AI Trickshot Puzzle

The Challenge

Pocket the 1, 2, and 9 balls in a single shot.

Controls

Click, drag, and release the Cue Ball to shoot.

\n \n

Ask Chalky

\n \n

If you fail the shot, I'll give you a tip.

\n\n\n\n`;\n}\n\n// --- BACKEND PHYSICS & AI LOGIC ---\nasync function runServerPhysics(\n initialPositions: { number: number; x: number; y: number }[],\n shotParams: { angle: number; power: number },\n) {\n let balls = initialPositions.map(p => ({ ...p, vx: 0, vy: 0, radius: 8, potted: false }));\n let collisionLog: string[] = [];\n const cueBall = balls.find(b => b.number === 0);\n if (cueBall) {\n cueBall.vx = Math.cos(shotParams.angle) * (shotParams.power || 0) * 0.2;\n cueBall.vy = Math.sin(shotParams.angle) * (shotParams.power || 0) * 0.2;\n }\n for (let step = 0; step < 400; step++) {\n let systemMoving = false;\n for (let i = 0; i < balls.length; i++)\n for (let j = i + 1; j < balls.length; j++) {\n const b1 = balls[i], b2 = balls[j];\n const dx = b2.x - b1.x, dy = b2.y - b1.y, dist = Math.sqrt(dx * dx + dy * dy);\n if (dist < b1.radius + b2.radius) {\n collisionLog.push(`Collision: Ball ${b1.number} hit Ball ${b2.number}.`);\n const angle = Math.atan2(dy, dx), sin = Math.sin(angle), cos = Math.cos(angle);\n let v1 = { x: b1.vx * cos + b1.vy * sin, y: b1.vy * cos - b1.vx * sin };\n let v2 = { x: b2.vx * cos + b2.vy * sin, y: b2.vy * cos - b2.vx * sin };\n [v1.x, v2.x] = [v2.x, v1.x];\n b1.vx = v1.x * cos - v1.y * sin;\n b1.vy = v1.y * cos + v1.x * sin;\n b2.vx = v2.x * cos - v2.y * sin;\n b2.vy = v2.y * cos + v2.x * sin;\n }\n }\n balls.forEach(b => {\n b.x += b.vx;\n b.y += b.vy;\n b.vx *= 0.985;\n b.vy *= 0.985;\n if (Math.abs(b.vx) > 0.01 || Math.abs(b.vy) > 0.01) systemMoving = true;\n });\n if (!systemMoving) break;\n }\n return { collisionLog: collisionLog.slice(0, 10) };\n}\n\nexport default async function(req: Request) {\n const url = new URL(req.url);\n if (req.method === \"POST\") {\n try {\n const { initialPositions, shotParams } = await req.json();\n const { collisionLog } = await runServerPhysics(initialPositions, shotParams);\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [\n { role: \"system\", content: TRICKSHOT_TUTOR_PROMPT },\n {\n role: \"user\",\n content:\n `The fixed goal is to pocket the 1, 2, and 9 balls in a single shot. Here is my failed attempt:\\nShot: Power ${\n (shotParams.power || 0).toFixed(0)\n }, Angle ${(shotParams.angle * 180 / Math.PI).toFixed(0)}deg\\nSimulation collision log:\\n${\n collisionLog.join(\"\\n\")\n }`,\n },\n ],\n response_format: { type: \"json_object\" },\n });\n const aiResponse = JSON.parse(completion.choices[0].message.content);\n return new Response(JSON.stringify({ aiResponse }), { headers: { \"Content-Type\": \"application/json\" } });\n } catch (error) {\n return new Response(JSON.stringify({ error: error.message || \"An error occurred.\" }), {\n status: 500,\n headers: { \"Content-Type\": \"application/json\" },\n });\n }\n }\n const sourceUrl = `https://${url.hostname}${url.pathname}`;\n return new Response(generateHtml(sourceUrl), { headers: { \"Content-Type\": \"text/html\" } });\n}","deleted","label","children",[],[99],"files",[149],"counts",{"_173":32,"_174":175,"_176":32,"_177":32},"pulls","branches",1,"forks","envVars","contents",[],"file",{"_103":150,"_95":99,"_151":152,"_153":-5,"_154":137,"_155":136,"_156":157,"_105":182,"_100":159,"_160":161,"_162":163,"_164":22,"_165":-5,"_98":183},["D",1751553151072],[99],"upstreamProject","upstreamBranch","emptyEnvVars","isNew","flags",[],"routes/_app.x.