Move to subfolder

environments/hack0/DeepSacrifice/server/llm/llm_feedback.ts

@@ -0,0 +1,76 @@
+/**
+ * Placeholder for LLM-based move aggression/brilliance scoring.
+ * Will use OpenAI GPT-4o-mini via callOpenAI in the future.
+ */
+import { callOpenAI } from "./openai_client";
+
+export async function scoreMoveAggression(
+  fenHistory: string[],
+  moveIdx: number,
+  moveSAN: string,
+): Promise<string> {
+  const prompt = `Given the following chess game FEN history (one FEN per move):\n${fenHistory.map((f, i) => `${i + 1}: ${f}`).join("\n")}\nEvaluate the aggression or brilliance of move #${moveIdx + 1} (${moveSAN}). Respond ONLY with a single digit from 1 (not aggressive) to 10 (extremely aggressive/brilliant). Be fast and concise.`;
+  const response = await callOpenAI(
+    [
+      {
+        role: "system",
+        content:
+          "You are a chess grandmaster evaluating move aggression and brilliance. Respond as quickly and concisely as possible.",
+      },
+      { role: "user", content: prompt },
+    ],
+    16,
+  );
+  return response;
+}
+
+/**
+ * Placeholder for LLM-based sacrifice justification.
+ * Will use OpenAI GPT-4o-mini via callOpenAI in the future.
+ */
+export async function justifySacrifice(
+  fenHistory: string[],
+  moveIdx: number,
+  moveSAN: string,
+): Promise<string> {
+  const prompt = `Given the following chess game FEN history (one FEN per move):\n${fenHistory.map((f, i) => `${i + 1}: ${f}`).join("\n")}\nWas the sacrifice in move #${moveIdx + 1} (${moveSAN}) justified? Reply in 1 short sentence: is the sacrifice justified or not, and why. Be fast and concise.`;
+  const response = await callOpenAI(
+    [
+      {
+        role: "system",
+        content:
+          "You are a chess grandmaster evaluating sacrifices. Respond as quickly and concisely as possible.",
+      },
+      { role: "user", content: prompt },
+    ],
+    16,
+  );
+  return response;
+}
+
+export async function scoreAndJustifyGame(
+  fenHistory: string[],
+  moveSANs: string[],
+): Promise<{ score: string; justification: string }[]> {
+  const prompt = `Given the following chess game FEN history (one FEN per move) and the corresponding SAN moves, evaluate each move for aggression/brilliance and sacrifice justification.\n\nFEN history (one per move):\n${fenHistory.map((f, i) => `${i + 1}: ${f}`).join("\n")}\n\nSAN moves (one per move):\n${moveSANs.map((san, i) => `${i + 1}: ${san}`).join("\n")}\n\nFor each move, respond with a JSON array of objects, each with:\n- score: a single digit from 1 (not aggressive) to 10 (extremely aggressive/brilliant)\n- justification: 1 short sentence on whether the move is a justified sacrifice or not, and why.\n\nExample:\n[{"score": "7", "justification": "The sacrifice is risky but justified."}, ...]\n\nRespond ONLY with the JSON array, nothing else.`;
+  const response = await callOpenAI(
+    [
+      {
+        role: "system",
+        content:
+          "You are a chess grandmaster evaluating a full game for aggression and sacrifice justification. Respond as quickly and concisely as possible.",
+      },
+      { role: "user", content: prompt },
+    ],
+    512,
+  );
+  try {
+    const parsed = JSON.parse(response);
+    if (Array.isArray(parsed)) {
+      return parsed;
+    }
+    throw new Error("Response is not an array");
+  } catch (e) {
+    throw new Error("Failed to parse LLM response as JSON: " + response);
+  }
+}

environments/hack0/DeepSacrifice/server/llm/openai_client.ts

@@ -0,0 +1,31 @@
+/**
+ * OpenAI API client for GPT-4o-mini feedback.
+ * Expects to run in Bun/Node.js where process.env is available.
+ * If you see a type error for 'process', install @types/node as a dev dependency.
+ */
+
+import OpenAI from "openai";
+import type { ChatCompletionMessageParam } from "openai/resources/chat/completions";
+
+const OPENAI_API_KEY = process.env.OPENAI_API_KEY || "";
+const MODEL = "gpt-4o-mini"; // Use gpt-4o-mini if available, else gpt-4o
+
+if (!OPENAI_API_KEY) {
+  throw new Error("OPENAI_API_KEY is not set");
+}
+
+const openai = new OpenAI({ apiKey: OPENAI_API_KEY });
+
+export async function callOpenAI(
+  messages: ChatCompletionMessageParam[],
+  max_tokens = 64,
+): Promise<string> {
+  const response = await openai.chat.completions.create({
+    model: MODEL,
+    messages,
+    max_tokens,
+  });
+  const content = response.choices[0]?.message?.content;
+  if (!content) throw new Error("No content returned from OpenAI");
+  return content.trim();
+}