diff --git a/agents/ResearchAgent-soln.js b/agents/ResearchAgent-soln.js
deleted file mode 100644
index b7fbf06..0000000
--- a/agents/ResearchAgent-soln.js
+++ /dev/null
@@ -1,81 +0,0 @@
-import { ChatOpenAI } from "langchain/chat_models/openai";
-import { LLMChain } from "langchain/chains";
-import { ZeroShotAgent } from "langchain/agents";
-import {
- ChatPromptTemplate,
- HumanMessagePromptTemplate,
- SystemMessagePromptTemplate,
-} from "langchain/prompts";
-import { AgentExecutor } from "langchain/agents";
-import SerpAPITool from "../tools/SerpAPI";
-import WebBrowserTool from "../tools/WebBrowser";
-
-/**
- *
- * WARNING: THIS IS THE SOLUTION! Please try coding before viewing this.
- *
- */
-const ResearchAgent = async (topic) => {
- console.log({ topic });
- try {
- // We'll give it tohe ability to search Google and also Browse the WEb
- // show the importance of returnDirect
- const SerpAPI = SerpAPITool();
- const WebBrowser = WebBrowserTool();
-
- console.log(SerpAPI.returnDirect);
- console.log(WebBrowser.returnDirect);
- // We put them into an array of tools
- const tools = [SerpAPI, WebBrowser];
-
- // We'll use the ZeroShotReactDescription which is the recommended tool for chat models
- // https://js.langchain.com/docs/modules/agents/agents/
- // The cool thing is that Langchain gives us a method that can create prompts specifically for that agent
- const promptTemplate = ZeroShotAgent.createPrompt(tools, {
- prefix: `Answer the following questions as best you can. You have access to the following tools:`,
- suffix: `Begin! Answer concisely. It's OK to say you don't know.`,
- });
-
- // Then we'll initialize it with a Prompt template again
- const chatPrompt = ChatPromptTemplate.fromPromptMessages([
- new SystemMessagePromptTemplate(promptTemplate),
- HumanMessagePromptTemplate.fromTemplate(`{input}`),
- ]);
-
- // And we'll initialize the model, what is chatOpenAI since we're using a ChatAgent
- const chat = new ChatOpenAI({});
- // We'll create an LLM chain which just a prompt template and a LLM or chat model
- const llmChain = new LLMChain({
- prompt: chatPrompt,
- llm: chat,
- });
- // then we'lll use that LLM chain as the basis of the agent
- // so basidcally our agent is made up of: Tools, LLM and Prompt Templates, making it highly customizable to our needs!
- const agent = new ZeroShotAgent({
- llmChain,
- allowedTools: tools.map((tool) => tool.name),
- });
- // Now we'll create an Executor instance which allows us to make queries to the agent
-
- const executor = AgentExecutor.fromAgentAndTools({
- agent,
- tools,
- returnIntermediateSteps: false,
- // Max iterations is important - because sometimes our agent can get confused. this can be from a variety of factors such as:
- // not using the right agent for the tool
- // prompts are not perfect – check out the document for crafting the perfect prompt!
- // always set this to a low number to start, or if you're not going to watch the output
- maxIterations: 3,
- // Always set verbose to true, there was a case where I didn't do this and realized my agent was going in loops
- verbose: true,
- });
-
- const result = await executor.run(`Who is ${topic}?`);
-
- return result;
- } catch (err) {
- console.error(err);
- return "Error in completing research";
- }
-};
-export default ResearchAgent;
diff --git a/agents/ResearchAgent.js b/agents/ResearchAgent.js
index ae1dda9..b7fbf06 100644
--- a/agents/ResearchAgent.js
+++ b/agents/ResearchAgent.js
@@ -10,16 +10,72 @@ import { AgentExecutor } from "langchain/agents";
import SerpAPITool from "../tools/SerpAPI";
import WebBrowserTool from "../tools/WebBrowser";
+/**
+ *
+ * WARNING: THIS IS THE SOLUTION! Please try coding before viewing this.
+ *
+ */
const ResearchAgent = async (topic) => {
console.log({ topic });
-
try {
- // do stuff
+ // We'll give it tohe ability to search Google and also Browse the WEb
+ // show the importance of returnDirect
+ const SerpAPI = SerpAPITool();
+ const WebBrowser = WebBrowserTool();
+
+ console.log(SerpAPI.returnDirect);
+ console.log(WebBrowser.returnDirect);
+ // We put them into an array of tools
+ const tools = [SerpAPI, WebBrowser];
+
+ // We'll use the ZeroShotReactDescription which is the recommended tool for chat models
+ // https://js.langchain.com/docs/modules/agents/agents/
+ // The cool thing is that Langchain gives us a method that can create prompts specifically for that agent
+ const promptTemplate = ZeroShotAgent.createPrompt(tools, {
+ prefix: `Answer the following questions as best you can. You have access to the following tools:`,
+ suffix: `Begin! Answer concisely. It's OK to say you don't know.`,
+ });
+
+ // Then we'll initialize it with a Prompt template again
+ const chatPrompt = ChatPromptTemplate.fromPromptMessages([
+ new SystemMessagePromptTemplate(promptTemplate),
+ HumanMessagePromptTemplate.fromTemplate(`{input}`),
+ ]);
+
+ // And we'll initialize the model, what is chatOpenAI since we're using a ChatAgent
+ const chat = new ChatOpenAI({});
+ // We'll create an LLM chain which just a prompt template and a LLM or chat model
+ const llmChain = new LLMChain({
+ prompt: chatPrompt,
+ llm: chat,
+ });
+ // then we'lll use that LLM chain as the basis of the agent
+ // so basidcally our agent is made up of: Tools, LLM and Prompt Templates, making it highly customizable to our needs!
+ const agent = new ZeroShotAgent({
+ llmChain,
+ allowedTools: tools.map((tool) => tool.name),
+ });
+ // Now we'll create an Executor instance which allows us to make queries to the agent
+
+ const executor = AgentExecutor.fromAgentAndTools({
+ agent,
+ tools,
+ returnIntermediateSteps: false,
+ // Max iterations is important - because sometimes our agent can get confused. this can be from a variety of factors such as:
+ // not using the right agent for the tool
+ // prompts are not perfect – check out the document for crafting the perfect prompt!
+ // always set this to a low number to start, or if you're not going to watch the output
+ maxIterations: 3,
+ // Always set verbose to true, there was a case where I didn't do this and realized my agent was going in loops
+ verbose: true,
+ });
+
+ const result = await executor.run(`Who is ${topic}?`);
return result;
} catch (err) {
console.error(err);
+ return "Error in completing research";
}
};
-
export default ResearchAgent;
diff --git a/app/components/Emoji.jsx b/app/components/Emoji.jsx
new file mode 100644
index 0000000..93452a9
--- /dev/null
+++ b/app/components/Emoji.jsx
@@ -0,0 +1,17 @@
+import React from "react";
+
+const Emoji = ({ color }) => {
+ console.log(color);
+ return (
+