mistral_reasoning.ipynb)                           | UI chat, tool calling           | Chainlit   |
+| [Composio Integration - AI email assistant ](third_party/Composio)                                                    | Agent, tool calling             | Composio   |
 | [corrective_rag_mistral.ipynb](third_party/langchain/corrective_rag_mistral.ipynb)                                    | RAG                             | Langchain  |
 | [distilabel_synthetic_dpo_dataset.ipynb](third_party/argilla/distilabel_synthetic_dpo_dataset.ipynb)                  | synthetic data                  | Argilla    |
 | [E2B Code Interpreter SDK with Codestral](third_party/E2B_Code_Interpreting)                                          | tool, agent                     | E2B        |

third_party/Composio/gmail-agent-js/.env.template

@@ -0,0 +1,2 @@
+MISTRAL_API_KEY=
+COMPOSIO_API_KEY=

third_party/Composio/gmail-agent-js/.gitignore

@@ -0,0 +1,2 @@
+node_modules
+.env

third_party/Composio/gmail-agent-js/email-send-example.js

@@ -0,0 +1,78 @@
+import dotenv from "dotenv";
+import { AgentExecutor, createToolCallingAgent } from "langchain/agents";
+import { LangchainToolSet } from "composio-core";
+import { ChatPromptTemplate } from "@langchain/core/prompts";
+import { ChatMistralAI } from "@langchain/mistralai";
+
+dotenv.config();
+
+// Initialize the toolset
+const toolset = new LangchainToolSet({ apiKey: process.env.COMPOSIO_API_KEY });
+
+// Create Gmail integration with Composio
+async function setupUserConnection(entityId) {
+  const entity = toolset.client.getEntity(entityId);
+  const connection = await entity.getConnection("gmail");
+
+  if (!connection) {
+    // If this entity/user hasn't already connected the account
+    const connection = await entity.initiateConnection("gmail");
+    console.log("Log in via: ", connection.redirectUrl);
+
+    return connection.waitUntilActive(100);
+  }
+  return connection;
+}
+
+async function executeAgent(entityName) {
+
+  // Create entity and set up user connection
+  const entity = await toolset.client.getEntity(entityName);
+  const a = await setupUserConnection(entity.id);
+
+  // Get the desired actions
+  const tools = await toolset.getTools(
+    {
+      actions: [
+        "gmail_send_email",
+        "gmail_fetch_emails",
+        "gmail_create_email_draft",
+        "gmail_create_label",
+      ],
+    },
+    entity.id
+  );
+
+  // Initialize the LLM
+  const llm = new ChatMistralAI({
+    model: "mistral-large-latest",
+    apiKey: process.env.MISTRAL_API_KEY,
+  });
+
+  // Create custom prompt
+  const prompt = ChatPromptTemplate.fromMessages([
+    [
+      "system",
+      "You are a helpful and thorough AI email assistant who can write and fetch emails, create draft mails and also create new gmail labels. Your goal is to understand the guidelines provided by the user and perform the specific actions requested by the user. If the user asks to fetch emails, use the actual data from the API response and print the result including these terms: Subject and Mail in an easy-to-read format.",
+    ],
+    ["human", "{input}"],
+    ["placeholder", "{agent_scratchpad}"],
+  ]);
+
+  // Create the agent
+  const agent = await createToolCallingAgent({
+    llm,
+    tools: tools,
+    prompt,
+  });
+
+  // Execute the agent
+  const agentExecutor = new AgentExecutor({ agent, tools, verbose: true });
+  const result = await agentExecutor.invoke({
+    // Customize the input -
+    input: "Send a mail to [email protected] saying Hello",
+  });
+
+  console.log(result.output);
+}
+executeAgent("default");