use chrono::{Datelike, Timelike, Utc};
use rusqlite::{params, Connection, OptionalExtension};
use serde::{Deserialize, Serialize};
use std::path::PathBuf;
use crate::config::config_dir;
#[derive(Debug, Clone, Serialize)]
pub struct DiscoveredPostRow {
pub tweet_id: String,
pub author_username: String,
pub text: String,
pub like_count: i64,
pub impression_count: i64,
pub last_source: String,
pub first_discovered_at: i64,
}
#[derive(Debug, Clone, Serialize)]
pub struct WatchlistEntry {
pub username: String,
pub user_id: Option<String>,
pub topic: Option<String>,
pub followers: i64,
pub added_at: i64,
}
#[derive(Debug, Clone, Serialize)]
pub struct HotReplyTarget {
pub username: String,
pub user_id: Option<String>,
pub target_followers: i64,
}
#[derive(Debug, Clone, Serialize)]
pub struct CrossPostCandidate {
pub tweet_id: String,
pub text_preview: String,
pub posted_at: i64,
pub impressions: i64,
pub likes: i64,
pub bookmarks: i64,
pub suggested_action: String,
}
#[derive(Debug, Clone, Serialize)]
pub struct HotTargetStats {
pub username: String,
pub sample_count: i64,
pub avg_impressions: f64,
pub avg_profile_clicks: f64,
pub reply_back_rate: f64,
pub last_reply_at: i64,
pub score: f64,
}
#[derive(Debug, Clone)]
pub struct PendingReply {
pub id: i64,
pub reply_tweet_id: String,
pub target_username: Option<String>,
pub performed_at: i64,
}
#[derive(Debug, Clone)]
pub struct FullSnapshot {
pub snapshot_at: i64,
pub minutes_since_post: i64,
pub likes: i64,
pub retweets: i64,
pub replies: i64,
pub impressions: i64,
pub bookmarks: i64,
pub quotes: i64,
pub profile_clicks: i64,
}
#[derive(Debug, Clone, Serialize)]
pub struct PostRecord {
pub tweet_id: String,
pub text: String,
pub content_type: String,
pub posted_at: i64,
pub preflight_score: Option<f64>,
pub latest_metrics: Option<SnapshotRecord>,
}
#[derive(Debug, Clone, Serialize)]
pub struct SnapshotRecord {
pub likes: i64,
pub retweets: i64,
pub replies: i64,
pub impressions: i64,
pub bookmarks: i64,
pub engagement_rate: f64,
}
#[derive(Debug, Clone, Serialize)]
pub struct TimingSlot {
pub day_of_week: i32,
pub hour_of_day: i32,
pub avg_impressions: f64,
pub avg_engagement_rate: f64,
pub sample_count: i64,
}
#[derive(Debug, Clone, Serialize)]
pub struct RecentVelocity {
pub posts_1h: i64,
pub posts_6h: i64,
pub posts_24h: i64,
pub standalone_24h: i64,
pub accelerating_post: Option<String>,
}
#[derive(Debug, Clone, Serialize)]
pub struct ReciprocityInfo {
pub total_engagements: i64,
pub replies_received: i64,
pub reply_rate: f64,
}
#[derive(Debug, Clone, Serialize)]
pub struct ReciprocatorInfo {
pub username: String,
pub total_engagements: i64,
pub replies_received: i64,
pub reply_rate: f64,
pub avg_followers: i64,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
pub enum ReplyStyle {
Question,
DataPoint,
Counterpoint,
Anecdote,
Humor,
Agreement,
}
impl ReplyStyle {
pub fn as_str(&self) -> &'static str {
match self {
Self::Question => "question",
Self::DataPoint => "data_point",
Self::Counterpoint => "counterpoint",
Self::Anecdote => "anecdote",
Self::Humor => "humor",
Self::Agreement => "agreement",
}
}
}
impl std::fmt::Display for ReplyStyle {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.write_str(self.as_str())
}
}
pub fn classify_reply(text: &str) -> ReplyStyle {
let lower = text.to_lowercase();
if text.contains('?') {
return ReplyStyle::Question;
}
if lower.contains('%')
|| lower.chars().any(|c| c.is_ascii_digit())
&& (lower.contains("study") || lower.contains("data") || lower.contains("stat"))
{
return ReplyStyle::DataPoint;
}
let counterpoint_markers = ["but ", "however", "actually", "disagree", "on the other hand"];
if counterpoint_markers.iter().any(|m| lower.contains(m)) {
return ReplyStyle::Counterpoint;
}
let anecdote_markers = [
"i've ", "i tested", "in my experience", "i found", "i noticed",
"i tried", "personally", "my own",
];
if anecdote_markers.iter().any(|m| lower.contains(m)) {
return ReplyStyle::Anecdote;
}
let humor_markers = ["lol", "lmao", "haha", "rofl", "😂", "🤣", "💀"];
if humor_markers.iter().any(|m| lower.contains(m)) {
return ReplyStyle::Humor;
}
ReplyStyle::Agreement
}
pub struct IntelStore {
conn: Connection,
}
impl IntelStore {
pub fn open() -> Result<Self, rusqlite::Error> {
let dir = config_dir();
std::fs::create_dir_all(&dir).ok();
let db_path: PathBuf = dir.join("xmaster.db");
Self::open_at(&db_path)
}
pub fn open_at(path: &std::path::Path) -> Result<Self, rusqlite::Error> {
let conn = Connection::open(path)?;
conn.pragma_update(None, "journal_mode", "wal")?;
conn.pragma_update(None, "busy_timeout", 5000)?;
conn.pragma_update(None, "synchronous", "NORMAL")?;
let store = Self { conn };
store.init_tables()?;
Ok(store)
}
fn init_tables(&self) -> Result<(), rusqlite::Error> {
self.conn.execute_batch(
"
CREATE TABLE IF NOT EXISTS posts (
id INTEGER PRIMARY KEY AUTOINCREMENT,
tweet_id TEXT UNIQUE NOT NULL,
text TEXT NOT NULL,
content_type TEXT NOT NULL DEFAULT 'text',
char_count INTEGER NOT NULL,
has_link BOOLEAN NOT NULL DEFAULT 0,
has_media BOOLEAN NOT NULL DEFAULT 0,
has_poll BOOLEAN NOT NULL DEFAULT 0,
hashtag_count INTEGER NOT NULL DEFAULT 0,
hook_text TEXT,
posted_at INTEGER NOT NULL,
day_of_week INTEGER NOT NULL,
hour_of_day INTEGER NOT NULL,
reply_to_id TEXT,
quote_of_id TEXT,
preflight_score REAL,
analysis_json TEXT,
analysis_version INTEGER DEFAULT 1,
scheduled_post_id TEXT,
local_day_of_week INTEGER,
local_hour_of_day INTEGER,
tz_offset_minutes INTEGER
);
CREATE TABLE IF NOT EXISTS metric_snapshots (
id INTEGER PRIMARY KEY AUTOINCREMENT,
tweet_id TEXT NOT NULL REFERENCES posts(tweet_id),
snapshot_at INTEGER NOT NULL,
minutes_since_post INTEGER NOT NULL,
likes INTEGER NOT NULL DEFAULT 0,
retweets INTEGER NOT NULL DEFAULT 0,
replies INTEGER NOT NULL DEFAULT 0,
impressions INTEGER NOT NULL DEFAULT 0,
bookmarks INTEGER NOT NULL DEFAULT 0,
quotes INTEGER NOT NULL DEFAULT 0,
profile_clicks INTEGER NOT NULL DEFAULT 0,
url_clicks INTEGER
);
CREATE TABLE IF NOT EXISTS engagement_actions (
id INTEGER PRIMARY KEY AUTOINCREMENT,
action_type TEXT NOT NULL,
target_tweet_id TEXT,
target_user_id TEXT,
target_username TEXT,
performed_at INTEGER NOT NULL,
