672 lines
27 KiB
Rust
672 lines
27 KiB
Rust
use axum::{
|
||
extract::State,
|
||
http::{HeaderMap, HeaderValue, Method, StatusCode},
|
||
routing::{get, post},
|
||
Json, Router,
|
||
};
|
||
use tower_http::cors::{CorsLayer, AllowOrigin, Any};
|
||
use jsonwebtoken::{decode, DecodingKey, Validation, Algorithm};
|
||
use serde::{Deserialize, Serialize};
|
||
use sqlx::postgres::PgPoolOptions;
|
||
use std::collections::HashMap;
|
||
use std::fs;
|
||
use std::sync::Arc;
|
||
use tokio::sync::Mutex;
|
||
use tracing::{error, info, warn};
|
||
use uuid::Uuid;
|
||
use rand::Rng;
|
||
use chrono::Timelike;
|
||
use std::time::Duration;
|
||
use std::time::{SystemTime, UNIX_EPOCH};
|
||
|
||
// ── JWT Claims ────────────────────────────────────────────────
|
||
|
||
#[derive(Debug, Deserialize)]
|
||
struct Claims {
|
||
#[serde(rename = "userId")]
|
||
user_id: String,
|
||
role: String,
|
||
}
|
||
|
||
fn verify_jwt(token: &str, secret: &str) -> Option<Claims> {
|
||
let key = DecodingKey::from_secret(secret.as_bytes());
|
||
let validation = Validation::new(Algorithm::HS256);
|
||
decode::<Claims>(token, &key, &validation).ok().map(|d| d.claims)
|
||
}
|
||
|
||
// ── Rate limiter ──────────────────────────────────────────────
|
||
|
||
struct RateLimiter {
|
||
buckets: Mutex<HashMap<String, Vec<u64>>>,
|
||
max_requests: usize,
|
||
window_secs: u64,
|
||
}
|
||
|
||
impl RateLimiter {
|
||
fn new(max_requests: usize, window_secs: u64) -> Self {
|
||
Self {
|
||
buckets: Mutex::new(HashMap::new()),
|
||
max_requests,
|
||
window_secs,
|
||
}
|
||
}
|
||
|
||
async fn check(&self, key: &str) -> bool {
|
||
let now = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
|
||
let mut buckets = self.buckets.lock().await;
|
||
let timestamps = buckets.entry(key.to_string()).or_default();
|
||
timestamps.retain(|t| now - *t <= self.window_secs);
|
||
if timestamps.len() >= self.max_requests {
|
||
false
|
||
} else {
|
||
timestamps.push(now);
|
||
true
|
||
}
|
||
}
|
||
}
|
||
|
||
// ── Shared state ───────────────────────────────────────────────
|
||
|
||
struct AppState {
|
||
db: sqlx::PgPool,
|
||
ollama_url: String,
|
||
model: String,
|
||
jobs: Vec<Job>,
|
||
leads: Arc<Mutex<LeadStore>>,
|
||
http_client: reqwest::Client,
|
||
jwt_secret: String,
|
||
rate_limiter: RateLimiter,
|
||
}
|
||
|
||
#[derive(Debug, Clone, Serialize)]
|
||
struct Lead {
|
||
title: String,
|
||
url: String,
|
||
source: String,
|
||
found_at: u64,
|
||
author: String,
|
||
date: String,
|
||
content: String,
|
||
}
|
||
|
||
struct LeadStore {
|
||
leads: Vec<Lead>,
|
||
max_size: usize,
|
||
}
|
||
|
||
#[derive(Debug, Deserialize)]
|
||
struct ScrapeResponse {
|
||
success: bool,
|
||
leads: Vec<ScrapeLead>,
|
||
flagged: bool,
|
||
flag_reason: Option<String>,
|
||
error: Option<String>,
|
||
}
|
||
|
||
#[derive(Debug, Deserialize, Clone)]
|
||
struct ScrapeLead {
|
||
title: String,
|
||
url: String,
|
||
author: String,
|
||
date: String,
|
||
content: String,
|
||
source: Option<String>,
|
||
}
|
||
|
||
impl LeadStore {
|
||
fn new(max_size: usize) -> Self {
|
||
Self { leads: Vec::new(), max_size }
