I’m working on a voice-powered booking system that uses Retell AI to talk to callers and n8n as the backend. Retell collects all the booking details, confirms the exact UK date and time with the caller, and then sends one clean function call into n8n, for example: Book a table for four under Michael on 25/11/2025 at 2 PM. n8n’s Calendar Agent is then supposed to interpret this text and create, update, or cancel the event in Google Calendar. What’s going wrong Even though Retell now always sends explicit UK-format dates (dd/mm/yyyy) after confirming with the caller, the Calendar Agent inside n8n sometimes reinterprets that date — usually by shifting the year backwards (e.g., 2025 → 2024) or otherwise assuming a different date before creating the Google Calendar event. There are no node errors; the workflow technically “works”, but the booked date in Google Calendar is sometimes wrong unless the caller corrects it manually during the conversation. A typical example Retell sends: Book a table for four under Paul on 26/11/2025 at 11 AM. But the Calendar Agent replies with: “Do you mean Wednesday 27/11/2024?” Which is obviously not what was said. This behaviour suggests the LLM inside the Calendar Agent is unnecessarily re-interpreting an already explicit date. What I've already tried I rewrote the Retell system message completely (the new 26 November version): Anchors today’s date Forces future-only dates Requires explicit dd/mm/yyyy internal conversion Confirms the exact UK date verbally with the caller before sending anything to n8n Ensures Retell never sends vague phrases like “next Tuesday” in the backend call I also updated the n8n Calendar Agent system message: If chatInput already contains a numeric UK date, it should trust that date 100% It should not convert “26/11/2025” into anything else It should only reject dates that are actually in the past Removed all automatic reinterpretation of relative dates when an explicit one is provided Removed availability checks unless the caller explicitly asks for it

Hey everyone, I’m stuck on something strange and hoping someone has seen this before. I switched my number’s inbound handler to a simple TwiML Bin to dial Retell’s SIP server. The number is definitely set to: Configure with: TwiML Bin A call comes in: Retel SIP dial (my TwiML Bin) Primary handler fails: Retel SIP dial No webhook No Studio Flow No Functions No Proxy No TwiML Apps (confirmed the TwiML Apps page is completely empty) The problem: Twilio still routes every call to a TwiML App URL that doesn’t exist anymore. In the call logs I see Twilio POSTing to this old handler: https://handler.twilio.com/twiml/EH2961c93e597b35174275e1ec153e1f10 But this TwiML App is NOT in my account and isn’t assigned anywhere. Every inbound call hits this ghost URL, then the SIP leg instantly returns “Busy”, and the call ends immediately. I’ve tried reselecting the Bin, saving again, switching to webhook and back, recreating the Bin—nothing changes. Where is Twilio pulling this old TwiML App from, and how do I force it to actually use the TwiML Bin that’s selected? Thanks!

Hey everyone, I’m trying to connect our Twilio BYOC trunk (bci-tech-trunk.sip.twilio.com) with our Retell voice agent so it can answer incoming calls. When we call the Twilio number, it connects but plays: “We’re sorry, an application error has occurred. Goodbye.” Twilio’s logs show Error 11200 (HTTP 404) — “Cannot POST /v2/handle-call” fromhttps://api.retellai.com/v2/handle-call?agent_id=agent_7ce1************03d&retell_api_key=key_2***********049.I also tried switching it to GET, but the same issue happens. We’ve checked all the Twilio settings — BYOC trunk, SIP domain, transport set to TLS, and tested /v2/handle-call, /v2/handle-call/twilio, and /v2/create-phone-call. Secure media has been tested on and off, same result. Could someone confirm what the correct Twilio webhook endpoint should be for this setup, and whether inbound SIP from Twilio should authenticate via IP/TLS trust instead of credentials? Troubleshooting I’ve Tried - Verified Twilio BYOC trunk and SIP domain setup - Tested both HTTP GET and POST - Tried multiple Retell endpoint paths (/v2/handle-call, /v2/handle-call/twilio, /v2/create-phone-call) - Confirmed origination target = sip:sip.retellai.com;transport=tls - Tested with Secure Media enabled and disabled - Retell webhook in place: https://bci-tech-trunk.pstn.twilio.com Still getting the same “application error” message and 404 logs from Retell’s API. Environment details: - Integration type: Twilio BYOC (SIP trunking) - Retell Workspace: BCI Tech - Agent ID: agent_7ce1************03d - Twilio Domain: bci-tech-trunk.sip.twilio.com - Secure Media: tested ON/OFF - Browser: Chrome - OS: Chrome OS Thanks for the help!

Hi community! We are not fixed on any particular kind of help but open to what can be offered. Ultimately our goal is to set up 4 core workflows that we can then sell as part of the ai business solutions agency i am trying to build. Aren't we all! The work flows are a personal assistant, social media content creator/poster, lead scraper and cold email outreach campaign systems. We will eventually look at a cold call/call response workflow too but for now just the first 4. All workflows have been taken from mainly Nates templates initially and built out from there. We would like to find someone we can work with ad hoc on an ongoing basis to just help us where needed and in the first instance help us make these workflows as good as possible and explain certain areas so we understand the technical side a bit more to be able to manage and maintain them. Is this something anyone here could help with and could we perhaps jump on a video call for 10 mins first to just chat through possibilities? This might give you a better understanding of what we are looking for and us an idea of what you can offer/costs etc. I am on UK time for any zoom/teams/google meet video calls. Thank you and I look forward to hearing from you! Best Chris

Hi everyone, I’m working on an Email Agent workflow that integrates with Gmail via n8n. At one point, everything was working as intended: Send Email — Worked perfectly Reply to Email— Was functioning reliably Summarise Emails — Also worked well Delete Email — Never worked correctly (wrong message deleted once, then failed) My goal is to restore full functionality for Send, Reply, and Summarise and only troubleshoot Delete later. Current Setup The workflow uses the following Gmail nodes: Get Many Messages (returns 3 emails, limit set, not simplified) Send Email (fields dynamically populated by AI) Reply to Email (uses ID extracted from Get Many Messages) Delete Email (same, but currently deprioritized) Tool routing and inputs are managed via a separate GPT-powered “main agent” that: Routes the user’s request to this Email Agent Dynamically generates fields like to, subject, body, and id using AI What Changed? Recently, summarizing and replying both started failing with: “Bad request - please check your parameters” or “OpenAI: Rate limit reached” The send function was still working… until even that failed during our latest testing. I suspect the issue is either: Tool names mismatching between system message and n8n node labels Message id not being passed correctly between GetManyMessages and the reply/delete nodes Or AI not generating required fields under some conditions What I’ve Tried: Confirmed correct settings: Return All = Off Limit = 3 Simplify = Off Sender filter = removed completely Ensured Send Email uses: js CopyEdit To: {{ $json.to }} Subject: {{ $json.subject || 'No Subject' }} Message: {{ $json.body || 'No message provided.' }} Ensured Reply and Delete use: js CopyEdit Message ID: {{ $json.id }} All inputs are generated via AI; no hardcoded expressions like $json.response[0].id What I Need Help With Why summarize and reply are failing even when Get Many Messages returns emails correctly How best to pass id from GetManyMessages → ReplyToEmail