got_reply_back BOOLEAN DEFAULT NULL,
target_followers INTEGER
);
CREATE TABLE IF NOT EXISTS watchlist_accounts (
id INTEGER PRIMARY KEY AUTOINCREMENT,
username TEXT UNIQUE NOT NULL,
user_id TEXT,
topic TEXT,
followers INTEGER NOT NULL DEFAULT 0,
added_at INTEGER NOT NULL
);
-- Add reply_tweet_id if not exists (safe for existing DBs)
-- SQLite doesn't support IF NOT EXISTS for ALTER, so we use a pragma check
CREATE TABLE IF NOT EXISTS timing_stats (
id INTEGER PRIMARY KEY AUTOINCREMENT,
day_of_week INTEGER NOT NULL,
hour_of_day INTEGER NOT NULL,
content_type TEXT NOT NULL DEFAULT 'all',
avg_impressions REAL,
avg_engagement_rate REAL,
sample_count INTEGER NOT NULL DEFAULT 0,
last_updated INTEGER NOT NULL
);
CREATE TABLE IF NOT EXISTS discovered_posts (
id INTEGER PRIMARY KEY AUTOINCREMENT,
tweet_id TEXT UNIQUE NOT NULL,
text TEXT NOT NULL,
author_id TEXT,
author_username TEXT,
created_at TEXT,
conversation_id TEXT,
referenced_tweets_json TEXT NOT NULL DEFAULT '[]',
like_count INTEGER,
retweet_count INTEGER,
reply_count INTEGER,
impression_count INTEGER,
bookmark_count INTEGER,
author_followers INTEGER,
media_urls_json TEXT NOT NULL DEFAULT '[]',
first_source TEXT NOT NULL,
last_source TEXT NOT NULL,
first_discovered_at INTEGER NOT NULL,
last_seen_at INTEGER NOT NULL
);
CREATE INDEX IF NOT EXISTS idx_discovered_author
ON discovered_posts(author_username);
CREATE INDEX IF NOT EXISTS idx_discovered_last_seen
ON discovered_posts(last_seen_at DESC);
CREATE INDEX IF NOT EXISTS idx_discovered_impressions
ON discovered_posts(impression_count DESC);
",
)?;
self.conn.execute_batch("BEGIN IMMEDIATE;")?;
let migrate_result = (|| -> Result<(), rusqlite::Error> {
let cols: Vec<String> = self.conn
.prepare("PRAGMA table_info(engagement_actions)")?
.query_map([], |row| row.get::<_, String>(1))?
.collect::<Result<Vec<_>, _>>()?;
if !cols.iter().any(|c| c == "reply_tweet_id") {
self.conn.execute_batch("ALTER TABLE engagement_actions ADD COLUMN reply_tweet_id TEXT;")?;
}
if !cols.iter().any(|c| c == "reply_style") {
self.conn.execute_batch("ALTER TABLE engagement_actions ADD COLUMN reply_style TEXT;")?;
}
let post_cols: Vec<String> = self.conn
.prepare("PRAGMA table_info(posts)")?
.query_map([], |row| row.get::<_, String>(1))?
.collect::<Result<Vec<_>, _>>()?;
if !post_cols.iter().any(|c| c == "analysis_json") {
self.conn.execute_batch("ALTER TABLE posts ADD COLUMN analysis_json TEXT;")?;
}
if !post_cols.iter().any(|c| c == "analysis_version") {
self.conn.execute_batch("ALTER TABLE posts ADD COLUMN analysis_version INTEGER DEFAULT 1;")?;
}
if !post_cols.iter().any(|c| c == "scheduled_post_id") {
self.conn.execute_batch("ALTER TABLE posts ADD COLUMN scheduled_post_id TEXT;")?;
}
let ms_cols: Vec<String> = self.conn
.prepare("PRAGMA table_info(metric_snapshots)")?
.query_map([], |row| row.get::<_, String>(1))?
.collect::<Result<Vec<_>, _>>()?;
if !ms_cols.iter().any(|c| c == "url_clicks") {
self.conn.execute_batch("ALTER TABLE metric_snapshots ADD COLUMN url_clicks INTEGER;")?;
}
Ok(())
})();
match migrate_result {
Ok(()) => self.conn.execute_batch("COMMIT;")?,
Err(e) => { self.conn.execute_batch("ROLLBACK;").ok(); return Err(e); }
}
self.conn.execute_batch("DROP VIEW IF EXISTS reply_outcomes;")?;
self.conn.execute_batch(
"CREATE VIEW reply_outcomes AS
SELECT ea.id AS action_id,
ea.target_tweet_id,
ea.target_user_id,
ea.target_username,
ea.target_followers,
ea.reply_tweet_id,
ea.reply_style,
ea.performed_at,
ea.got_reply_back,
ms.likes,
ms.retweets,
ms.replies,
ms.impressions,
ms.bookmarks,
ms.quotes,
ms.profile_clicks,
ms.url_clicks,
CAST(ms.snapshot_at AS INTEGER) AS snapshot_at,
ms.minutes_since_post
FROM engagement_actions ea
LEFT JOIN metric_snapshots ms
ON ms.tweet_id = ea.reply_tweet_id
AND ms.id = (
SELECT MAX(ms2.id) FROM metric_snapshots ms2
WHERE ms2.tweet_id = ea.reply_tweet_id
)
WHERE ea.action_type = 'reply'
AND ea.reply_tweet_id IS NOT NULL;"
)?;
Ok(())
}
pub fn classify_reply_style(text: &str) -> ReplyStyle {
classify_reply(text)
}
pub fn add_watchlist(&self, username: &str, user_id: Option<&str>, topic: Option<&str>, followers: i64) -> Result<(), rusqlite::Error> {
let now = Utc::now().timestamp();
self.conn.execute(
"INSERT OR REPLACE INTO watchlist_accounts (username, user_id, topic, followers, added_at)
VALUES (LOWER(?1), ?2, ?3, ?4, ?5)",
params![username.to_lowercase(), user_id, topic, followers, now],
)?;
Ok(())
}
pub fn remove_watchlist(&self, username: &str) -> Result<bool, rusqlite::Error> {
let changed = self.conn.execute(
"DELETE FROM watchlist_accounts WHERE username = LOWER(?1)",
params![username.to_lowercase()],
)?;
Ok(changed > 0)
}
pub fn list_watchlist(&self) -> Result<Vec<WatchlistEntry>, rusqlite::Error> {
let mut stmt = self.conn.prepare(
"SELECT username, user_id, topic, followers, added_at FROM watchlist_accounts ORDER BY added_at DESC"
)?;
let rows = stmt.query_map([], |row| {
Ok(WatchlistEntry {
username: row.get(0)?,
user_id: row.get(1)?,
topic: row.get(2)?,
followers: row.get(3)?,
added_at: row.get(4)?,
})
})?;
rows.collect()
}
pub fn log_reply(
&self,
target_tweet_id: &str,
target_user_id: Option<&str>,
target_username: Option<&str>,
target_followers: Option<i64>,
reply_tweet_id: &str,
reply_style: Option<&ReplyStyle>,
) -> Result<(), rusqlite::Error> {
let performed_at = Utc::now().timestamp();
self.conn.execute(
"INSERT INTO engagement_actions
(action_type, target_tweet_id, target_user_id, target_username,
performed_at, target_followers, reply_tweet_id, reply_style)
VALUES ('reply', ?1, ?2, LOWER(?3), ?4, ?5, ?6, ?7)",
params![
target_tweet_id,
target_user_id,
target_username.map(|u| u.to_lowercase()),
performed_at,
target_followers,
reply_tweet_id,
reply_style.map(|s| s.as_str()),
],
)?;
Ok(())
}
pub fn get_pending_replies(&self, max_age_hours: i64) -> Result<Vec<PendingReply>, rusqlite::Error> {
let cutoff = Utc::now().timestamp() - (max_age_hours * 3600);
let mut stmt = self.conn.prepare(
"SELECT id, reply_tweet_id, target_username, performed_at
FROM engagement_actions
WHERE action_type = 'reply' AND got_reply_back IS NULL
AND reply_tweet_id IS NOT NULL AND performed_at > ?1
