Allow multiple source CSV files for normalisation. Implement real model node. Add normalizarion prompt. Implement normalization setup. Start on RAG retreival functions

agent/nodes/model.ts

@@ -1,24 +1,27 @@
-// import { SystemMessage } from "@langchain/core/messages";
-// import { GraphNode } from "@langchain/langgraph";
-// import { MessagesState } from "../state";
-// import { arithmeticTools } from "../tools/arithmetic";
-// import { ChatOpenAI } from "@langchain/openai"
+import { HumanMessage, SystemMessage } from "@langchain/core/messages";
+import { GraphNode } from "@langchain/langgraph";
+import { MessagesState } from "../state";
+import { ChatOpenAI } from "@langchain/openai"
+import { hydratePrompt } from "../prompts/hydratePrompt";
 
-// const model = new ChatOpenAI({
-//   model: "gpt-5-mini"
-// });
+export function createModelNode(tools: any, promptPath: string): GraphNode<typeof MessagesState> {
+    return async (state) => {
+        const sysPrompt = hydratePrompt(promptPath, state.disinformationTitle)
 
-// const modelWithTools = model.bindTools(arithmeticTools);
+        const model = new ChatOpenAI({
+            model: "gpt-5-mini"
+        });
+        const modelWithTools = model.bindTools(tools);
 
-// export const llmCall: GraphNode<typeof MessagesState> = async (state) => {
-//   const response = await modelWithTools.invoke([
-//     new SystemMessage(
-//       "You are a helpful assistant tasked with performing arithmetic on a set of inputs. Any calculation, no matter how trivial, should be done with tools.  Output the final answer with %%% on each side"
-//     ),
-//     ...state.messages,
-//   ]);
-//   return {
-//     messages: [response],
-//     llmCalls: 1,
-//   };
-// };
+        const response = await modelWithTools.invoke([
+            new SystemMessage(
+                sysPrompt
+            ),
+            ...state.messages,
+        ]);
+
+        return {
+            messages: [response]
+        };
+    };
+}