|
||
}
|
||
|
||
fn push(&mut self, lead: Lead) {
|
||
if !self.leads.iter().any(|l| l.url == lead.url) {
|
||
self.leads.insert(0, lead);
|
||
self.leads.truncate(self.max_size);
|
||
}
|
||
}
|
||
|
||
fn recent(&self, max_age_secs: u64, limit: usize) -> Vec<Lead> {
|
||
let now = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
|
||
self.leads.iter()
|
||
.filter(|l| now.saturating_sub(l.found_at) <= max_age_secs)
|
||
.take(limit)
|
||
.cloned()
|
||
.collect()
|
||
}
|
||
}
|
||
|
||
#[derive(Debug, Clone, Serialize, Deserialize)]
|
||
struct Job {
|
||
job_title: String,
|
||
keywords: Vec<String>,
|
||
industry: String,
|
||
description: String,
|
||
}
|
||
|
||
#[derive(Debug, Deserialize)]
|
||
struct ChatRequest {
|
||
message: String,
|
||
}
|
||
|
||
#[derive(Debug, Serialize)]
|
||
struct ChatResponse {
|
||
response: String,
|
||
}
|
||
|
||
#[derive(Debug, Serialize)]
|
||
struct JobsResponse {
|
||
jobs: Vec<Job>,
|
||
}
|
||
|
||
#[derive(Debug, Serialize)]
|
||
struct HealthResponse {
|
||
status: String,
|
||
model: String,
|
||
}
|
||
|
||
// ── Ollama API types ───────────────────────────────────────────
|
||
|
||
#[derive(Debug, Serialize)]
|
||
struct OllamaChatMessage {
|
||
role: String,
|
||
content: String,
|
||
}
|
||
|
||
#[derive(Debug, Serialize)]
|
||
struct OllamaRequest {
|
||
model: String,
|
||
messages: Vec<OllamaChatMessage>,
|
||
stream: bool,
|
||
options: OllamaOptions,
|
||
}
|
||
|
||
#[derive(Debug, Serialize)]
|
||
struct OllamaOptions {
|
||
temperature: f32,
|
||
num_predict: u32,
|
||
}
|
||
|
||
#[derive(Debug, Deserialize)]
|
||
struct OllamaResponse {
|
||
message: Option<OllamaResponseMessage>,
|
||
}
|
||
|
||
#[derive(Debug, Deserialize)]
|
||
struct OllamaResponseMessage {
|
||
content: String,
|
||
}
|
||
|
||
// ── Helpers ────────────────────────────────────────────────────
|
||
|
||
fn truncate(s: &str, max: usize) -> String {
|
||
s.chars().take(max).collect()
|
||
}
|
||
|
||
fn extract_claims(headers: &HeaderMap, state: &AppState) -> Result<Claims, (StatusCode, String)> {
|
||
let auth_header = headers.get("Authorization").and_then(|v| v.to_str().ok()).unwrap_or("");
|
||
let token = auth_header.strip_prefix("Bearer ").unwrap_or("");
|
||
let claims = verify_jwt(token, &state.jwt_secret).ok_or_else(|| {
|
||
(StatusCode::UNAUTHORIZED, "Unauthorized".to_string())
|
||
})?;
|
||
match claims.role.to_lowercase().as_str() {
|
||
"sales" | "admin" | "super_admin" => Ok(claims),
|
||
_ => Err((StatusCode::FORBIDDEN, "Forbidden".to_string())),
|
||
}
|
||
}
|
||
|
||
fn format_leads_output(leads: &[Lead]) -> String {
|
||
if leads.is_empty() {
|
||
return "No new requests found yet.".to_string();
|
||
}
|
||
leads
|
||
.iter()
|
||
.enumerate()
|
||
.map(|(i, l)| {
|
||
let author = if l.author.is_empty() { "Unknown" } else { &l.author };
|
||
let date = truncate(&l.date, 10);
|
||
format!(
|
||
"{}. {}\n {}\n {}\n {}",
|
||
i + 1,
|
||
author,
|
||
date,
|
||
l.title,
|
||
l.url
|
||
)
|
||
})
|
||
.collect::<Vec<_>>()
|
||
.join("\n")
|
||
}
|
||
|
||
fn build_system_prompt(jobs: &[Job], leads: &[Lead]) -> String {
|
||