ORDER BY performed_at DESC"
)?;
let rows = stmt.query_map(params![cutoff], |row| {
Ok(PendingReply {
id: row.get(0)?,
reply_tweet_id: row.get(1)?,
target_username: row.get(2)?,
performed_at: row.get(3)?,
})
})?;
rows.collect()
}
pub fn set_reply_back(&self, action_id: i64, got_reply: bool) -> Result<(), rusqlite::Error> {
self.conn.execute(
"UPDATE engagement_actions SET got_reply_back = ?1 WHERE id = ?2",
params![got_reply as i32, action_id],
)?;
Ok(())
}
pub fn log_post(
&self,
tweet_id: &str,
text: &str,
content_type: &str,
reply_to: Option<&str>,
quote_of: Option<&str>,
preflight_score: Option<f64>,
analysis_json: Option<&str>,
scheduled_post_id: Option<&str>,
) -> Result<(), rusqlite::Error> {
let now = Utc::now();
let local_now = chrono::Local::now();
let posted_at = now.timestamp();
let day_of_week = now.weekday().num_days_from_monday() as i32; let hour_of_day = now.hour() as i32; let local_day = local_now.weekday().num_days_from_monday() as i32;
let local_hour = local_now.hour() as i32;
let tz_offset = local_now.offset().local_minus_utc() / 60;
let char_count = text.chars().count() as i32;
let has_link = text.contains("http://") || text.contains("https://");
let hashtag_count = text.matches('#').count() as i32;
let hook_text: String = text.chars().take(140).collect();
self.conn.execute(
"INSERT OR IGNORE INTO posts
(tweet_id, text, content_type, char_count, has_link, has_media, has_poll,
hashtag_count, hook_text, posted_at, day_of_week, hour_of_day,
reply_to_id, quote_of_id, preflight_score,
analysis_json, analysis_version, scheduled_post_id,
local_day_of_week, local_hour_of_day, tz_offset_minutes)
VALUES (?1,?2,?3,?4,?5,0,0,?6,?7,?8,?9,?10,?11,?12,?13,?14,1,?15,?16,?17,?18)",
params![
tweet_id,
text,
content_type,
char_count,
has_link,
hashtag_count,
hook_text,
posted_at,
day_of_week,
hour_of_day,
reply_to,
quote_of,
preflight_score,
analysis_json,
scheduled_post_id,
local_day,
local_hour,
tz_offset,
],
)?;
Ok(())
}
#[allow(clippy::too_many_arguments)]
pub fn record_published_post(
&self,
tweet_id: &str,
text: &str,
content_type: &str,
reply_to: Option<&str>,
quote_of: Option<&str>,
preflight_score: Option<f64>,
analysis_json: Option<&str>,
scheduled_post_id: Option<&str>,
) -> Result<(), rusqlite::Error> {
self.log_post(
tweet_id,
text,
content_type,
reply_to,
quote_of,
preflight_score,
analysis_json,
scheduled_post_id,
)
}
pub fn record_discovered_posts(
&self,
source: &str,
tweets: &[crate::providers::xapi::TweetData],
) -> Result<(), rusqlite::Error> {
if tweets.is_empty() {
return Ok(());
}
let now = Utc::now().timestamp();
let tx = self.conn.unchecked_transaction()?;
{
let mut stmt = tx.prepare_cached(
"INSERT INTO discovered_posts (
tweet_id, text, author_id, author_username, created_at,
conversation_id, referenced_tweets_json,
like_count, retweet_count, reply_count, impression_count,
bookmark_count, author_followers, media_urls_json,
first_source, last_source, first_discovered_at, last_seen_at
) VALUES (?1,?2,?3,?4,?5,?6,?7,?8,?9,?10,?11,?12,?13,?14,?15,?15,?16,?16)
ON CONFLICT(tweet_id) DO UPDATE SET
text = excluded.text,
author_id = COALESCE(excluded.author_id, discovered_posts.author_id),
author_username = COALESCE(excluded.author_username, discovered_posts.author_username),
created_at = COALESCE(excluded.created_at, discovered_posts.created_at),
conversation_id = COALESCE(excluded.conversation_id, discovered_posts.conversation_id),
referenced_tweets_json = CASE
WHEN excluded.referenced_tweets_json <> '[]' THEN excluded.referenced_tweets_json
ELSE discovered_posts.referenced_tweets_json END,
like_count = COALESCE(excluded.like_count, discovered_posts.like_count),
retweet_count = COALESCE(excluded.retweet_count, discovered_posts.retweet_count),
reply_count = COALESCE(excluded.reply_count, discovered_posts.reply_count),
impression_count = COALESCE(excluded.impression_count, discovered_posts.impression_count),
bookmark_count = COALESCE(excluded.bookmark_count, discovered_posts.bookmark_count),
author_followers = COALESCE(excluded.author_followers, discovered_posts.author_followers),
media_urls_json = CASE
WHEN excluded.media_urls_json <> '[]' THEN excluded.media_urls_json
ELSE discovered_posts.media_urls_json END,
last_source = excluded.last_source,
last_seen_at = excluded.last_seen_at",
)?;
for t in tweets {
let m = t.public_metrics.as_ref();
let refs_json = serde_json::to_string(
&t.referenced_tweets.as_deref().unwrap_or(&[])
).unwrap_or_else(|_| "[]".into());
let media_json = serde_json::to_string(&t.media_urls)
.unwrap_or_else(|_| "[]".into());
stmt.execute(params![
t.id,
t.text,
t.author_id,
t.author_username,
t.created_at,
t.conversation_id,
refs_json,
m.map(|m| m.like_count as i64),
m.map(|m| m.retweet_count as i64),
m.map(|m| m.reply_count as i64),
m.map(|m| m.impression_count as i64),
m.map(|m| m.bookmark_count as i64),
t.author_followers.map(|f| f as i64),
media_json,
source,
now,
])?;
}
}
tx.commit()?;
Ok(())
}
pub fn record_discovered_post(
&self,
source: &str,
tweet: &crate::providers::xapi::TweetData,
) -> Result<(), rusqlite::Error> {
self.record_discovered_posts(source, std::slice::from_ref(tweet))
}
pub fn query_discovered_posts(
&self,
topic: Option<&str>,
author: Option<&str>,
min_likes: Option<i64>,
limit: usize,
) -> Result<Vec<DiscoveredPostRow>, rusqlite::Error> {
let mut sql = String::from(
"SELECT tweet_id, COALESCE(author_username,''), text,
COALESCE(like_count,0), COALESCE(impression_count,0),
last_source, first_discovered_at
FROM discovered_posts WHERE 1=1"
);
let mut param_idx = 1usize;
let topic_idx = if topic.is_some() {
sql.push_str(&format!(" AND text LIKE '%' || ?{param_idx} || '%'"));
let idx = param_idx; param_idx += 1; Some(idx)
} else { None };
let author_idx = if author.is_some() {
sql.push_str(&format!(" AND author_username LIKE '%' || ?{param_idx} || '%'"));
let idx = param_idx; param_idx += 1; Some(idx)
} else { None };
let likes_idx = if min_likes.is_some() {
sql.push_str(&format!(" AND like_count >= ?{param_idx}"));
let idx = param_idx; param_idx += 1; Some(idx)
} else { None };
sql.push_str(&format!(" ORDER BY COALESCE(impression_count,0) DESC LIMIT ?{param_idx}"));