$handle.$name.branch.$branch.history_.$",{"_93":94,"_95":96,"_97":43,"_98":99,"_101":192,"_180":197,"_200":201,"_226":227,"_235":-7,"_236":32,"_237":22},{"_103":104,"_95":96,"_105":193,"_107":-5,"_108":22,"_109":110,"_111":-5,"_112":-5,"_14":113,"_114":-5,"_115":22,"_116":-5,"_117":32,"_16":118,"_119":120,"_121":120,"_122":123,"_124":123,"_125":123,"_126":120,"_127":120,"_23":194,"_131":195,"_134":196,"_146":22},["D",1751546870883],{"_93":94,"_129":130},{"_93":94,"_133":130},{"_100":136,"_95":43,"_103":137,"_138":104,"_139":-5,"_140":-5,"_141":-5,"_142":-5,"_143":-5,"_144":96,"_145":94},{"_103":150,"_95":99,"_151":152,"_153":-5,"_154":137,"_155":136,"_156":157,"_105":198,"_100":159,"_160":161,"_162":163,"_164":22,"_165":-5,"_98":199},["D",1751553151072],[99],"history",[202,204,209,214,219,223],{"_103":150,"_154":137,"_151":152,"_95":99,"_162":163,"_160":161,"_100":159,"_153":-5,"_105":203,"_155":136,"_165":-5},["D",1751553151072],{"_103":205,"_154":137,"_151":152,"_95":99,"_162":206,"_160":161,"_100":207,"_153":-5,"_105":208,"_155":159,"_165":-5},"2e1a1688-581a-11f0-87f0-f69ea79377d9","// @ts-ignore\nimport { OpenAI } from \"https://esm.town/v/std/openai?v=4\";\n\n// --- AI BEHAVIORAL GUIDELINES ---\nconst GENERATION_STYLE_GUIDE =\n \"As the AI Trickshot Tutor, you must adopt the persona of a seasoned, world-class billiards coach. Your name is 'Chalky'. Your tone is calm, analytical, and encouraging. You use professional, precise terminology (e.g., 'cue ball deflection,' 'angle of incidence,' 'follow-through'). You never sound like a generic chatbot. You are direct but supportive, aiming to build the player's understanding of the physics and geometry of the game. Your analysis should be insightful and your suggestions actionable, as if you were standing by the table providing expert guidance.\";\n\nconst TRICKSHOT_TUTOR_PROMPT = `\nYou are 'Chalky', an expert trickshot billiards coach. Your sole purpose is to analyze a player's failed attempt at a trickshot and provide clear, actionable feedback.\n\nThe player has already attempted a shot and seen the result. They are now asking for your expert analysis. You will be provided with:\n1. Their stated goal.\n2. The initial positions of the balls on the table.\n3. The parameters of their shot attempt (cue angle, power).\n4. A log of the key collision events that occurred during the simulation of their shot.\n\nYour task is to analyze this information and determine why the shot failed. You must then provide constructive feedback.\n\nYou MUST respond ONLY with a single, minified JSON object with two keys: \"analysis\" and \"suggestion\".\n- \"analysis\": A concise, expert explanation of what went wrong with the player's shot, based on the collision log. For example, \"The cue ball struck the 1-ball too squarely, transferring most of its energy and leaving it with insufficient momentum to reach the second rail.\"\n- \"suggestion\": A concrete, actionable adjustment the player should make. For example, \"For your next attempt, aim for a finer cut on the 1-ball. Adjust your aim slightly to the left and reduce power to approximately 70% to better control the cue ball's path after contact.\"\n\n${GENERATION_STYLE_GUIDE}\n`;\n\n// --- HTML & FRONTEND ---\nfunction generateHtml(sourceUrl: string) {\n return `\n\n\n \n \n AI Trickshot Tutor (2.5D)\n \n \n \n\n\n\n