let job_list: Vec<String> = jobs
|
||
.iter()
|
||
.map(|j| format!("- {} ({}): {}", j.job_title, j.industry, j.description))
|
||
.collect();
|
||
let job_list_str = job_list.join("\n");
|
||
|
||
let lead_summary: Vec<String> = leads
|
||
.iter()
|
||
.map(|l| format!("{} | {} | {}", l.author, l.title, l.url))
|
||
.collect();
|
||
let lead_summary_str = if lead_summary.is_empty() {
|
||
"None yet.".to_string()
|
||
} else {
|
||
lead_summary.join("\n")
|
||
};
|
||
|
||
format!(
|
||
"You are a Sales AI Assistant for Coast IT CRM.\n\n\
|
||
Available job categories to target:\n{}\n\n\
|
||
Recent leads context:\n{}\n\n\
|
||
Rules:\n\
|
||
- When asked about leads, answer concisely under 150 words.\n\
|
||
- If asked to suggest a sales strategy, give brief actionable advice.\n\
|
||
- Be direct and professional. No fluff.",
|
||
job_list_str, lead_summary_str
|
||
)
|
||
}
|
||
|
||
// ── Chat handler ───────────────────────────────────────────────
|
||
|
||
async fn handle_chat(
|
||
State(state): State<Arc<AppState>>,
|
||
headers: HeaderMap,
|
||
Json(req): Json<ChatRequest>,
|
||
) -> Result<Json<ChatResponse>, (StatusCode, String)> {
|
||
let claims = extract_claims(&headers, &state)?;
|
||
|
||
if !state.rate_limiter.check(&claims.user_id).await {
|
||
return Err((StatusCode::TOO_MANY_REQUESTS, "Rate limit exceeded".to_string()));
|
||
}
|
||
|
||
let msg_lower = req.message.to_lowercase();
|
||
let msg_words: Vec<&str> = msg_lower.split_whitespace().collect();
|
||
let has_listing = msg_words.iter().any(|w| ["listings", "listing", "leads", "links", "lists"].contains(w));
|
||
let has_show = msg_lower.contains("show me") || msg_lower.contains("give me") || msg_lower.contains("pull");
|
||
let has_job = msg_words.contains(&"jobs") || msg_words.contains(&"job");
|
||
if has_listing || (has_show && has_job) || (has_show && msg_lower.contains("links")) || msg_lower.contains("recent leads") {
|
||
let now = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
|
||
let base_url = "http://localhost:3008/scrape/facebook";
|
||
use std::fmt::Write;
|
||
let mut service_url = base_url.to_string();
|
||
if let Ok(Some((_, path))) = sqlx::query_as::<_, (uuid::Uuid, String)>(
|
||
"SELECT id, profile_path FROM facebook_accounts \
|
||
WHERE is_active = TRUE AND flagged = FALSE \
|
||
ORDER BY last_scrape_at ASC NULLS FIRST LIMIT 1"
|
||
)
|
||
.fetch_optional(&state.db)
|
||
.await
|
||
{
|
||
let encoded: String = path.chars().map(|c| match c {
|
||
'A'..='Z' | 'a'..='z' | '0'..='9' | '.' | '-' | '_' | '~' => c.to_string(),
|
||
_ => format!("%{:02X}", c as u8),
|
||
}).collect();
|
||
write!(service_url, "?profile_path={}&force=true", encoded).unwrap();
|
||
info!("Calling Python scrape at: {}?profile_path=...&force=true", base_url);
|
||
} else {
|
||
warn!("No active Facebook account found for on-demand scrape");
|
||
}
|
||
let req_builder = state.http_client.post(&service_url);
|
||
|
||
match req_builder.send().await {
|
||
Ok(resp) => {
|
||
let status = resp.status();
|
||
let body = resp.text().await.unwrap_or_default();
|
||
info!("Python scrape response ({}): {} bytes", status, body.len());