let mut stmt = self.conn.prepare(&sql)?;
let mut bind_idx = 1usize;
if let Some(_) = topic_idx { stmt.raw_bind_parameter(bind_idx, topic.unwrap())?; bind_idx += 1; }
if let Some(_) = author_idx { stmt.raw_bind_parameter(bind_idx, author.unwrap())?; bind_idx += 1; }
if let Some(_) = likes_idx { stmt.raw_bind_parameter(bind_idx, min_likes.unwrap())?; bind_idx += 1; }
stmt.raw_bind_parameter(bind_idx, limit as i64)?;
let mut rows = Vec::new();
let mut raw = stmt.raw_query();
while let Some(row) = raw.next()? {
rows.push(DiscoveredPostRow {
tweet_id: row.get(0)?,
author_username: row.get(1)?,
text: row.get(2)?,
like_count: row.get(3)?,
impression_count: row.get(4)?,
last_source: row.get(5)?,
first_discovered_at: row.get(6)?,
});
}
Ok(rows)
}
pub fn discovered_posts_count(&self) -> Result<i64, rusqlite::Error> {
self.conn.query_row("SELECT COUNT(*) FROM discovered_posts", [], |r| r.get(0))
}
#[allow(clippy::too_many_arguments)]
pub fn log_metric_snapshot(
&self,
tweet_id: &str,
likes: i64,
retweets: i64,
replies: i64,
impressions: i64,
bookmarks: i64,
quotes: i64,
profile_clicks: i64,
minutes_since_post: i64,
url_clicks: Option<i64>,
) -> Result<(), rusqlite::Error> {
let snapshot_at = Utc::now().timestamp();
self.conn.execute(
"INSERT INTO metric_snapshots
(tweet_id, snapshot_at, minutes_since_post, likes, retweets, replies,
impressions, bookmarks, quotes, profile_clicks, url_clicks)
VALUES (?1,?2,?3,?4,?5,?6,?7,?8,?9,?10,?11)",
params![
tweet_id,
snapshot_at,
minutes_since_post,
likes,
retweets,
replies,
impressions,
bookmarks,
quotes,
profile_clicks,
url_clicks,
],
)?;
Ok(())
}
pub fn latest_snapshot_full(
&self,
tweet_id: &str,
) -> Result<Option<FullSnapshot>, rusqlite::Error> {
self.conn
.query_row(
"SELECT snapshot_at, minutes_since_post,
likes, retweets, replies,
impressions, bookmarks, quotes, profile_clicks
FROM metric_snapshots
WHERE tweet_id = ?1
ORDER BY snapshot_at DESC
LIMIT 1",
params![tweet_id],
|row| {
Ok(FullSnapshot {
snapshot_at: row.get(0)?,
minutes_since_post: row.get(1)?,
likes: row.get(2)?,
retweets: row.get(3)?,
replies: row.get(4)?,
impressions: row.get(5)?,
bookmarks: row.get(6)?,
quotes: row.get(7)?,
profile_clicks: row.get(8)?,
})
},
)
.optional()
}
pub fn log_engagement(
&self,
action_type: &str,
target_tweet_id: Option<&str>,
target_user_id: Option<&str>,
target_username: Option<&str>,
target_followers: Option<i64>,
) -> Result<(), rusqlite::Error> {
let performed_at = Utc::now().timestamp();
self.conn.execute(
"INSERT INTO engagement_actions
(action_type, target_tweet_id, target_user_id, target_username,
performed_at, target_followers)
VALUES (?1,?2,?3,?4,?5,?6)",
params![
action_type,
target_tweet_id,
target_user_id,
target_username,
performed_at,
target_followers,
],
)?;
Ok(())
}
pub fn get_post_history(&self, limit: i64) -> Result<Vec<PostRecord>, rusqlite::Error> {
let mut stmt = self.conn.prepare(
"SELECT tweet_id, text, content_type, posted_at, preflight_score
FROM posts ORDER BY posted_at DESC LIMIT ?1",
)?;
let rows = stmt.query_map(params![limit], |row| {
Ok(PostRecord {
tweet_id: row.get(0)?,
text: row.get(1)?,
content_type: row.get(2)?,
posted_at: row.get(3)?,
preflight_score: row.get(4)?,
latest_metrics: None, })
})?;
let mut posts: Vec<PostRecord> = rows.collect::<Result<Vec<_>, _>>()?;
for post in &mut posts {
post.latest_metrics = self.latest_snapshot(&post.tweet_id)?;
}
Ok(posts)
}
fn latest_snapshot(&self, tweet_id: &str) -> Result<Option<SnapshotRecord>, rusqlite::Error> {
self.conn
.query_row(
"SELECT likes, retweets, replies, impressions, bookmarks
FROM metric_snapshots
WHERE tweet_id = ?1
ORDER BY snapshot_at DESC LIMIT 1",
params![tweet_id],
|row| {
let likes: i64 = row.get(0)?;
let retweets: i64 = row.get(1)?;
let replies: i64 = row.get(2)?;
let impressions: i64 = row.get(3)?;
let bookmarks: i64 = row.get(4)?;
let engagement_rate = if impressions > 0 {
(likes + retweets + replies) as f64 / impressions as f64
} else {
0.0
};
Ok(SnapshotRecord {
likes,
retweets,
replies,
impressions,
bookmarks,
engagement_rate,
})
},
)
.optional()
}
pub fn get_timing_heatmap(&self) -> Result<Vec<TimingSlot>, rusqlite::Error> {
let mut stmt = self.conn.prepare(
"SELECT p.day_of_week, p.hour_of_day,
AVG(ms.impressions) AS avg_imp,
AVG(CASE WHEN ms.impressions > 0
THEN (ms.likes + ms.retweets + ms.replies + ms.quotes) * 1.0
/ ms.impressions ELSE 0 END) AS avg_er,
COUNT(DISTINCT p.tweet_id) AS cnt
FROM posts p
JOIN metric_snapshots ms ON ms.tweet_id = p.tweet_id
GROUP BY p.day_of_week, p.hour_of_day
ORDER BY avg_er DESC",
)?;
let rows = stmt.query_map([], |row| {
Ok(TimingSlot {
day_of_week: row.get(0)?,
hour_of_day: row.get(1)?,
avg_impressions: row.get(2)?,
avg_engagement_rate: row.get(3)?,
sample_count: row.get(4)?,
})
})?;
rows.collect()
}
pub fn get_best_posting_times(
&self,
content_type: Option<&str>,
top_n: i64,
) -> Result<Vec<TimingSlot>, rusqlite::Error> {
let (sql, use_filter) = match content_type {
Some(_) => (
"SELECT p.day_of_week, p.hour_of_day,
AVG(ms.impressions) AS avg_imp,
AVG(CASE WHEN ms.impressions > 0
THEN (ms.likes+ms.retweets+ms.replies+ms.quotes)*1.0
/ ms.impressions ELSE 0 END) AS avg_er,
COUNT(DISTINCT p.tweet_id) AS cnt
FROM posts p
JOIN metric_snapshots ms ON ms.tweet_id = p.tweet_id
WHERE p.content_type = ?1
GROUP BY p.day_of_week, p.hour_of_day
HAVING cnt >= 2
ORDER BY avg_er DESC
LIMIT ?2",
true,
),
None => (
"SELECT p.day_of_week, p.hour_of_day,
AVG(ms.impressions) AS avg_imp,
AVG(CASE WHEN ms.impressions > 0
THEN (ms.likes+ms.retweets+ms.replies+ms.quotes)*1.0
/ ms.impressions ELSE 0 END) AS avg_er,
COUNT(DISTINCT p.tweet_id) AS cnt
FROM posts p
JOIN metric_snapshots ms ON ms.tweet_id = p.tweet_id
GROUP BY p.day_of_week, p.hour_of_day
HAVING cnt >= 2
ORDER BY avg_er DESC
LIMIT ?1",
false,
),
};
let mut stmt = self.conn.prepare(sql)?;
let map_row = |row: &rusqlite::Row| -> rusqlite::Result<TimingSlot> {
Ok(TimingSlot {
day_of_week: row.get(0)?,
hour_of_day: row.get(1)?,
avg_impressions: row.get(2)?,
avg_engagement_rate: row.get(3)?,
sample_count: row.get(4)?,
})
};
let results: Vec<TimingSlot> = if use_filter {
stmt.query_map(params![content_type.unwrap(), top_n], map_row)?