AI Trickshot Tutor 2.5D

1. Your Goal

\n \n

2. Shoot!

Click and drag the Cue Ball to aim and set power. Release to shoot.

\n \n

3. Ask Chalky

\n \n

After your shot, I'll analyze it and tell you how to improve.

\n\n\n\n`;\n}\n\n// --- BACKEND PHYSICS & AI LOGIC ---\n// Note: This physics simulation is intentionally simple to run quickly on the server.\n// It is used by the AI to understand the shot dynamics and generate its advice.\nasync function runServerPhysics(\n initialPositions: { number: number; x: number; y: number }[],\n shotParams: { angle: number; power: number },\n) {\n let balls = initialPositions.map(p => ({\n x: p.x,\n y: p.y,\n number: p.number,\n vx: 0,\n vy: 0,\n radius: 8,\n potted: false,\n }));\n let collisionLog: string[] = [];\n const SIMULATION_STEPS = 400;\n const FRICTION = 0.985;\n const POWER_SCALE = 0.2;\n\n // Apply initial shot\n const cueBall = balls.find(b => b.number === 0);\n if (cueBall) {\n const power = shotParams.power || 0;\n const angle = shotParams.angle || 0;\n cueBall.vx = Math.cos(angle) * power * POWER_SCALE;\n cueBall.vy = Math.sin(angle) * power * POWER_SCALE;\n }\n\n for (let step = 0; step < SIMULATION_STEPS; step++) {\n let systemMoving = false;\n // Collision detection\n for (let i = 0; i < balls.length; i++) {\n for (let j = i + 1; j < balls.length; j++) {\n const b1 = balls[i], b2 = balls[j];\n if (b1.potted || b2.potted) continue;\n const dx = b2.x - b1.x, dy = b2.y - b1.y;\n const dist = Math.sqrt(dx * dx + dy * dy);\n if (dist < b1.radius + b2.radius) {\n collisionLog.push(`Collision: Ball ${b1.number} hit Ball ${b2.number}.`);\n const angle = Math.atan2(dy, dx);\n const sin = Math.sin(angle), cos = Math.cos(angle);\n let v1 = { x: b1.vx * cos + b1.vy * sin, y: b1.vy * cos - b1.vx * sin };\n let v2 = { x: b2.vx * cos + b2.vy * sin, y: b2.vy * cos - b2.vx * sin };\n [v1.x, v2.x] = [v2.x, v1.x]; // Swap x-velocities\n b1.vx = v1.x * cos - v1.y * sin;\n b1.vy = v1.y * cos + v1.x * sin;\n b2.vx = v2.x * cos - v2.y * sin;\n b2.vy = v2.y * cos + v2.x * sin;\n }\n }\n }\n\n balls.forEach(b => {\n b.x += b.vx;\n b.y += b.vy;\n b.vx *= FRICTION;\n b.vy *= FRICTION;\n if (Math.abs(b.vx) > 0.01 || Math.abs(b.vy) > 0.01) systemMoving = true;\n });\n\n if (!systemMoving) break;\n }\n return { collisionLog: collisionLog.slice(0, 10) }; // Limit log length\n}\n\nexport default async function(req: Request) {\n const url = new URL(req.url);\n\n if (req.method === \"POST\") {\n try {\n const { goal, initialPositions, shotParams } = await req.json();\n if (!goal || !initialPositions || !shotParams) {\n throw new Error(\"Missing required parameters.\");\n }\n\n // Run a quick simulation on the server to generate a collision log for the AI\n const { collisionLog } = await runServerPhysics(initialPositions, shotParams);\n\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [\n { role: \"system\", content: TRICKSHOT_TUTOR_PROMPT },\n {\n role: \"user\",\n content: `My goal: \"${goal}\"\\nInitial ball positions: ${\n JSON.stringify(initialPositions.map(p => p.number))\n }\\nMy shot: Power ${shotParams.power.toFixed(0)}, Angle ${\n (shotParams.angle * 180 / Math.PI).toFixed(0)\n }deg\\nSimulation collision log: \\n${collisionLog.join(\"\\n\")}`,\n },\n ],\n response_format: { type: \"json_object\" },\n });\n\n const aiResponse = JSON.parse(completion.choices[0].message.content);\n\n return new Response(JSON.stringify({ aiResponse }), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n } catch (error) {\n console.error(\"AI Tutor Error:\", error);\n return new Response(JSON.stringify({ error: error.message || \"An unknown error occurred.