🚀 Features

← Home

---

Feature 1 — Connection Automation

Tab: Live Feed (left panel → Execute button)

Finds decision makers at target companies and sends personalized connection requests.

How it works

1. Enter target company names (one per line) in the left panel
2. Click Execute
3. Chrome opens, searches LinkedIn for people at each company
4. Each profile is scored (AI or keyword matching)
5. Connection request sent with an AI-written note
6. Logged to connections.csv

Controls

Caps & Limits

• WEEKLY_CAP=100 — tracked in the ring widget
• Configurable via .env
• No daily cap — removed

Ring Widget

Three concentric arcs show:

• Outer ring — weekly connections (vs 100 cap)
• Middle ring — today's connections (display only, no daily cap)
• Inner ring — connections sent with a message note

---

Feature 2 — First Message (Msg Prospect)

Tab: Msg Prospect

Sends AI-personalized first messages to new connections who haven't been messaged yet.

How it works

1. Click the Msg Prospect tab
2. Set Messages per Run slider (1–10)
3. Click Send Messages
4. Chrome opens, navigates to connections page
5. Scrapes visible connection cards via JavaScript DOM evaluation
6. Filters out anyone already in messages.csv
7. For each new connection:
   • Navigates directly to their compose URL
   • Ollama generates a warm, non-salesy first message
   • Types message and clicks Send
   • Logs to messages.csv
   • Seeds followups.csv with status=pending

Message Prompt Design

Write a short, friendly LinkedIn first message to {first_name},
who just connected with me. Their role is: {role}.
Rules:
- 2-3 sentences max
- Warm and genuine, not salesy
- Reference their role if known
- Do NOT mention jobs, recruiting, or opportunities
- Start with 'Hi {first_name}'
- Return ONLY the message text

Dedup Logic

• Primary: normalized profile_url match against messages.csv
• Fallback: lowercased name match (for manually-added rows)

---

Feature 3 — Follow-Up Sequence with Reply Detection

Tab: Follow Up

Sends a single gentle follow-up to connections who haven't replied after a set number of days.

How it works

1. Click the Follow Up tab
2. Set Wait before follow-up slider (0 = test mode, 1–14 days)
3. Set Follow-ups per Run slider (1–10)
4. Click Send Follow-Ups
5. followups.csv is seeded from messages.csv (idempotent — safe to run repeatedly)
6. Filters to: status = pending AND first_msg_sent_at ≤ now - wait_days
7. For each candidate:
   • Navigates to their LinkedIn profile page
   • Clicks the Message button directly on the profile
   • Compose overlay opens on top of the profile (no page jump)
   • Checks last message bubble class to detect reply
   • If replied → marks status=replied, moves on
   • If not replied → Ollama generates follow-up → types → Send
   • Updates followups.csv → status=followed_up

Reply Detection

DOM selectors checked inside the open compose overlay:
  .msg-s-event-listitem--other  →  they sent last  →  replied ✓
  .msg-s-event-listitem--self   →  we sent last    →  no reply
  (none found)                  →  unknown         →  treat as no reply

Status Values in followups.csv

Follow-Up Prompt Design

Write a short, gentle follow-up LinkedIn message to {first_name},
a {role}, who hasn't replied yet.
My original message was: "{original_message[:120]}"
Rules:
- 1-2 sentences max
- Warm, not pushy
- Do NOT mention jobs, recruiting, or opportunities
- Do NOT repeat the original message verbatim
- Start with 'Hi {first_name}'
- Return ONLY the message text

NOTE — One Follow-Up Only The system sends exactly one follow-up per connection. Once status=followed_up, that person is never contacted again automatically.

---

Common Behaviors Across All Features

5-Second Send Countdown

Before every send (first message or follow-up), the live feed shows:

Sending in 5s — click Stop to cancel...

This gives you a window to abort any individual message.

CAPTCHA Detection

If LinkedIn shows a CAPTCHA at any point, the run pauses and logs a warning. You solve it manually, then restart.

Speed Modes

AI Quote Stripping

Both message generators apply .strip().strip('"').strip("'") to remove any quote characters the AI wraps around its output.