.collect::<Result<Vec<_>, _>>()?
} else {
stmt.query_map(params![top_n], map_row)?
.collect::<Result<Vec<_>, _>>()?
};
Ok(results)
}
pub fn get_recent_post_velocity(&self) -> Result<RecentVelocity, rusqlite::Error> {
let now = Utc::now().timestamp();
let posts_1h: i64 = self.conn.query_row(
"SELECT COUNT(*) FROM posts
WHERE posted_at > ?1 - 3600",
params![now],
|r| r.get(0),
)?;
let posts_6h: i64 = self.conn.query_row(
"SELECT COUNT(*) FROM posts
WHERE posted_at > ?1 - 21600",
params![now],
|r| r.get(0),
)?;
let posts_24h: i64 = self.conn.query_row(
"SELECT COUNT(*) FROM posts
WHERE posted_at > ?1 - 86400",
params![now],
|r| r.get(0),
)?;
let standalone_24h: i64 = self.conn.query_row(
"SELECT COUNT(*) FROM posts
WHERE posted_at > ?1 - 86400
AND content_type != 'reply'",
params![now],
|r| r.get(0),
)?;
let accelerating_post: Option<String> = self
.conn
.query_row(
"SELECT s1.tweet_id
FROM metric_snapshots s1
JOIN metric_snapshots s2 ON s1.tweet_id = s2.tweet_id
AND s2.id = (SELECT MAX(id) FROM metric_snapshots
WHERE tweet_id = s1.tweet_id AND id < s1.id)
JOIN posts p ON p.tweet_id = s1.tweet_id
WHERE s1.id = (SELECT MAX(id) FROM metric_snapshots
WHERE tweet_id = s1.tweet_id)
AND p.posted_at > ?1 - 86400
AND s1.impressions > s2.impressions * 1.5
ORDER BY (s1.impressions - s2.impressions) DESC
LIMIT 1",
params![now],
|r| r.get(0),
)
.optional()?;
Ok(RecentVelocity {
posts_1h,
posts_6h,
posts_24h,
standalone_24h,
accelerating_post,
})
}
pub fn update_timing_stats(&self) -> Result<(), rusqlite::Error> {
let now = Utc::now().timestamp();
let tx = self.conn.unchecked_transaction()?;
tx.execute_batch("DELETE FROM timing_stats")?;
tx.execute(
"INSERT INTO timing_stats
(day_of_week, hour_of_day, content_type,
avg_impressions, avg_engagement_rate, sample_count, last_updated)
SELECT p.day_of_week, p.hour_of_day, p.content_type,
AVG(ms.impressions),
AVG(CASE WHEN ms.impressions > 0
THEN (ms.likes+ms.retweets+ms.replies+ms.quotes)*1.0
/ ms.impressions ELSE 0 END),
COUNT(DISTINCT p.tweet_id),
?1
FROM posts p
JOIN metric_snapshots ms ON ms.tweet_id = p.tweet_id
GROUP BY p.day_of_week, p.hour_of_day, p.content_type",
params![now],
)?;
tx.execute(
"INSERT INTO timing_stats
(day_of_week, hour_of_day, content_type,
avg_impressions, avg_engagement_rate, sample_count, last_updated)
SELECT p.day_of_week, p.hour_of_day, 'all',
AVG(ms.impressions),
AVG(CASE WHEN ms.impressions > 0
THEN (ms.likes+ms.retweets+ms.replies+ms.quotes)*1.0
/ ms.impressions ELSE 0 END),
COUNT(DISTINCT p.tweet_id),
?1
FROM posts p
JOIN metric_snapshots ms ON ms.tweet_id = p.tweet_id
GROUP BY p.day_of_week, p.hour_of_day",
params![now],
)?;
tx.commit()?;
Ok(())
}
pub fn get_engagement_reciprocity(
&self,
username: &str,
) -> Result<Option<ReciprocityInfo>, rusqlite::Error> {
self.conn
.query_row(
"SELECT COUNT(*) AS total,
SUM(CASE WHEN got_reply_back = 1 THEN 1 ELSE 0 END) AS replies_back
FROM engagement_actions
WHERE target_username = ?1",
params![username],
|row| {
let total: i64 = row.get(0)?;
if total == 0 {
return Ok(None);
}
let replies_received: i64 = row.get(1)?;
Ok(Some(ReciprocityInfo {
total_engagements: total,
replies_received,
reply_rate: if total > 0 {
replies_received as f64 / total as f64
} else {
0.0
},
}))
},
)
}
pub fn find_hot_reply_targets(
&self,
min_impressions: i64,
min_profile_clicks: i64,
min_target_followers: i64,
max_age_hours: i64,
) -> Result<Vec<HotReplyTarget>, rusqlite::Error> {
let cutoff = Utc::now().timestamp() - max_age_hours * 3600;
let mut stmt = self.conn.prepare(
"SELECT DISTINCT LOWER(ea.target_username),
ea.target_user_id,
COALESCE(ea.target_followers, 0)
FROM engagement_actions ea
LEFT JOIN metric_snapshots ms
ON ms.tweet_id = ea.reply_tweet_id
AND ms.id = (
SELECT MAX(ms2.id) FROM metric_snapshots ms2
WHERE ms2.tweet_id = ea.reply_tweet_id
)
LEFT JOIN watchlist_accounts wa
ON wa.username = LOWER(ea.target_username)
WHERE ea.action_type = 'reply'
AND ea.target_username IS NOT NULL
AND ea.reply_tweet_id IS NOT NULL
AND CAST(ea.performed_at AS INTEGER) >= ?1
AND COALESCE(ea.target_followers, 0) >= ?2
AND wa.username IS NULL
AND (COALESCE(ms.impressions, 0) >= ?3
OR COALESCE(ms.profile_clicks, 0) >= ?4
OR COALESCE(ea.got_reply_back, 0) = 1)
ORDER BY COALESCE(ea.target_followers, 0) DESC"
)?;
let rows = stmt.query_map(
params![cutoff, min_target_followers, min_impressions, min_profile_clicks],
|row| {
Ok(HotReplyTarget {
username: row.get(0)?,
user_id: row.get(1)?,
target_followers: row.get(2)?,
})
},
)?;
rows.collect()
}
pub fn find_cross_post_candidates(
&self,
min_impressions: i64,
max_age_days: i64,
max_results: i64,
) -> Result<Vec<CrossPostCandidate>, rusqlite::Error> {
let cutoff = Utc::now().timestamp() - max_age_days * 24 * 3600;
let mut stmt = self.conn.prepare(
"SELECT p.tweet_id,
p.text,
p.posted_at,
MAX(ms.impressions) AS peak_impressions,
MAX(ms.likes) AS peak_likes,
MAX(ms.bookmarks) AS peak_bookmarks
FROM posts p
JOIN metric_snapshots ms ON ms.tweet_id = p.tweet_id
WHERE p.content_type NOT IN ('reply', 'thread_reply')
AND p.posted_at >= ?1
GROUP BY p.tweet_id, p.text, p.posted_at
HAVING MAX(ms.impressions) >= ?2
ORDER BY MAX(ms.impressions) DESC
LIMIT ?3"
)?;
let rows = stmt.query_map(
params![cutoff, min_impressions, max_results],
|row| {
let tweet_id: String = row.get(0)?;
let text: String = row.get(1)?;
let posted_at: i64 = row.get(2)?;
let impressions: i64 = row.get(3)?;
let likes: i64 = row.get(4)?;
let bookmarks: i64 = row.get(5)?;
Ok(CrossPostCandidate {
tweet_id: tweet_id.clone(),
text_preview: if text.chars().count() > 120 {
let truncated: String = text.chars().take(120).collect();
format!("{truncated}...")