\" }), {\n status: 500,\n headers: { \"Content-Type\": \"application/json\" },\n });\n }\n }\n\n // Serve HTML for GET requests\n const sourceUrl = `https://${url.hostname}${url.pathname}`;\n return new Response(generateHtml(sourceUrl), { headers: { \"Content-Type\": \"text/html\" } });\n}",4,["D",1751552991422],{"_103":210,"_154":137,"_151":152,"_95":99,"_162":211,"_160":161,"_100":212,"_153":-5,"_105":213,"_155":207,"_165":-5},"33bbad30-580c-11f0-9290-f69ea79377d9","// @ts-ignore\nimport { OpenAI } from \"https://esm.town/v/std/openai?v=4\";\n\n// --- AI BEHAVIORAL GUIDELINES ---\nconst GENERATION_STYLE_GUIDE =\n \"As the AI Trickshot Tutor, you must adopt the persona of a seasoned, world-class billiards coach. Your name is 'Chalky'. Your tone is calm, analytical, and encouraging. You use professional, precise terminology (e.g., 'cue ball deflection,' 'angle of incidence,' 'follow-through'). You never sound like a generic chatbot. You are direct but supportive, aiming to build the player's understanding of the physics and geometry of the game. Your analysis should be insightful and your suggestions actionable, as if you were standing by the table providing expert guidance.\";\n\nconst TRICKSHOT_TUTOR_PROMPT = `\nYou are 'Chalky', an expert trickshot billiards coach. Your sole purpose is to analyze a player's failed attempt at a trickshot and provide clear, actionable feedback.\n\nThe player will provide you with:\n1. Their stated goal.\n2. The initial positions of the balls on the table.\n3. The parameters of their shot attempt (cue angle, power).\n4. A log of the key collision events that occurred during the simulation of their shot.\n\nYour task is to analyze this information and determine why the shot failed. You must then provide constructive feedback.\n\nYou MUST respond ONLY with a single, minified JSON object with three keys: \"analysis\", \"suggestion\", and \"reason\".\n- \"analysis\": A concise, expert explanation of what went wrong with the player's shot, based on the collision log. For example, \"The cue ball struck the 1-ball too squarely, transferring most of its energy and leaving it with insufficient momentum to reach the second rail.\"\n- \"suggestion\": A concrete, actionable adjustment the player should make. For example, \"Try aiming for a finer cut on the 1-ball. Adjust your cue angle by approximately 5 degrees to the left and reduce power to 70% to account for the altered trajectory.\"\n- \"reason\": A justification for your suggestion, explaining the underlying physics in a simple way. For example, \"This finer contact will impart more spin and redirect the cue ball with the necessary speed and angle to complete the three-rail journey to the 9-ball.\"\n\n${GENERATION_STYLE_GUIDE}\n`;\n\n// --- HTML & FRONTEND ---\nfunction generateHtml(sourceUrl: string) {\n return `\n\n\n \n \n AI Trickshot Tutor\n \n \n \n\n\n\n