|
||
if body.starts_with('{') {
|
||
match serde_json::from_str::<ScrapeResponse>(&body) {
|
||
Ok(scrape_resp) => {
|
||
info!("Scraped {} leads from Facebook", scrape_resp.leads.len());
|
||
if scrape_resp.leads.is_empty() && scrape_resp.error.is_some() {
|
||
warn!("Python returned error: {:?}", scrape_resp.error);
|
||
}
|
||
let mut store = state.leads.lock().await;
|
||
for item in &scrape_resp.leads {
|
||
store.push(Lead {
|
||
title: truncate(&item.title, 120),
|
||
url: item.url.clone(),
|
||
source: item.source.clone().unwrap_or_else(|| "facebook".to_string()),
|
||
found_at: now,
|
||
author: truncate(&item.author, 60),
|
||
date: truncate(&item.date, 30),
|
||
content: truncate(&item.content, 300),
|
||
});
|
||
}
|
||
}
|
||
Err(e) => {
|
||
warn!("Failed to parse Python scrape response: {} body: {}", e, &body[..body.len().min(200)]);
|
||
}
|
||
}
|
||
} else {
|
||
warn!("Python returned non-JSON response: {}", &body[..body.len().min(200)]);
|
||
}
|
||
}
|
||
Err(e) => {
|
||
error!("Scraper request error: {} - URL: {}", e, base_url);
|
||
}
|
||
}
|
||
|
||
let recent_leads = state.leads.lock().await.recent(604800, 20);
|
||
let response = format_leads_output(&recent_leads);
|
||
let _ = sqlx::query(
|
||
"INSERT INTO ai_conversations (id, user_id, role, message, response) VALUES ($1, $2, $3, $4, $5)",
|
||
)
|
||
.bind(Uuid::new_v4())
|
||
.bind(Uuid::parse_str(&claims.user_id).unwrap_or(Uuid::nil()))
|
||
.bind(&claims.role)
|
||
.bind(&req.message)
|
||
.bind(&response)
|
||
.execute(&state.db)
|
||
.await;
|
||
return Ok(Json(ChatResponse { response }));
|
||
}
|
||
|
||
let recent_leads = state.leads.lock().await.recent(604800, 20);
|
||
let system_prompt = build_system_prompt(&state.jobs, &recent_leads);
|
||
let message_text = req.message.clone();
|
||
|
||
let ollama_req = OllamaRequest {
|
||
model: state.model.clone(),
|
||
messages: vec![
|
||
OllamaChatMessage { role: "system".to_string(), content: system_prompt },
|
||
OllamaChatMessage { role: "user".to_string(), content: req.message.clone() },
|
||
],
|
||
stream: false,
|
||
options: OllamaOptions { temperature: 0.7, num_predict: 1024 },
|
||
};
|
||
|
||
let resp = state.http_client
|
||
.post(format!("{}/api/chat", state.ollama_url))
|
||
.json(&ollama_req)
|
||
.send()
|
||
.await
|
||
.map_err(|e| {
|
||
error!("Ollama request failed: {}", e);
|
||
(StatusCode::SERVICE_UNAVAILABLE, "AI service unavailable".to_string())
|
||
})?;
|
||
|
||
let ollama_resp: OllamaResponse = resp.json().await.map_err(|e| {
|
||
error!("Failed to parse Ollama response: {}", e);
|
||
(StatusCode::SERVICE_UNAVAILABLE, "AI response parse error".to_string())
|
||
})?;
|
||
|
||
let response_text = ollama_resp.message.map(|m| m.content).unwrap_or_default();
|
||
|
||
let _ = sqlx::query(
|
||
"INSERT INTO ai_conversations (id, user_id, role, message, response) VALUES ($1, $2, $3, $4, $5)",
|
||
)
|
||
.bind(Uuid::new_v4())
|
||
.bind(Uuid::parse_str(&claims.user_id).unwrap_or(Uuid::nil()))
|
||
.bind(&claims.role)
|
||
.bind(&message_text)
|
||
.bind(&response_text)
|
||
.execute(&state.db)
|
||
.await;
|
||
|
||
