# { "Depends": "py-genlayer:1jb45aa8ynh2a9c9xn3b7qqh8sm5q93hwfp7jqmwsfhh8jpz09h6" } import json from genlayer import * class RallyCheck(gl.Contract): """ RallyCheck -- Intelligent Contract on GenLayer. Predicts the score a tweet would receive in a Rally.fun campaign BEFORE the user submits it. Helps creators optimize tweets for maximum earnings. Architecture: 1. User provides tweet URL + Rally campaign URL 2. Frontend fetches tweet data (twitterapi.io) 3. Contract scrapes campaign rules live from rally.fun 4. Contract scores tweet against campaign criteria 5. Returns predicted score + improvement suggestions """ results: TreeMap[str, str] def __init__(self): pass @gl.public.write def predict_score( self, tweet_url: str, campaign_url: str, api_data: str, ) -> None: # LAYER 1: Fetch live campaign rules from rally.fun def fetch_campaign_rules() -> str: try: page_content = gl.nondet.web.render( campaign_url, mode="text" ) page_content = page_content[:5000] if page_content else "" extract_prompt = f""" You are extracting campaign rules from a Rally.fun campaign page. PAGE CONTENT: {page_content} Extract campaign rules into JSON. Use defaults if fields not visible. Return ONLY this JSON, no markdown: {{ "campaign_name": "", "project": "", "topic": "", "required_hashtags": ["<#tag1>"], "required_mentions": ["<@handle1>"], "content_guidelines": "", "min_sorsa_score": , "verified_only": , "old_tweets_allowed": }} """ response = gl.nondet.exec_prompt(extract_prompt) response = response.strip() if response.startswith("```"): response = response.split("\n", 1)[-1] if response.endswith("```"): response = response.rsplit("```", 1)[0] return response.strip() except Exception as e: return json.dumps({ "campaign_name": "unknown", "project": "unknown", "topic": "unknown", "required_hashtags": [], "required_mentions": [], "content_guidelines": "follow general Rally guidelines", "min_sorsa_score": 10, "verified_only": False, "old_tweets_allowed": False, "fetch_error": str(e)[:100], }) campaign_rules = gl.eq_principle.prompt_comparative( fetch_campaign_rules, principle=( "Both outputs should agree on campaign topic, project, " "and required hashtags/mentions. Other fields can vary." ), ) # LAYER 2: Score tweet against campaign criteria def evaluate() -> str: prompt = f""" You are a Rally.fun campaign scoring oracle. Predict how Rally's AI will score a tweet against the active campaign criteria. Score on three tiers (each 0.0 - 1.0): CORE GATES (must pass): - Content Alignment: tweet matches campaign topic? - Information Accuracy: facts in tweet correct? - Campaign Compliance: follows campaign rules (hashtags, mentions)? - Originality and Authenticity: original, not copy-paste? QUALITY METRICS: - Engagement Potential: will it drive real discussion? - Technical Quality: well-written, clear, formatted? ENGAGEMENT METRICS (predict potential): - Retweets, Likes, Replies, Quality of Replies, Followers of Repliers, Impressions =================================== TWEET URL: {tweet_url} =================================== TWEET DATA (twitterapi.io): {api_data} =================================== CAMPAIGN URL: {campaign_url} CAMPAIGN RULES (live from rally.fun): {campaign_rules} =================================== SCORING: - Content Alignment: 1.0 if directly addresses campaign topic, 0.0 if off-topic - Information Accuracy: 1.0 if facts correct, lower if unverified/wrong - Campaign Compliance: 1.0 if all required hashtags/mentions present - Originality: 1.0 if unique perspective, 0.0 if obvious copy-paste - Engagement Potential: hook quality, controversy, value (0-1.0) - Technical Quality: clarity, grammar, formatting, length (0-1.0) - Engagement metrics: based on actual viewCount, likeCount, etc. For unposted tweets (low metrics), predict based on content quality. Return ONLY this JSON, no markdown: {{ "predicted_total_score": <0.0 to 10.0>, "verdict": "QUALIFIES | BORDERLINE | DOES_NOT_QUALIFY", "summary": "", "core_gates": {{ "content_alignment": <0.0-1.0>, "information_accuracy": <0.0-1.0>, "campaign_compliance": <0.0-1.0>, "originality_authenticity": <0.0-1.0> }}, "quality_metrics": {{ "engagement_potential": <0.0-1.0>, "technical_quality": <0.0-1.0> }}, "engagement_metrics": {{ "retweets": <0.0-1.0>, "likes": <0.0-1.0>, "replies": <0.0-1.0>, "quality_of_replies": <0.0-1.0>, "followers_of_repliers": <0.0-1.0>, "impressions": <0.0-1.0> }}, "improvement_suggestions": [ "", "", "" ], "campaign_info": {{ "campaign_name": "", "project": "", "matched_hashtags": [""], "missing_hashtags": [""], "matched_mentions": [""], "missing_mentions": [""] }} }} KEY RULES: - Required hashtags missing -> campaign_compliance = 0.0 - Tweet off-topic -> content_alignment = 0.0 - Generic tweet (just "gm", "lfg") -> all core gates low - Originality high for unique take, low for copy-paste """ response = gl.nondet.exec_prompt(prompt) response = response.strip() if response.startswith("```"): response = response.split("\n", 1)[-1] if response.endswith("```"): response = response.rsplit("```", 1)[0] return response.strip() verdict = gl.eq_principle.prompt_comparative( evaluate, principle=( "Both outputs should agree on the verdict (QUALIFIES / " "BORDERLINE / DOES_NOT_QUALIFY). Individual metric scores " "should be within 0.2 of each other. Suggestions can vary " "in wording." ), ) key = f"{tweet_url}::{campaign_url}" self.results[key] = verdict @gl.public.view def get_result(self, tweet_url: str, campaign_url: str) -> str: key = f"{tweet_url}::{campaign_url}" return self.results.get(key, "Not analyzed yet") @gl.public.view def get_all_results(self) -> dict[str, str]: return {k: v for k, v in self.results.items()} @gl.public.view def has_result(self, tweet_url: str, campaign_url: str) -> bool: key = f"{tweet_url}::{campaign_url}" return key in self.results