AI Trickshot Tutor

Setup

\n Your Goal:\n \n

\n \n

Shot Control

\n Power: 50\n \n

\n \n

Chalky's Analysis

Set up the balls and describe your goal. Then, take a shot. I'll tell you what went right, what went wrong, and how to improve.

\n\n\n\n`;\n}\n\n// --- BACKEND PHYSICS & AI LOGIC ---\n\nclass Vec2 {\n x: number;\n y: number;\n constructor(x = 0, y = 0) {\n this.x = x;\n this.y = y;\n }\n add(v: Vec2) {\n return new Vec2(this.x + v.x, this.y + v.y);\n }\n sub(v: Vec2) {\n return new Vec2(this.x - v.x, this.y - v.y);\n }\n mul(s: number) {\n return new Vec2(this.x * s, this.y * s);\n }\n get length() {\n return Math.sqrt(this.x * this.x + this.y * this.y);\n }\n normalize() {\n const len = this.length;\n return len > 0 ? new Vec2(this.x / len, this.y / len) : new Vec2();\n }\n dot(v: Vec2) {\n return this.x * v.x + this.y * v.y;\n }\n}\n\nclass Ball {\n pos: Vec2;\n vel: Vec2;\n radius: number;\n mass: number;\n number: number;\n potted: boolean = false;\n\n constructor(number: number, x: number, y: number) {\n this.number = number;\n this.pos = new Vec2(x, y);\n this.vel = new Vec2();\n this.radius = 8;\n this.mass = 1;\n }\n}\n\nasync function runPhysicsSimulation(\n initialPositions: { number: number; x: number; y: number }[],\n shotParams: { angle: number; power: number },\n) {\n const TABLE_WIDTH = 600;\n const TABLE_HEIGHT = 300;\n const POCKET_RADIUS = 12;\n const FRICTION = 0.985;\n const SIMULATION_STEPS = 500;\n\n let balls: Ball[] = initialPositions.map(p => new Ball(p.number, p.x, p.y));\n let collisionLog: string[] = [];\n const simulationFrames: { x: number; y: number; potted: boolean }[][] = [];\n\n const pockets = [\n new Vec2(0, 0),\n new Vec2(TABLE_WIDTH / 2, 0),\n new Vec2(TABLE_WIDTH, 0),\n new Vec2(0, TABLE_HEIGHT),\n new Vec2(TABLE_WIDTH / 2, TABLE_HEIGHT),\n new Vec2(TABLE_WIDTH, TABLE_HEIGHT),\n ];\n\n // Apply initial shot\n const cueBall = balls.find(b => b.number === 0);\n if (cueBall) {\n const force = shotParams.power / 10;\n cueBall.vel = new Vec2(\n Math.cos(shotParams.angle) * force,\n Math.sin(shotParams.angle) * force,\n );\n }\n\n for (let step = 0; step < SIMULATION_STEPS; step++) {\n // Check for collisions\n for (let i = 0; i < balls.length; i++) {\n for (let j = i + 1; j < balls.length; j++) {\n const b1 = balls[i];\n const b2 = balls[j];\n if (b1.potted || b2.potted) continue;\n\n const distVec = b1.pos.sub(b2.pos);\n const dist = distVec.length;\n const min_dist = b1.radius + b2.radius;\n\n if (dist < min_dist) {\n collisionLog.push(`Collision between Ball ${b1.number} and Ball ${b2.number}.`);\n const normal = distVec.normalize();\n const tangent = new Vec2(-normal.y, normal.x);\n const overlap = min_dist - dist;\n b1.pos = b1.pos.add(normal.mul(overlap / 2));\n b2.pos = b2.pos.sub(normal.mul(overlap / 2));\n\n const v1n = b1.vel.dot(normal);\n const v1t = b1.vel.dot(tangent);\n const v2n = b2.vel.dot(normal);\n const v2t = b2.vel.dot(tangent);\n\n const v1n_final = (v1n * (b1.mass - b2.mass) + 2 * b2.mass * v2n) / (b1.mass + b2.mass);\n const v2n_final = (v2n * (b2.mass - b1.mass) + 2 * b1.mass * v1n) / (b1.mass + b2.mass);\n\n const v1n_vec = normal.mul(v1n_final);\n const v1t_vec = tangent.mul(v1t);\n const v2n_vec = normal.mul(v2n_final);\n const v2t_vec = tangent.mul(v2t);\n\n b1.vel = v1n_vec.add(v1t_vec);\n b2.vel = v2n_vec.add(v2t_vec);\n }\n }\n }\n\n let systemMoving = false;\n // Update positions and handle cushions/pockets\n balls.forEach(b => {\n if (b.potted) return;\n\n // Pocket check\n for (const pocket of pockets) {\n if (b.pos.sub(pocket).length < POCKET_RADIUS) {\n b.potted = true;\n b.vel = new Vec2();\n collisionLog.push(`Ball ${b.number} was pocketed.`);\n break;\n }\n }\n if (b.potted) return;\n\n // Cushion collisions\n if (b.pos.x - b.radius < 0 || b.pos.x + b.radius > TABLE_WIDTH) {\n b.vel.x *= -1;\n b.pos.x = b.pos.x - b.radius < 0 ? b.radius : TABLE_WIDTH - b.radius;\n collisionLog.push(`Ball ${b.number} hit a vertical cushion.`);\n }\n if (b.pos.y - b.radius < 0 || b.pos.y + b.radius > TABLE_HEIGHT) {\n b.vel.y *= -1;\n b.pos.y = b.pos.y - b.radius < 0 ? b.radius : TABLE_HEIGHT - b.radius;\n collisionLog.push(`Ball ${b.number} hit a horizontal cushion.