Ok(Json(ChatResponse { response: response_text }))
|
||
}
|
||
|
||
// ── Jobs handler ───────────────────────────────────────────────
|
||
|
||
async fn handle_jobs(
|
||
State(state): State<Arc<AppState>>,
|
||
headers: HeaderMap,
|
||
) -> Result<Json<JobsResponse>, (StatusCode, String)> {
|
||
let _claims = extract_claims(&headers, &state)?;
|
||
if !state.rate_limiter.check(&_claims.user_id).await {
|
||
return Err((StatusCode::TOO_MANY_REQUESTS, "Rate limit exceeded".to_string()));
|
||
}
|
||
Ok(Json(JobsResponse { jobs: state.jobs.clone() }))
|
||
}
|
||
|
||
// ── Health handler ─────────────────────────────────────────────
|
||
|
||
async fn handle_health(
|
||
State(state): State<Arc<AppState>>,
|
||
) -> Json<HealthResponse> {
|
||
Json(HealthResponse {
|
||
status: "ok".to_string(),
|
||
model: state.model.clone(),
|
||
})
|
||
}
|
||
|
||
// ── Main ───────────────────────────────────────────────────────
|
||
|
||
#[tokio::main]
|
||
async fn main() {
|
||
tracing_subscriber::fmt()
|
||
.with_env_filter(
|
||
tracing_subscriber::EnvFilter::try_from_default_env()
|
||
.unwrap_or_else(|_| "crm_ai=info,tower_http=info".into()),
|
||
)
|
||
.init();
|
||
|
||
dotenvy::dotenv().ok();
|
||
|
||
let database_url = std::env::var("DATABASE_URL").expect("DATABASE_URL must be set");
|
||
let jwt_secret = std::env::var("JWT_SECRET").expect("JWT_SECRET must be set");
|
||
let ollama_url = std::env::var("OLLAMA_BASE_URL").unwrap_or_else(|_| "http://localhost:11434".to_string());
|
||
let model = std::env::var("AI_MODEL").unwrap_or_else(|_| "dolphin-phi".to_string());
|
||
let host = std::env::var("AI_HOST").unwrap_or_else(|_| "127.0.0.1".to_string());
|
||
let port: u16 = std::env::var("AI_PORT").unwrap_or_else(|_| "3001".to_string()).parse().expect("AI_PORT must be a number");
|
||
|
||
let jobs_path = std::env::var("JOBS_PATH").unwrap_or_else(|_| "data/ai/jobs.jsonl".to_string());
|
||
let jobs_content = fs::read_to_string(&jobs_path).unwrap_or_default();
|
||
let jobs: Vec<Job> = jobs_content.lines().filter(|l| !l.trim().is_empty()).filter_map(|l| serde_json::from_str(l).ok()).collect();
|
||
|
||
info!("Loaded {} job categories, model: {}, Ollama: {}", jobs.len(), model, ollama_url);
|
||
|
||
let db = PgPoolOptions::new()
|
||
.max_connections(20)
|
||
.connect(&database_url)
|
||
.await
|
||
.expect("Failed to connect to database");
|
||
|
||
info!("Connected to PostgreSQL");
|
||
|
||
let http_client = reqwest::Client::builder()
|
||
.timeout(Duration::from_secs(300))
|
||
.build()
|
||
.expect("Failed to build HTTP client");
|
||
|
||
let lead_store = Arc::new(Mutex::new(LeadStore::new(100)));
|
||
|
||
let state = Arc::new(AppState {
|
||
db,
|
||
ollama_url,
|
||
model,
|
||
jobs,
|
||
leads: lead_store.clone(),
|
||
http_client,
|
||
jwt_secret,
|
||
rate_limiter: RateLimiter::new(30, 60),
|
||
});
|
||
|
||
let cors_origins_env = std::env::var("CORS_ORIGINS").unwrap_or_else(|_| "http://localhost:3006,http://127.0.0.1:3006".to_string());
|
||
let cors_origins: Vec<HeaderValue> = cors_origins_env.split(',')
|
||
.filter_map(|o| { let t = o.trim(); if t.is_empty() { None } else { t.parse().ok() } })