} else {
text
},
posted_at,
impressions,
likes,
bookmarks,
suggested_action: format!(
"Screenshot https://x.com/i/status/{tweet_id} and repost as an Instagram still via the clinstagram skill"
),
})
},
)?;
rows.collect()
}
pub fn rank_hot_reply_targets(
&self,
days: i64,
min_samples: i64,
min_avg_impressions: f64,
min_avg_profile_clicks: f64,
) -> Result<Vec<HotTargetStats>, rusqlite::Error> {
let cutoff = Utc::now().timestamp() - days * 24 * 3600;
let mut stmt = self.conn.prepare(
"SELECT LOWER(ea.target_username) AS username,
COUNT(*) AS sample_count,
AVG(COALESCE(ms.impressions, 0)) AS avg_imps,
AVG(COALESCE(ms.profile_clicks, 0)) AS avg_clicks,
AVG(CASE WHEN ea.got_reply_back = 1 THEN 1.0 ELSE 0.0 END) AS reply_back_rate,
CAST(MAX(CAST(ea.performed_at AS INTEGER)) AS INTEGER) AS last_reply_at
FROM engagement_actions ea
LEFT JOIN metric_snapshots ms
ON ms.tweet_id = ea.reply_tweet_id
AND ms.id = (
SELECT MAX(ms2.id) FROM metric_snapshots ms2
WHERE ms2.tweet_id = ea.reply_tweet_id
)
WHERE ea.action_type = 'reply'
AND ea.target_username IS NOT NULL
AND ea.reply_tweet_id IS NOT NULL
AND CAST(ea.performed_at AS INTEGER) >= ?1
GROUP BY LOWER(ea.target_username)
HAVING sample_count >= ?2
AND avg_imps >= ?3
AND avg_clicks >= ?4"
)?;
let mut rows: Vec<HotTargetStats> = stmt
.query_map(
params![cutoff, min_samples, min_avg_impressions, min_avg_profile_clicks],
|row| {
let avg_imps: f64 = row.get(2)?;
let avg_clicks: f64 = row.get(3)?;
let reply_back_rate: f64 = row.get(4)?;
let imps_norm = (avg_imps / 1000.0).min(1.0);
let clicks_norm = (avg_clicks / 10.0).min(1.0);
let score = 0.55 * imps_norm + 0.25 * clicks_norm + 0.20 * reply_back_rate;
Ok(HotTargetStats {
username: row.get(0)?,
sample_count: row.get(1)?,
avg_impressions: avg_imps,
avg_profile_clicks: avg_clicks,
reply_back_rate,
last_reply_at: row.get(5)?,
score,
})
},
)?
.collect::<Result<Vec<_>, _>>()?;
rows.sort_by(|a, b| {
b.score
.partial_cmp(&a.score)
.unwrap_or(std::cmp::Ordering::Equal)
});
Ok(rows)
}
pub fn get_top_reciprocators(
&self,
min_followers: i64,
limit: i64,
) -> Result<Vec<ReciprocatorInfo>, rusqlite::Error> {
let mut stmt = self.conn.prepare(
"SELECT target_username,
COUNT(*) AS total,
SUM(CASE WHEN got_reply_back = 1 THEN 1 ELSE 0 END) AS replies_back,
AVG(target_followers) AS avg_fol
FROM engagement_actions
WHERE target_username IS NOT NULL
AND target_followers >= ?1
GROUP BY target_username
HAVING replies_back > 0
ORDER BY (CAST(replies_back AS REAL) / total) DESC
LIMIT ?2",
)?;
let rows = stmt.query_map(params![min_followers, limit], |row| {
let total: i64 = row.get(1)?;
let replies_received: i64 = row.get(2)?;
let avg_fol: f64 = row.get(3)?;
Ok(ReciprocatorInfo {
username: row.get(0)?,
total_engagements: total,
replies_received,
reply_rate: if total > 0 {
replies_received as f64 / total as f64
} else {
0.0
},
avg_followers: avg_fol as i64,
})
})?;
rows.collect()
}
}
#[cfg(test)]
mod tests {
use super::*;
use tempfile::tempdir;
fn test_store() -> IntelStore {
let dir = tempdir().unwrap();
let db_path = dir.path().join("test.db");
let store = IntelStore::open_at(&db_path).unwrap();
std::mem::forget(dir);
store
}
#[test]
fn log_and_retrieve_post() {
let store = test_store();
store
.log_post("tweet_001", "Hello world!", "opinion", None, None, Some(85.0), None, None)
.unwrap();
let posts = store.get_post_history(10).unwrap();
assert_eq!(posts.len(), 1);
assert_eq!(posts[0].tweet_id, "tweet_001");
assert_eq!(posts[0].text, "Hello world!");
assert_eq!(posts[0].content_type, "opinion");
assert_eq!(posts[0].preflight_score, Some(85.0));
assert!(posts[0].latest_metrics.is_none());
}
#[test]
fn duplicate_tweet_id_ignored() {
let store = test_store();
store
.log_post("tweet_dup", "First", "opinion", None, None, None, None, None)
.unwrap();
store
.log_post("tweet_dup", "Second", "opinion", None, None, None, None, None)
.unwrap();
let posts = store.get_post_history(10).unwrap();
assert_eq!(posts.len(), 1);
assert_eq!(posts[0].text, "First"); }
#[test]
fn engagement_logging() {
let store = test_store();
store
.log_engagement("like", Some("t_100"), None, Some("testuser"), Some(5000))
.unwrap();
store
.log_engagement("reply", Some("t_101"), None, Some("testuser"), Some(5000))
.unwrap();
let info = store.get_engagement_reciprocity("testuser").unwrap();
assert!(info.is_some());
let info = info.unwrap();
assert_eq!(info.total_engagements, 2);
}
#[test]
fn timing_heatmap_empty_db() {
let store = test_store();
let heatmap = store.get_timing_heatmap().unwrap();
assert!(heatmap.is_empty());
}
#[test]
fn recent_velocity_empty_db() {
let store = test_store();
let velocity = store.get_recent_post_velocity().unwrap();
assert_eq!(velocity.posts_1h, 0);
assert_eq!(velocity.posts_6h, 0);
assert_eq!(velocity.posts_24h, 0);
assert_eq!(velocity.standalone_24h, 0);
assert!(velocity.accelerating_post.is_none());
}
#[test]
fn cross_post_candidates_filters_and_excludes_replies() {
let store = test_store();
store
.log_post("big_hit", "a viral thought", "text", None, None, None, None, None)
.unwrap();
store
.log_metric_snapshot("big_hit", 400, 50, 30, 12_000, 100, 10, 5, 30, None)
.unwrap();
store
.log_post("small_post", "meh", "text", None, None, None, None, None)
.unwrap();
store
.log_metric_snapshot("small_post", 10, 1, 0, 1_000, 5, 0, 0, 20, None)
.unwrap();
store
.log_post("viral_reply", "nailed it", "reply", Some("other"), None, None, None, None)
.unwrap();
store
.log_metric_snapshot("viral_reply", 500, 60, 40, 15_000, 80, 15, 20, 40, None)
.unwrap();
let candidates = store.find_cross_post_candidates(5_000, 14, 10).unwrap();
assert_eq!(candidates.len(), 1, "only big_hit qualifies");
assert_eq!(candidates[0].tweet_id, "big_hit");
assert_eq!(candidates[0].impressions, 12_000);
assert!(candidates[0].suggested_action.contains("clinstagram"));
}
#[test]
fn standalone_24h_excludes_replies() {
let store = test_store();
store
.log_post("p_1", "a standalone post", "text", None, None, None, None, None)
.unwrap();
store
.log_post("p_2", "reply to something", "reply", Some("other_id"), None, None, None, None)
.unwrap();
store