`);\n }\n\n b.vel = b.vel.mul(FRICTION);\n if (b.vel.length < 0.1) b.vel = new Vec2();\n else systemMoving = true;\n\n b.pos = b.pos.add(b.vel);\n });\n\n simulationFrames.push(balls.map(b => ({ x: b.pos.x, y: b.pos.y, potted: b.potted })));\n\n if (!systemMoving) break; // End early if nothing is moving\n }\n\n return { simulation: simulationFrames, log: collisionLog };\n}\n\nexport default async function(req: Request) {\n const url = new URL(req.url);\n\n if (req.method === \"POST\") {\n try {\n const { goal, initialPositions, shotParams } = await req.json();\n\n if (!goal || !initialPositions || !shotParams) {\n throw new Error(\"Missing required parameters: goal, initialPositions, or shotParams.\");\n }\n\n const { simulation, log: collisionLog } = await runPhysicsSimulation(initialPositions, shotParams);\n\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [\n { role: \"system\", content: TRICKSHOT_TUTOR_PROMPT },\n {\n role: \"user\",\n content: `My goal: \"${goal}\"\\nInitial ball positions: ${JSON.stringify(initialPositions)}\\nMy shot: ${\n JSON.stringify(shotParams)\n }\\nSimulation collision log: \\n${collisionLog.slice(0, 10).join(\"\\n\")}`,\n }, // Limit log length\n ],\n response_format: { type: \"json_object\" },\n });\n\n const aiResponse = JSON.parse(completion.choices[0].message.content);\n\n return new Response(JSON.stringify({ simulation, aiResponse, collisionLog }), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n } catch (error) {\n console.error(\"AI Tutor Error:\", error);\n return new Response(JSON.stringify({ error: error.message || \"An unknown error occurred.\" }), {\n status: 500,\n headers: { \"Content-Type\": \"application/json\" },\n });\n }\n }\n\n // Serve HTML for GET requests\n const sourceUrl = `https://${url.hostname}${url.pathname}`;\n return new Response(generateHtml(sourceUrl), { headers: { \"Content-Type\": \"text/html\" } });\n}",3,["D",1751546987916],{"_103":215,"_154":137,"_151":152,"_95":99,"_162":211,"_160":216,"_100":217,"_153":-5,"_105":218,"_155":212,"_165":-5},"291be21e-580c-11f0-9031-f69ea79377d9","script",2,["D",1751546970093],{"_103":220,"_154":137,"_151":152,"_95":99,"_162":221,"_160":216,"_100":175,"_153":-5,"_105":222,"_155":217,"_165":-5},"041a5bf8-580c-11f0-9d08-f69ea79377d9","// @ts-ignore\nimport { OpenAI } from \"https://esm.town/v/std/openai?v=4\";\n\n// --- AI BEHAVIORAL GUIDELINES ---\nconst GENERATION_STYLE_GUIDE =\n \"As the AI Trickshot Tutor, you must adopt the persona of a seasoned, world-class billiards coach. Your name is 'Chalky'. Your tone is calm, analytical, and encouraging. You use professional, precise terminology (e.g., 'cue ball deflection,' 'angle of incidence,' 'follow-through'). You never sound like a generic chatbot. You are direct but supportive, aiming to build the player's understanding of the physics and geometry of the game. Your analysis should be insightful and your suggestions actionable, as if you were standing by the table providing expert guidance.\";\n\nconst TRICKSHOT_TUTOR_PROMPT = `\nYou are 'Chalky', an expert trickshot billiards coach. Your sole purpose is to analyze a player's failed attempt at a trickshot and provide clear, actionable feedback.\n\nThe player will provide you with:\n1. Their stated goal.\n2. The initial positions of the balls on the table.\n3. The parameters of their shot attempt (cue angle, power).\n4. A log of the key collision events that occurred during the simulation of their shot.\n\nYour task is to analyze this information and determine why the shot failed. You must then provide constructive feedback.\n\nYou MUST respond ONLY with a single, minified JSON object with three keys: \"analysis\", \"suggestion\", and \"reason\".\n- \"analysis\": A concise, expert explanation of what went wrong with the player's shot, based on the collision log. For example, \"The cue ball struck the 1-ball too squarely, transferring most of its energy and leaving it with insufficient momentum to reach the second rail.\"\n- \"suggestion\": A concrete, actionable adjustment the player should make. For example, \"Try aiming for a finer cut on the 1-ball. Adjust your cue angle by approximately 5 degrees to the left and reduce power to 70% to account for the altered trajectory.\"\n- \"reason\": A justification for your suggestion, explaining the underlying physics in a simple way. For example, \"This finer contact will impart more spin and redirect the cue ball with the necessary speed and angle to complete the three-rail journey to the 9-ball.\"\n\n${GENERATION_STYLE_GUIDE}\n`;\n\n// --- HTML & FRONTEND ---\nfunction generateHtml(sourceUrl: string) {\n return `\n\n\n \n \n AI Trickshot Tutor\n \n \n \n\n\n\n