|
||
.collect();
|
||
let cors = CorsLayer::new()
|
||
.allow_origin(AllowOrigin::list(cors_origins))
|
||
.allow_methods([Method::GET, Method::POST])
|
||
.allow_headers(Any);
|
||
|
||
let app = Router::new()
|
||
.route("/ai/chat", post(handle_chat))
|
||
.route("/ai/jobs", get(handle_jobs))
|
||
.route("/health", get(handle_health))
|
||
.layer(cors)
|
||
.with_state(state.clone());
|
||
|
||
let addr = format!("{}:{}", host, port);
|
||
info!("CRM AI server listening on {}", addr);
|
||
|
||
let listener = tokio::net::TcpListener::bind(&addr)
|
||
.await
|
||
.expect("Failed to bind address");
|
||
|
||
let bg_leads = lead_store.clone();
|
||
let bg_db = state.db.clone();
|
||
let bg_url = std::env::var("SCRAPER_URL").unwrap_or_else(|_| "http://localhost:3008".to_string()) + "/scrape/facebook";
|
||
tokio::spawn(async move {
|
||
let client = match reqwest::Client::builder()
|
||
.timeout(Duration::from_secs(300))
|
||
.build()
|
||
{
|
||
Ok(c) => c,
|
||
Err(e) => {
|
||
error!("Failed to build background HTTP client: {} — scraper disabled", e);
|
||
return;
|
||
}
|
||
};
|
||
// Initial delay to let user on-demand requests take priority
|
||
tokio::time::sleep(Duration::from_secs(300)).await;
|
||
|
||
loop {
|
||
// 10% random cycle skip — makes pattern non-periodic
|
||
if rand::thread_rng().gen_range(0..100) < 10 {
|
||
info!("Skipping this scrape cycle (random 10% skip)");
|
||
let jitter = rand::thread_rng().gen_range(16200..19800);
|
||
tokio::time::sleep(Duration::from_secs(jitter)).await;
|
||
continue;
|
||
}
|
||
|
||
// Skip night hours (23:00 – 06:00)
|
||
let hour = chrono::Local::now().hour();
|
||
if hour < 6 || hour >= 23 {
|
||
info!("Night hours ({}) — skipping scrape", hour);
|
||
let jitter = rand::thread_rng().gen_range(16200..19800);
|
||
tokio::time::sleep(Duration::from_secs(jitter)).await;
|
||
continue;
|
||
}
|
||
|
||
let now = SystemTime::now().duration_since(UNIX_EPOCH).unwrap_or_default().as_secs();
|
||
|
||
// Pick next active un-flagged account
|
||
let account = sqlx::query_as::<_, (uuid::Uuid, String)>(
|
||
"SELECT id, profile_path FROM facebook_accounts \
|
||
WHERE is_active = TRUE AND flagged = FALSE \
|
||
ORDER BY last_scrape_at ASC NULLS FIRST LIMIT 1"
|
||
)
|
||
.fetch_optional(&bg_db)
|
||
.await;
|
||
|
||
match account {
|
||
Ok(Some((account_id, profile_path))) => {
|
||
match client.post(&bg_url).query(&[("profile_path", &profile_path)]).send().await {
|
||
Ok(resp) => {
|
||
if resp.status().is_success() {
|
||
match resp.json::<ScrapeResponse>().await {
|
||
Ok(data) => {
|
||
let leads_count = data.leads.len() as i32;
|
||
if data.flagged {
|
||
let _ = sqlx::query(
|
||
"UPDATE facebook_accounts SET flagged = TRUE, flagged_at = NOW(), \
|
||
flagged_reason = $2, last_error_at = NOW(), \
|
||
last_error_message = $3, consecutive_failures = consecutive_failures + 1 \
|
||
WHERE id = $1"
|
||
)
|
||
.bind(account_id)
|
||
.bind(&data.flag_reason)
|
||
.bind(&data.error)
|
||
.execute(&bg_db)
|
||
.await;
|
||
warn!("Facebook account {} flagged: {:?}", account_id, data.flag_reason);
|
||