.log_post("p_3", "thread opener", "thread_hook", None, None, None, None, None)
.unwrap();
let v = store.get_recent_post_velocity().unwrap();
assert_eq!(v.posts_24h, 3, "all three rows count toward posts_24h");
assert_eq!(v.standalone_24h, 2, "reply must be excluded from standalone_24h");
}
#[test]
fn metric_snapshot_persists_url_clicks() {
let store = test_store();
store
.log_post("tweet_url", "Check this link", "text", None, None, None, None, None)
.unwrap();
store
.log_metric_snapshot("tweet_url", 5, 1, 0, 200, 0, 0, 3, 10, Some(42))
.unwrap();
let clicks: Option<i64> = store
.conn
.query_row(
"SELECT url_clicks FROM metric_snapshots WHERE tweet_id = 'tweet_url'",
[],
|r| r.get(0),
)
.unwrap();
assert_eq!(clicks, Some(42));
}
#[test]
fn metric_snapshot_persists_url_clicks_null_when_none() {
let store = test_store();
store
.log_post("tweet_null", "no link", "text", None, None, None, None, None)
.unwrap();
store
.log_metric_snapshot("tweet_null", 1, 0, 0, 10, 0, 0, 0, 5, None)
.unwrap();
let clicks: Option<i64> = store
.conn
.query_row(
"SELECT url_clicks FROM metric_snapshots WHERE tweet_id = 'tweet_null'",
[],
|r| r.get(0),
)
.unwrap();
assert_eq!(clicks, None);
}
#[test]
fn metric_snapshot_and_retrieval() {
let store = test_store();
store
.log_post("tweet_metrics", "Test post", "opinion", None, None, None, None, None)
.unwrap();
store
.log_metric_snapshot("tweet_metrics", 10, 5, 3, 1000, 2, 1, 50, 60, None)
.unwrap();
let posts = store.get_post_history(10).unwrap();
assert_eq!(posts.len(), 1);
let metrics = posts[0].latest_metrics.as_ref().unwrap();
assert_eq!(metrics.likes, 10);
assert_eq!(metrics.retweets, 5);
assert_eq!(metrics.replies, 3);
assert_eq!(metrics.impressions, 1000);
assert_eq!(metrics.bookmarks, 2);
assert!(metrics.engagement_rate > 0.0);
}
#[test]
fn engagement_reciprocity_unknown_user() {
let store = test_store();
let info = store.get_engagement_reciprocity("nobody").unwrap();
assert!(info.is_none());
}
#[test]
fn update_timing_stats_empty_db() {
let store = test_store();
store.update_timing_stats().unwrap();
}
#[test]
fn classify_reply_question() {
assert_eq!(classify_reply("What do you think about this?"), ReplyStyle::Question);
}
#[test]
fn classify_reply_counterpoint() {
assert_eq!(classify_reply("However, the evidence suggests otherwise"), ReplyStyle::Counterpoint);
}
#[test]
fn classify_reply_anecdote() {
assert_eq!(classify_reply("I've been using this protocol for 6 months"), ReplyStyle::Anecdote);
}
#[test]
fn classify_reply_humor() {
assert_eq!(classify_reply("lol that's incredible"), ReplyStyle::Humor);
}
#[test]
fn classify_reply_agreement_fallback() {
assert_eq!(classify_reply("Great insight, totally agree"), ReplyStyle::Agreement);
}
#[test]
fn record_published_post_works() {
let store = test_store();
store
.record_published_post(
"tweet_pub_001",
"Test published post",
"text",
None,
None,
Some(75.0),
Some(r#"{"score":75}"#),
Some("sched_123"),
)
.unwrap();
let posts = store.get_post_history(10).unwrap();
assert_eq!(posts.len(), 1);
assert_eq!(posts[0].tweet_id, "tweet_pub_001");
}
#[test]
fn log_reply_with_style() {
let store = test_store();
let style = classify_reply("What makes you think that?");
store
.log_reply("target_001", None, Some("testuser"), Some(5000), "reply_001", Some(&style))
.unwrap();
let pending = store.get_pending_replies(24).unwrap();
assert_eq!(pending.len(), 1);
assert_eq!(pending[0].reply_tweet_id, "reply_001");
}
#[test]
fn discovered_posts_upsert_and_query() {
let store = test_store();
let tweet = crate::providers::xapi::TweetData {
id: "dp_123".into(),
text: "Longevity research is the future".into(),
author_id: Some("user1".into()),
author_username: Some("testuser".into()),
created_at: Some("2026-01-01T00:00:00Z".into()),
conversation_id: None,
referenced_tweets: None,
public_metrics: Some(crate::providers::xapi::TweetMetrics {
like_count: 42,
retweet_count: 5,
reply_count: 3,
impression_count: 2000,
bookmark_count: 1,
}),
author_followers: Some(500),
media_urls: vec![],
};
store.record_discovered_post("search", &tweet).unwrap();
assert_eq!(store.discovered_posts_count().unwrap(), 1);
store.record_discovered_post("timeline", &tweet).unwrap();
assert_eq!(store.discovered_posts_count().unwrap(), 1);
let last: String = store.conn.query_row(
"SELECT last_source FROM discovered_posts WHERE tweet_id = 'dp_123'", [], |r| r.get(0)
).unwrap();
assert_eq!(last, "timeline");
let first: String = store.conn.query_row(
"SELECT first_source FROM discovered_posts WHERE tweet_id = 'dp_123'", [], |r| r.get(0)
).unwrap();
assert_eq!(first, "search");
let results = store.query_discovered_posts(Some("longevity"), None, None, 10).unwrap();
assert_eq!(results.len(), 1);
assert_eq!(results[0].tweet_id, "dp_123");
let results = store.query_discovered_posts(None, None, Some(100), 10).unwrap();
assert_eq!(results.len(), 0); let results = store.query_discovered_posts(None, None, Some(10), 10).unwrap();
assert_eq!(results.len(), 1);
let results = store.query_discovered_posts(None, Some("testuser"), None, 10).unwrap();
assert_eq!(results.len(), 1);
}
fn log_outgoing_reply(
store: &IntelStore,
target_tweet_id: &str,
target_user_id: Option<&str>,
target_username: &str,
target_followers: i64,
reply_tweet_id: &str,
) {
store
.log_post(reply_tweet_id, "reply body", "text", Some(target_tweet_id), None, None, None, None)
.unwrap();
store
.log_reply(
target_tweet_id,
target_user_id,
Some(target_username),
Some(target_followers),
reply_tweet_id,
None,
)
.unwrap();
}
#[test]
fn find_hot_reply_targets_returns_high_impression_reply() {
let store = test_store();
log_outgoing_reply(&store, "target_tweet_1", Some("uid_42"), "hottarget", 5000, "my_reply_1");
store
.log_metric_snapshot("my_reply_1", 10, 2, 1, 500, 0, 0, 3, 15, None)
.unwrap();
let promoted = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert_eq!(promoted.len(), 1);
assert_eq!(promoted[0].username, "hottarget");
assert_eq!(promoted[0].user_id.as_deref(), Some("uid_42"));
assert_eq!(promoted[0].target_followers, 5000);
}
#[test]
