🟢 *Meridian — Main Branch Update*

Meridian is more than a pool screener now. It's a Telegram-native DLMM operator you can actually talk to.

*What's new*

*Management* — obvious close/claim decisions are handled in code first. less LLM waste, more deterministic outcomes.

*Exit logic* — trailing take-profit, low-yield closes, cleaner out-of-range handling, better post-close accounting.

*Screening* — token age filters, deployer blocklist, OKX enrichment, ATH distance checks, wash-trading hard filter, rugpull flags as caution context.

*Config & setup* — cleaner overall. user-config.example.json now clearly documents available settings.

---

*What you can do from Telegram*

— check wallet, balances, positions, fees, recent performance
— close positions, claim fees, set notes, swap tokens
— analyze pools and tokens before deploying
— find best candidates or deploy into a specific pool
— track smart wallets and check if they're in a pool
— blacklist tokens and block deployer wallets
— study top LPers and build lessons over time
— tune thresholds and risk parameters from chat
— update the bot from chat

*Built-in commands*
/briefing /positions /close <n> /set <n> <note>

Everything else is natural language.

---

*Important*

Every user runs their own Meridian instance — your own bot token, wallet, config, memory, blacklist, smart wallet list, and lessons. This is not a shared bot. It's your own operator.

*Net effect:* fewer bad entries, better exit discipline, better research context, and a much more usable Telegram workflow overall.

jangan lupa git pull, kalo ada error pas mulai coba npm install dulu

been fixing minor bug that had huge impact, pls update, both main experimental now is side-by-side or it's basically the same things

notable update adalah saran dari temen-temen kemaren buat filtering token washtrade, karena gua pake apinya okx, washtrade bakal ke hardfilter atau 100%

tapi menariknya di sini ada 1 risk yang gua filter yaitu rugpull, kalo di okx kalian bisa cek tab risk terus beberapa token ada tulisan rugpull

nah ini sebenernya agak false alarm karena banyak token kaya chicky, abcdefgh dan beberapa token kefilter, tapi rugpull ini bukan hardfilter.

rugpull risk ini diinclude ke data top candidates, dengan instruksi kalo rugpull tanpa smart wallet = skip

jadi saran saya cari smart wallet sebanyak banyaknya di lpagent caranya :

- cek pool token runner
- ke lpagent.io
- search pool
- pergi ke top lpers
- dan chat ke agent kalian buat add this smartwallet <wallet> <label>
- atau kalian juga bisa paste pool ke agent kalian dan bilang study top lpers and add to smart wallet list (harus punya lpagent api key)

contoh token bener yang diflag rugpull sama okx

kalo washtrading gini langsung hardfilter alias llm ga bakal liat token dengan flag washtrading di top candidates

added new feature, now the log on vps is streamed live on telegram

with this you can know whether your agent is still doin something

do your job or just hallucinating

reporting something without tool calls = halu

🧪 *Experimental Update — PVP Guard*

Meridian can now detect PVP setups — where the same symbol is running on a competing mint with a real pool eating into liquidity.

*What counts as a real rival*
— same exact symbol
— different mint
— holders ≥ 500
— Jupiter fees ≥ 30
— DLMM pool with TVL > 5k

Dead clones and low-traction duplicates don't trigger it.

*Three modes*

avoidPvpSymbols: false + blockPvpSymbols: false
→ ignore PVP completely

avoidPvpSymbols: true + blockPvpSymbols: false
→ warn only

avoidPvpSymbols: true + blockPvpSymbols: true
→ hard block

*How it runs*

Normal screening runs first. Only the top 2 finalists get checked for real PVP rivals — no extra overhead on the full candidate list.

*Net effect:* better protection against split-liquidity traps, no noise from dead clones, and you control how aggressive the guard is.

buat temen temen yang baru set up tolong minimal threshold tvl/organic score minimal di set

di user-config.json/minta agent kalian buat ganti
min tvl bisa 5k/10k
organic score di 60 keatas

agar kejadian agent maksa ke pool yang kosong gini ga keulang.

mending kalian belajar dlmm dulu deh

discord.gg/meteoraidn

🟢🧪 *Meridian Update — Experimental + Main*

*Repeated OOR protection*
If a pool closes out-of-range 3 times in a row, Meridian puts it on cooldown. The cooldown also applies to the token mint — so the agent won't reopen the same token through a different pool.

Defaults:
oorCooldownTriggerCount: 3
oorCooldownHours: 12

Both are configurable.

*Dry-run improvement*
Dry-run no longer requires real SOL balance. You can screen and simulate deploys with an empty wallet without hitting the balance guard.