AI Trickshot Tutor

Setup

\n Your Goal:\n \n

\n \n

Shot Control

\n Power: 50\n \n

\n \n

Chalky's Analysis

Set up the balls and describe your goal. Then, take a shot. I'll tell you what went right, what went wrong, and how to improve.

\n\n\n\n`;\n}\n\n// --- BACKEND PHYSICS & AI LOGIC ---\n\nclass Vec2 {\n x: number;\n y: number;\n constructor(x = 0, y = 0) {\n this.x = x;\n this.y = y;\n }\n add(v: Vec2) {\n return new Vec2(this.x + v.x, this.y + v.y);\n }\n sub(v: Vec2) {\n return new Vec2(this.x - v.x, this.y - v.y);\n }\n mul(s: number) {\n return new Vec2(this.x * s, this.y * s);\n }\n get length() {\n return Math.sqrt(this.x * this.x + this.y * this.y);\n }\n normalize() {\n const len = this.length;\n return len > 0 ? new Vec2(this.x / len, this.y / len) : new Vec2();\n }\n dot(v: Vec2) {\n return this.x * v.x + this.y * v.y;\n }\n}\n\nclass Ball {\n pos: Vec2;\n vel: Vec2;\n radius: number;\n mass: number;\n number: number;\n potted: boolean = false;\n\n constructor(number: number, x: number, y: number) {\n this.number = number;\n this.pos = new Vec2(x, y);\n this.vel = new Vec2();\n this.radius = 8;\n this.mass = 1;\n }\n}\n\nasync function runPhysicsSimulation(\n initialPositions: { number: number; x: number; y: number }[],\n shotParams: { angle: number; power: number },\n) {\n const TABLE_WIDTH = 600;\n const TABLE_HEIGHT = 300;\n const POCKET_RADIUS = 12;\n const FRICTION = 0.985;\n const SIMULATION_STEPS = 500;\n\n let balls: Ball[] = initialPositions.map(p => new Ball(p.number, p.x, p.y));\n let collisionLog: string[] = [];\n const simulationFrames: { x: number; y: number; potted: boolean }[][] = [];\n\n const pockets = [\n new Vec2(0, 0), new Vec2(TABLE_WIDTH / 2, 0), new Vec2(TABLE_WIDTH, 0),\n new Vec2(0, TABLE_HEIGHT), new Vec2(TABLE_WIDTH / 2, TABLE_HEIGHT), new Vec2(TABLE_WIDTH, TABLE_HEIGHT),\n ];\n\n // Apply initial shot\n const cueBall = balls.find(b => b.number === 0);\n if (cueBall) {\n const force = shotParams.power / 10;\n cueBall.vel = new Vec2(\n Math.cos(shotParams.angle) * force,\n Math.sin(shotParams.angle) * force,\n );\n }\n\n for (let step = 0; step < SIMULATION_STEPS; step++) {\n // Check for collisions\n for (let i = 0; i < balls.length; i++) {\n for (let j = i + 1; j < balls.length; j++) {\n const b1 = balls[i];\n const b2 = balls[j];\n if (b1.potted || b2.potted) continue;\n\n const distVec = b1.pos.sub(b2.pos);\n const dist = distVec.length;\n const min_dist = b1.radius + b2.radius;\n\n if (dist < min_dist) {\n collisionLog.push(`Collision between Ball ${b1.number} and Ball ${b2.number}.`);\n const normal = distVec.normalize();\n const tangent = new Vec2(-normal.y, normal.x);\n const overlap = min_dist - dist;\n b1.pos = b1.pos.add(normal.mul(overlap / 2));\n b2.pos = b2.pos.sub(normal.mul(overlap / 2));\n \n const v1n = b1.vel.dot(normal);\n const v1t = b1.vel.dot(tangent);\n const v2n = b2.vel.dot(normal);\n const v2t = b2.vel.dot(tangent);\n\n const v1n_final = (v1n * (b1.mass - b2.mass) + 2 * b2.mass * v2n) / (b1.mass + b2.mass);\n const v2n_final = (v2n * (b2.mass - b1.mass) + 2 * b1.mass * v1n) / (b1.mass + b2.mass);\n\n const v1n_vec = normal.mul(v1n_final);\n const v1t_vec = tangent.mul(v1t);\n const v2n_vec = normal.mul(v2n_final);\n const v2t_vec = tangent.mul(v2t);\n\n b1.vel = v1n_vec.add(v1t_vec);\n b2.vel = v2n_vec.add(v2t_vec);\n }\n }\n }\n \n let systemMoving = false;\n // Update positions and handle cushions/pockets\n balls.forEach(b => {\n if (b.potted) return;\n \n // Pocket check\n for(const pocket of pockets) {\n if (b.pos.sub(pocket).length < POCKET_RADIUS) {\n b.potted = true;\n b.vel = new Vec2();\n collisionLog.push(`Ball ${b.number} was pocketed.