fn find_hot_reply_targets_skips_low_follower_targets() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "smallfry", 500, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 1000, 0, 0, 5, 10, None)
.unwrap();
let promoted = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert!(promoted.is_empty(), "should filter target below min_target_followers");
}
#[test]
fn find_hot_reply_targets_skips_already_watchlisted() {
let store = test_store();
store
.add_watchlist("alreadyhere", Some("uid_1"), None, 10_000)
.unwrap();
log_outgoing_reply(&store, "t1", Some("uid_1"), "alreadyhere", 10_000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 1000, 0, 0, 5, 10, None)
.unwrap();
let promoted = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert!(promoted.is_empty(), "should not re-promote watchlist members");
}
#[test]
fn find_hot_reply_targets_promotes_on_profile_clicks_alone() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "clickytarget", 2000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 10, 0, 0, 2, 5, None)
.unwrap();
let promoted = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert_eq!(promoted.len(), 1);
assert_eq!(promoted[0].username, "clickytarget");
}
#[test]
fn find_hot_reply_targets_promotes_on_reply_back() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "replier", 5000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 10, 0, 0, 0, 3, None)
.unwrap();
let action_id: i64 = store
.conn
.query_row(
"SELECT id FROM engagement_actions WHERE reply_tweet_id = 'r1'",
[],
|r| r.get(0),
)
.unwrap();
store.set_reply_back(action_id, true).unwrap();
let promoted = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert_eq!(promoted.len(), 1);
assert_eq!(promoted[0].username, "replier");
}
#[test]
fn find_hot_reply_targets_skips_stale_replies() {
let store = test_store();
store
.log_post("r1", "old reply body", "text", Some("t1"), None, None, None, None)
.unwrap();
let old_ts = chrono::Utc::now().timestamp() - 30 * 24 * 3600;
store
.conn
.execute(
"INSERT INTO engagement_actions
(action_type, target_tweet_id, target_user_id, target_username,
performed_at, target_followers, reply_tweet_id, reply_style)
VALUES ('reply', 't1', NULL, LOWER('oldtarget'), ?1, 5000, 'r1', NULL)",
params![old_ts],
)
.unwrap();
store
.log_metric_snapshot("r1", 0, 0, 0, 10_000, 0, 0, 10, 30, None)
.unwrap();
let promoted = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert!(promoted.is_empty(), "should skip replies older than max_age_hours");
}
#[test]
fn rank_hot_reply_targets_aggregates_by_username() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "target_a", 3000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 100, 0, 0, 1, 5, None)
.unwrap();
log_outgoing_reply(&store, "t2", None, "target_a", 3000, "r2");
store
.log_metric_snapshot("r2", 0, 0, 0, 300, 0, 0, 3, 5, None)
.unwrap();
let ranked = store.rank_hot_reply_targets(7, 1, 0.0, 0.0).unwrap();
assert_eq!(ranked.len(), 1);
assert_eq!(ranked[0].username, "target_a");
assert_eq!(ranked[0].sample_count, 2);
assert!((ranked[0].avg_impressions - 200.0).abs() < 0.01);
assert!((ranked[0].avg_profile_clicks - 2.0).abs() < 0.01);
}
#[test]
fn rank_hot_reply_targets_respects_min_samples() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "target_a", 3000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 1000, 0, 0, 5, 5, None)
.unwrap();
let ranked = store.rank_hot_reply_targets(7, 2, 0.0, 0.0).unwrap();
assert!(ranked.is_empty(), "single sample should be filtered by min_samples=2");
}
#[test]
fn rank_hot_reply_targets_computes_reply_back_rate() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "target_a", 3000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 100, 0, 0, 0, 5, None)
.unwrap();
let id1: i64 = store
.conn
.query_row(
"SELECT id FROM engagement_actions WHERE reply_tweet_id = 'r1'",
[],
|r| r.get(0),
)
.unwrap();
store.set_reply_back(id1, true).unwrap();
log_outgoing_reply(&store, "t2", None, "target_a", 3000, "r2");
store
.log_metric_snapshot("r2", 0, 0, 0, 100, 0, 0, 0, 5, None)
.unwrap();
let id2: i64 = store
.conn
.query_row(
"SELECT id FROM engagement_actions WHERE reply_tweet_id = 'r2'",
[],
|r| r.get(0),
)
.unwrap();
store.set_reply_back(id2, false).unwrap();
let ranked = store.rank_hot_reply_targets(7, 1, 0.0, 0.0).unwrap();
assert_eq!(ranked.len(), 1);
assert!((ranked[0].reply_back_rate - 0.5).abs() < 0.01);
}
#[test]
fn hot_reply_targets_end_to_end_promotion_loop() {
let store = test_store();
log_outgoing_reply(&store, "t1", Some("uid_new"), "newhot", 5000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 500, 0, 0, 3, 15, None)
.unwrap();
let hot = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert_eq!(hot.len(), 1);
for row in &hot {
store
.add_watchlist(&row.username, row.user_id.as_deref(), None, row.target_followers)
.unwrap();
}
let wl = store.list_watchlist().unwrap();
assert_eq!(wl.len(), 1);
assert_eq!(wl[0].username, "newhot");
assert_eq!(wl[0].user_id.as_deref(), Some("uid_new"));
assert_eq!(wl[0].followers, 5000);
assert!(wl[0].topic.is_none(), "auto-promotion should preserve NULL topic");
let hot_again = store
.find_hot_reply_targets(100, 1, 1_000, 24 * 14)
.unwrap();
assert!(hot_again.is_empty());
}
#[test]
fn rank_hot_reply_targets_applies_days_filter() {
let store = test_store();
log_outgoing_reply(&store, "t1", None, "fresh_target", 3000, "r1");
store
.log_metric_snapshot("r1", 0, 0, 0, 1000, 0, 0, 5, 5, None)
.unwrap();
store
.log_post("r2", "stale reply body", "text", Some("t2"), None, None, None, None)
.unwrap();
let old_ts = chrono::Utc::now().timestamp() - 20 * 24 * 3600;
store
.conn
.execute(
"INSERT INTO engagement_actions
(action_type, target_tweet_id, target_user_id, target_username,
performed_at, target_followers, reply_tweet_id, reply_style)
VALUES ('reply', 't2', NULL, LOWER('stale_target'), ?1, 3000, 'r2', NULL)",
params![old_ts],
)
.unwrap();
store
.log_metric_snapshot("r2", 0, 0, 0, 1000, 0, 0, 5, 5, None)
.unwrap();
let ranked = store.rank_hot_reply_targets(7, 1, 0.0, 0.0).unwrap();
assert_eq!(ranked.len(), 1);
assert_eq!(ranked[0].username, "fresh_target");
}
}