let reason = data.flag_reason.as_deref().unwrap_or("unknown");
|
||
let _ = sqlx::query(
|
||
"INSERT INTO notifications (user_id, type, title, description, link) \
|
||
SELECT id, 'warning', 'Facebook Account Flagged', \
|
||
$1 || ' - ' || COALESCE($2, 'unknown reason'), \
|
||
NULL \
|
||
FROM users u JOIN user_roles ur ON ur.user_id = u.id \
|
||
JOIN roles r ON r.id = ur.role_id \
|
||
WHERE r.name IN ('ADMIN', 'SUPER_ADMIN')"
|
||
)
|
||
.bind(&account_id.to_string())
|
||
.bind(reason)
|
||
.execute(&bg_db)
|
||
.await;
|
||
} else if data.success {
|
||
let _ = sqlx::query(
|
||
"UPDATE facebook_accounts SET last_scrape_at = NOW(), \
|
||
last_success_at = NOW(), consecutive_failures = 0, \
|
||
updated_at = NOW() WHERE id = $1"
|
||
)
|
||
.bind(account_id)
|
||
.execute(&bg_db)
|
||
.await;
|
||
|
||
let mut store = bg_leads.lock().await;
|
||
for item in &data.leads {
|
||
store.push(Lead {
|
||
title: truncate(&item.title, 120),
|
||
url: item.url.clone(),
|
||
source: item.source.clone().unwrap_or_else(|| "facebook".to_string()),
|
||
found_at: now,
|
||
author: truncate(&item.author, 60),
|
||
date: truncate(&item.date, 30),
|
||
content: truncate(&item.content, 300),
|
||
});
|
||
}
|
||
info!("Scraped {} leads from Facebook account {}", leads_count, account_id);
|
||
} else {
|
||
// Increment failures; auto-flag if >= 3 consecutive
|
||
let _ = sqlx::query(
|
||
"UPDATE facebook_accounts SET last_error_at = NOW(), \
|
||
last_error_message = $2, consecutive_failures = consecutive_failures + 1, \
|
||
flagged = CASE WHEN consecutive_failures + 1 >= 3 THEN TRUE ELSE flagged END, \
|
||
flagged_reason = CASE WHEN consecutive_failures + 1 >= 3 THEN 'too_many_failures' ELSE flagged_reason END, \
|
||
flagged_at = CASE WHEN consecutive_failures + 1 >= 3 THEN NOW() ELSE flagged_at END, \
|
||
updated_at = NOW() WHERE id = $1"
|
||
)
|
||
.bind(account_id)
|
||
.bind(&data.error)
|
||
.execute(&bg_db)
|
||
.await;
|
||
warn!("Facebook scrape failed for account {}: {:?}", account_id, data.error);
|
||
}
|
||
|
||
let _ = sqlx::query(
|
||
"INSERT INTO facebook_scrape_logs \
|
||
(account_id, started_at, completed_at, success, leads_found, error_message, detected_flag) \
|
||
VALUES ($1, NOW() - interval '5 hours', NOW(), $2, $3, $4, $5)"
|
||
)
|
||
.bind(account_id)
|
||
.bind(data.success && !data.flagged)
|
||
.bind(leads_count)
|
||
.bind(&data.error)
|
||
.bind(&data.flag_reason)
|
||
.execute(&bg_db)
|
||
.await;
|
||
}
|
||
Err(e) => {
|
||
warn!("Failed to parse scraper JSON: {}", e);
|
||
}
|
||
}
|
||
} else {
|
||
warn!("Scraper returned status: {}", resp.status());
|
||
}
|
||
}
|
||
Err(e) => {
|
||
warn!("Scraper request failed: {}", e);
|
||
}
|
||
}
|
||
}
|
||
Ok(None) => {
|
||
info!("No active Facebook accounts available — skipping scrape cycle");
|
||
}
|
||
Err(e) => {
|
||
warn!("Failed to query Facebook accounts: {}", e);
|
||
}
|
||
}
|
||
|
||
let jitter = rand::thread_rng().gen_range(16200..19800); // 4.5h – 5.5h
|
||
tokio::time::sleep(Duration::from_secs(jitter)).await;
|
||
}
|
||
});
|
||
|
||
axum::serve(listener, app)
|
||
.await
|
||
.expect("Server failed");
|
||
}
|