Quick Actions
System Health
Checking services…
Recent Activity
Loading…
What it does
The Mission
Treelance makes outbound calls to candidates in the Trees OS talent network.
It acts as a warm, curious AI connector — not a screener. It listens far more
than it talks, follows emotional energy, and builds a picture of who the
candidate really is, so matching becomes faster and more relevant.
Identity: Who is Treelance?
The agent introduces itself honestly as an AI and represents Trees OS.
It holds a warm, unhurried conversation — never scripted, never extractive.
Warm & Curious
Honest about being AI
Speaks 20% — Listens 80%
Multilingual
Barge-in support
Tech Stack
| Language | TypeScript (Node.js) |
| Telephony | Twilio — outbound calls |
| STT | Deepgram Nova-2 |
| LLM | Gemini 2.5 Flash |
| TTS | Cartesia |
| Database | Supabase (PostgreSQL) |
| Calendar | Google Calendar API |
| Session Memory | Redis |
| Post-processing | Gemini batch pipeline |
Live Configuration
Loading provider config…
Call Volume
Total Calls
—
all time
This Week
—
Mon – today
Last Week
—
previous week
Trend
—
week over week
Daily Activity — last 14 days
Channel Breakdown
Loading…
Summary
Sim Runs
0
this session
Server
—
status
System Configuration
Loading…
Call History
Transcripts
Every completed call with its full transcript and AI summary. Click any row to expand.
Loading transcripts…
Behavioural Design
The Psychology Behind Treelance
Every response the agent gives is shaped by evidence-based conversational psychology.
These are the invisible tools built into Treelance's system prompt — the techniques
that make candidates feel genuinely heard and comfortable sharing more than they planned.
Outbound Calls
Calls
Call a candidate from the database or make a manual call to any number for testing.
Manual Call
Call Any Number
For testing — enter a name and phone number directly.
Name
Phone Number
Candidates
Call from Database
Select a candidate from your talent pool.
Search
Select Candidate
Call History
Filter by
Simulation Lab
Test the Agent
Run LLM-only simulations against pre-built candidate personas or a custom scenario.
Each simulation scores the agent on coverage, violations, and conversation quality.
No phone call required — the LLM plays both sides.
Select a scenario
Running…
Prompt Editor
System Prompt
Edit the prompt that drives every call. Changes are saved instantly and used on the next call — no code editing needed.
Master Prompt
The full core instructions sent to the AI on every call. Use
{{DATE_TIME}} as a placeholder — it gets replaced with the live Malaysia time automatically.Secondary Prompt
Extra instructions appended after the master. Use this to quickly tweak behaviour or experiment — without touching the master.