The lead engine,
explained end to end.

01System overview

From a brand-new lead to a booked roof inspection — the whole journey runs itself. Here is what happens end to end, then the moving parts underneath.

The big picture

When a new homeowner lead lands in GoHighLevel, the system takes it from there. It texts a satellite photo of their roof, and Sarah — the voice AI — calls within seconds, qualifies them, and books a roof inspection straight into the calendar. Reminders, reschedules, and follow-up all run on their own. Your team only steps in to show up and inspect the roof, and every step is written back to the CRM, so nothing is invisible.

The moving parts

Five layers do the work; n8n sits in the middle — it listens to GoHighLevel, makes the decisions, and orchestrates Retell, Twilio and the CRM.

How a lead enters

Outcome A new contact tagged non-booked-funnel (or non-booked-funnel-lead) in GoHighLevel is the starting gun: GHL's WF-1 workflow fires the First Touch webhook (the next section), and from there n8n runs the whole journey.

Satellite roof image

Outcome A satellite photo of the lead's roof is saved onto their CRM contact — which First Touch then attaches to the intro text it sends the homeowner. (If the photo can't be built, First Touch just sends text only.)

n8n · Stonebriar_Roof_Image — actual build

6 nodes · Webhook -> GET Contact -> Build Map URL -> PUT Image URL -> Tag -> Respond OKLead Journey

Trigger

HTTP POST to stonebriar/roof_image from Stonebriar_First_Touch immediately after guard checks pass. Payload: {contact_id}.

Module flow

1. Webhook—Listens at stonebriar/roof_image; accepts POST with contact_id stonebriar/roof_image
2. GET Contact—Fetches full GHL contact to extract address1, city, state fields for the map URL /contacts/{id} address1 city state
3. Build Map URL—Code node: encodes the property address into a Google Static Maps API URL with satellite maptype, 600x400 size, zoom=20; API key placeholder still present in code comment Google Static Maps API satellite zoom=20
4. PUT Roof Image URL—Writes the constructed satellite URL to GHL custom field WRDAthnXubJbA9GrYQbn on the contact WRDAthnXubJbA9GrYQbn /contacts/{id}
5. Tag roof-image-ready—Adds tag roof-image-ready to confirm the image field is populated and ready for SMS roof-image-ready /contacts/{id}/tags
6. Respond OK—Returns 200 to First_Touch so it can proceed to read the field after the 10-second wait

Speed-to-lead first touch

Outcome Within minutes of a new lead, the homeowner gets a text with a satellite photo of their own roof and Sarah calls them (twice if there's no answer). Whatever happens on the call is then handled by Post-call.

n8n · Stonebriar_First_Touch — actual build

37 nodes · Guard Check -> [skip] Note + stop | [pass] Respond 200 -> Roof Image -> SMS -> Tag -> Wait -> Recheck -> Dial 1 -> Wait 3min -> Recheck -> Dial 2Lead Journey

Trigger

HTTP POST to stonebriar/first-touch from GHL Workflow-1 (fired by the non-booked-funnel tag). Payload: {contact_id}. Fires within seconds of Lead_Intake completing the tag write.

Module flow

1. Webhook + Get Contact—Receives contact_id; immediately fetches full GHL contact record for guard evaluation contact_id
2. Guard Check + Skip? IF—Code node evaluates skip conditions (DND, first-touch-sent, booked, ai-active); IF routes to Respond Skipped with reason note, or Respond Accepted to continue async DND first-touch-sent booked ai-active
3. Window Check (always open)—Code node always returns in_window=true -- send window removed 2026-06-15 per Jay; gate retained for fast re-enabling in_window
4. Trigger Roof Image + Wait 10s—HTTP POST to stonebriar/roof_image; waits 10 seconds for async satellite map build to complete stonebriar/roof_image WRDAthnXubJbA9GrYQbn
5. Extract Roof URL + Verify Image—Reads roof image URL from custom field WRDAthnXubJbA9GrYQbn; HTTP GET verifies image returns 200 with image/* content-type WRDAthnXubJbA9GrYQbn content-type
6. Build Message + Send SMS—Assembles intro SMS with roof photo; sends via GHL conversations API (with image if valid, text-only fallback if URL fails) /conversations/messages roof photo
7. Write Note + Tag first-touch-sent—Writes CRM note confirming SMS send result; adds tag first-touch-sent to prevent duplicate re-send first-touch-sent /contacts/{id}/notes
8. Wait 10 Minutes + Recheck (Dial 1 guard)—Waits 1 minute (node label says 10 min; actual config is 1 min); re-fetches contact to check for post-touch outcome tags before dialing Wait 10 Minutes
9. Voice Decision + Voice Skip? (Dial 1)—Code node checks if contact gained engaged/booked/callback/ai-active tags during the wait; IF skips Dial 1 and writes note if already answered engaged-not-booked callback-scheduled
10. Trigger Voice Bridge Dial 1 + Note—HTTP POST to voice_trigger_call webhook (9f45335d); writes CRM note with Dial 1 outcome (success or error) 9f45335d-da8b-47c4-9465-6e9e6b2c571b Dial 1
11. Wait 3 Min + Recheck For Answer (Dial 2 guard)—Dial 2 gap: waits 3 minutes, re-fetches contact to check if Dial 1 was answered before placing Dial 2 Dial2 gap
12. Trigger Voice Bridge Dial 2 + Note—HTTP POST to same voice_trigger_call webhook for second call attempt (iPhone call-screen bypass); writes Dial 2 outcome note Dial 2 call-screen bypass