`);\n break;\n }\n }\n if (b.potted) return;\n\n // Cushion collisions\n if (b.pos.x - b.radius < 0 || b.pos.x + b.radius > TABLE_WIDTH) {\n b.vel.x *= -1;\n b.pos.x = b.pos.x - b.radius < 0 ? b.radius : TABLE_WIDTH - b.radius;\n collisionLog.push(`Ball ${b.number} hit a vertical cushion.`);\n }\n if (b.pos.y - b.radius < 0 || b.pos.y + b.radius > TABLE_HEIGHT) {\n b.vel.y *= -1;\n b.pos.y = b.pos.y - b.radius < 0 ? b.radius : TABLE_HEIGHT - b.radius;\n collisionLog.push(`Ball ${b.number} hit a horizontal cushion.`);\n }\n\n b.vel = b.vel.mul(FRICTION);\n if (b.vel.length < 0.1) b.vel = new Vec2();\n else systemMoving = true;\n \n b.pos = b.pos.add(b.vel);\n });\n\n simulationFrames.push(balls.map(b => ({ x: b.pos.x, y: b.pos.y, potted: b.potted })));\n\n if (!systemMoving) break; // End early if nothing is moving\n }\n\n return { simulation: simulationFrames, log: collisionLog };\n}\n\nexport default async function(req: Request) {\n const url = new URL(req.url);\n \n if (req.method === \"POST\") {\n try {\n const { goal, initialPositions, shotParams } = await req.json();\n\n if (!goal || !initialPositions || !shotParams) {\n throw new Error(\"Missing required parameters: goal, initialPositions, or shotParams.\");\n }\n\n const { simulation, log: collisionLog } = await runPhysicsSimulation(initialPositions, shotParams);\n \n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o\",\n messages: [\n { role: \"system\", content: TRICKSHOT_TUTOR_PROMPT },\n { role: \"user\", content: `My goal: \"${goal}\"\\nInitial ball positions: ${JSON.stringify(initialPositions)}\\nMy shot: ${JSON.stringify(shotParams)}\\nSimulation collision log: \\n${collisionLog.slice(0, 10).join('\\n')}` }, // Limit log length\n ],\n response_format: { type: \"json_object\" },\n });\n \n const aiResponse = JSON.parse(completion.choices[0].message.content);\n\n return new Response(JSON.stringify({ simulation, aiResponse, collisionLog }), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n\n } catch (error) {\n console.error(\"AI Tutor Error:\", error);\n return new Response(JSON.stringify({ error: error.message || \"An unknown error occurred.\" }), {\n status: 500,\n headers: { \"Content-Type\": \"application/json\" },\n });\n }\n }\n\n // Serve HTML for GET requests\n const sourceUrl = `https://${url.hostname}${url.pathname}`;\n return new Response(generateHtml(sourceUrl), { headers: { \"Content-Type\": \"text/html\" } });\n}",["D",1751546907994],{"_103":224,"_154":137,"_151":152,"_95":99,"_162":72,"_160":216,"_100":32,"_153":-5,"_105":225,"_155":175,"_165":-5},"edfb2154-580b-11f0-9290-f69ea79377d9",["D",1751546870896],"versions",[228,230,231,232,233,234],{"_180":202,"_229":204},"previous",{"_180":204,"_229":209},{"_180":209,"_229":214},{"_180":214,"_229":219},{"_180":219,"_229":223},{"_180":223,"_229":-5},"versionNumber","page","hasNext","actionData","errors"]

History for main.tsx