feat: Add a planning summary step in planner subgraph (#13)

src/subgraphs/planner/index.ts

@@ -8,7 +8,12 @@
 import { END, Send, START, StateGraph } from "@langchain/langgraph";
 import { PlannerGraphState, PlannerGraphStateObj } from "./types.js";
 import { GraphConfiguration } from "../../types.js";
-import { generateAction, generatePlan, takeAction } from "./nodes/index.js";
+import {
+  generateAction,
+  generatePlan,
+  summarizer,
+  takeAction,
+} from "./nodes/index.js";
 import { isAIMessage, ToolMessage } from "@langchain/core/messages";
 
 function takeActionOrGeneratePlan(
@@ -52,6 +57,7 @@ const workflow = new StateGraph(PlannerGraphStateObj, GraphConfiguration)
   .addNode("generate-plan-context-action", generateAction)
   .addNode("take-plan-action", takeAction)
   .addNode("generate-plan", generatePlan)
+  .addNode("summarizer", summarizer)
   .addEdge(START, "generate-plan-context-action")
   .addConditionalEdges(
     "generate-plan-context-action",
@@ -59,7 +65,8 @@ const workflow = new StateGraph(PlannerGraphStateObj, GraphConfiguration)
     ["take-plan-action", "generate-plan"],
   )
   .addEdge("take-plan-action", "generate-plan-context-action")
-  .addEdge("generate-plan", END);
+  .addEdge("generate-plan", "summarizer")
+  .addEdge("summarizer", END);
 
 // TODO: Fix zod types
 export const plannerGraph = workflow.compile() as any;

src/subgraphs/planner/nodes/index.ts

@@ -1,3 +1,4 @@
 export * from "./generate-message.js";
 export * from "./take-action.js";
 export * from "./generate-plan.js";
+export * from "./summarizer.js";

src/subgraphs/planner/nodes/summarizer.ts

@@ -0,0 +1,78 @@
+import { z } from "zod";
+import { GraphConfig } from "../../../types.js";
+import { PlannerGraphState, PlannerGraphUpdate } from "../types.js";
+import { loadModel, Task } from "../../../utils/load-model.js";
+import { isHumanMessage, ToolMessage } from "@langchain/core/messages";
+import { getMessageContentString } from "../../../utils/message-content.js";
+
+const systemPrompt = `You are operating as a terminal-based agentic coding assistant built by LangChain. It wraps LLM models to enable natural language interaction with a local codebase. You are expected to be precise, safe, and helpful.
+
+You've been given a task to summarize the messages in your conversation history. You just finished gathering context to be used when generating an development plan to address the user's request.
+You do not want to keep the entire conversation history, but instead you want to keep the most relevant and important snippets for future context
+
+You MUST adhere to the following criteria when summarizing the conversation history:
+- Retain context such as file paths, versions, and installed software.
+- Do not retain any full code snippets.
+- Do not retain any full file contents.
+- Ensure your summary is concise, but useful for future context.
+
+Here is the user's initial request
+## User request:
+{USER_REQUEST}
+
+With all of this in mind, please carefully summarize and condense the following conversation history. Ensure you pass this condensed context to the \`condense_planning_context\` tool.
+`;
+
+const formatPrompt = (userRequest: string): string =>
+  systemPrompt.replace("{USER_REQUEST}", userRequest);
+
+const condenseContextToolSchema = z.object({
+  context: z
+    .string()
+    .describe("The condensed context to be used when generating a plan."),
+});
+const condenseContextTool = {
+  name: "condense_planning_context",
+  description:
+    "Condense the conversation history into a concise summary, while still retaining the most relevant and important snippets.",
+  schema: condenseContextToolSchema,
+};
+
+export async function summarizer(
+  state: PlannerGraphState,
+  config: GraphConfig,
+): Promise<PlannerGraphUpdate> {
+  const model = await loadModel(config, Task.PLANNER);
+  const modelWithTools = model.bindTools([condenseContextTool], {
+    tool_choice: condenseContextTool.name,
+  });
+
+  const firstUserMessage = state.messages.find(isHumanMessage);
+
+  const response = await modelWithTools.invoke([
+    {
+      role: "system",
+      content: formatPrompt(
+        getMessageContentString(
+          firstUserMessage?.content ?? "No user request provided.",
+        ),
+      ),
+    },
+    ...state.plannerMessages,
+  ]);
+
+  const toolCall = response.tool_calls?.[0];
+  if (!toolCall) {
+    throw new Error("Failed to generate plan");
+  }
+
+  const toolMessage = new ToolMessage({
+    tool_call_id: toolCall.id ?? "",
+    name: toolCall.name,
+    content: `Successfully summarized planning context.`,
+  });
+
+  return {
+    messages: [response, toolMessage],
+  };
+}

src/utils/message-content.ts

@@ -0,0 +1,10 @@
+import { MessageContent } from "@langchain/core/messages";
+
+export function getMessageContentString(content: MessageContent): string {
+  if (typeof content === "string") return content;
+
+  return content
+    .filter((c): c is { type: "text"; text: string } => c.type === "text")
+    .map((c) => c.text)
+    .join(" ");
+}