*Provider compatibility fix*
Some LLM providers reject role: system and caused errors like invalid message role: system (2013). Meridian now retries automatically with a compatibility fallback instead of failing the request.

*Security hardening*
— no more auto-registration on first Telegram message
— bot control now requires a configured chat ID + allowed user IDs
— self-update and config mutation paths are tightened
— narrative and memory text from untrusted sources is treated as hostile prompt data

*Net effect:* less re-entry into bad OOR tokens, easier dry-run testing, more stable replies across providers, and a safer bot overall.

jujur aja kalo kalian kalo profit gua ga dapet apa-apa, gua ga ngasih tip/fee ke wallet gua pribadi di codebase.

kalo kalian masih tanya hal-hal basic mending tanyain ke claude/codex.

kalo kalian ngerti cara pakenya kalian bakal opit dan vice versa.

good luck all, jangan lupa join discussion group ada @meridianhelper_bot yang siap bantu jawab pertanyaan "itu-itu" aja kalian.

discussion group

https://t.me/+0uBDfKzcm5Y2M2Y1

🟢 *Meridian Update — Trailing TP Double-Check*

Meridian now confirms both sides of a trailing TP decision before acting.

*The problem*
A single bad API tick could set a fake peak, or trigger an early close on a fake crash.

*What changed*
— if PnL suddenly spikes, Meridian waits 15s and re-checks before accepting it as the real peak
— if trailing TP says close, Meridian waits 15s and re-checks before actually closing

*Example 1 — fake peak*
Real PnL is around -5%. One tick reports 29%. Meridian doesn't trust it. 15s later it's back near -5% — that peak is rejected.

*Example 2 — fake crash*
Confirmed peak is 10%. One tick reports -8%. Instead of closing instantly, Meridian waits 15s. Next read comes back at 7% — no close.

*Net effect:* don't close from one fake spike. don't close from one fake crash.

🟢🧪 *Meridian Update — PM2 Telegram Fix*

If you run Meridian under PM2, Telegram now works both ways.

*What was broken*
PM2 users received outbound notifications fine — deploys, closes, alerts. But messages sent to the agent never got through. The agent couldn't hear you.

*Why*
Meridian only started inbound Telegram polling in TTY mode. PM2 runs in non-TTY, so polling never started.

*What changed*
Inbound polling now starts in both modes — normal terminal and PM2.

*What works now under PM2*
— free-text requests to the agent
— /positions, /close, /set and other commands
— live tool progress and final replies
— outbound notifications unchanged

Before: one-way. Now: both ways.

cara update tinggal bilang update ke tele

config ga bakal ilang

halo temen-temen, dikarenakan Meridian akan mengikuti hackathon frontier,

sebagai bentuk apresiasi kalian terhadap pembuatan framework ini bisa difollow akun meridian di X ya

https://x.com/meridian_agent

🟢🧪 Meridian Update — LPAgent Integration

LPAgent is now the primary source for live position economics when LPAGENT_API_KEY is set.

What LPAgent handles
— current value
— PnL and PnL %
— fee values

What Meteora still handles
— open position discovery
— bin/range data
— OOR checks
— age and yield-related management fields

Why it matters
LPAgent data has been more reliable than Meteora for live PnL/value. Management decisions now run on better economics data while Meteora handles protocol and range state.

If LPAGENT_API_KEY is not set, Meridian falls back to the old Meteora-only path automatically.

How to get an API key
— go to https://portal.lpagent.io
— register with a Gmail account
— generate your key from the API section

How to wire it in
Add one line to your .env:
LPAGENT_API_KEY=lpagent_your_api_key_here

No other config change needed.

With key: LPAgent-backed live PnL. Without key: Meteora fallback.

udah seminggu sejak meridian open-source dan bisa diakses banyak orang, dan udah sekitar 3 minggu sejak pertama rilis, banyak trial & error sebelum gua open ke public, terimakasih untuk beta tester yang sudah memberikan feedback.

gua minta tolong kalian yang udah berproses buat *flex* win atau lossnya di discussion group buat nyemangatin atau sekedar ngasih tau kalo pake ai agent ngga sesusah itu atau ngga segampang itu.

mulai dari gua, setelah update 29 maret gua ngerasa agent gua udah bisa gua tinggal, banyak fitur security guard yang udah meningkat berkat feedback temen temen beta testers.

sekarang agent gua udah gua lepas tangan, 100% lepas tangan.

terimakasih temen-temen yang sudah mau nyoba meridian ini.