Outbound voice call dial

Outcome Sarah places the outbound call as herself, already knowing the homeowner's name, storm-damage details and timezone. When the call ends, Post-call logs it and updates the CRM.

n8n · StoneBriar_voice_trigger_call — actual build

16 nodes · config -> search_contact -> Get Custom Fields/Values -> Map Fields/Campaign -> GetAppointments -> Build Context -> suppress IF -> concurrency loop -> trigger_callLead Journey

Trigger

HTTP POST to webhook path 9f45335d-da8b-47c4-9465-6e9e6b2c571b. Called by: Stonebriar_First_Touch (Dial 1 and Dial 2), GHL Automation directly for non-first-touch leads, and Tag_Bridge_Poller for the callback-requested bridge. Payload: {phone, contact_id, first_name, email, and optional custom data}.

Module flow

1. config—Set node: stores retell_api key, from_phone=+12149721282, GHL API key, and location ID U9eehdYVTaneVBVevw0b key_0e3e94c358e8b7827192512dcf1f +12149721282 U9eehdYVTaneVBVevw0b
2. search_contact—GHL /contacts/search by phone to fetch the full contact record /contacts/search
3. Get_Custom_Fields + Get_Custom_Values—Fetches all GHL custom field definitions for the location and location-level custom values containing campaign metadata /customFields /customValues
4. Map_Custom_Fields + Map_campaign_values—Two code nodes: first maps GHL field IDs to human-readable names for the contact; second extracts campaign values (storm_damage_type, campaign_name) customFields[] storm_damage_type campaign_name
5. Call StoneBriar_GetAppointments_SubWorkflow—Executes sub-workflow to fetch existing confirmed future appointments -- GR-001 dispatch-time booking state recheck before dialing GetAppointments_SubWorkflow GR-001
6. Build Universal Agent Context—Code node: assembles all data into named Retell variables -- roof_age, roof_type, lead_stage, call_number=1, previous_call_summary, prospect_timezone=America/Chicago hardcoded, suppressCall flag retell_llm_dynamic_variables prospect_timezone America/Chicago suppressCall
7. If (suppress check)—IF node: if suppressCall=true (contact already booked or no phone), routes to Note_Call_Suppressed (writes CRM note) and stops suppressCall Note_Call_Suppressed
8. get_concurrency + If_Concurrency_Reached + Concurrency_Wait—GET /get-concurrency from Retell; if current_concurrency >= 40, loops on Concurrency_Wait node until a slot opens /get-concurrency current_concurrency >= 40 Concurrency_Wait
9. trigger_call—HTTP POST to Retell /v2/create-phone-call with agent_c7489b1000369067c5ef69e0dc, from +12149721282, to phone, and full retell_llm_dynamic_variables bundle agent_c7489b1000369067c5ef69e0dc /v2/create-phone-call

Sub-workflows used: StoneBriar_GetAppointments_SubWorkflow

Mid-call calendar tools

Outcome While Sarah is live on the call she can check times, book, reschedule or look up the inspection on the spot — the booking lands on the CRM calendar and the contact is tagged booked, all before the call ends.

n8n · StoneBriar_Voice_Calendar_Bookings — actual build

49 nodes · Webhook -> Set Variables -> Switch -> [5 routes each terminating in respondToWebhook]Lead Journey

Trigger

HTTP POST to f9d1c743-5530-4ac0-9770-63aab57dac06 from the Retell voice agent whenever she calls one of 5 registered function tools mid-call. The body.name field identifies which tool. responseMode=responseNode means a respondToWebhook node must fire within Retell's tool timeout.

Module flow

1. Webhook -> Set Variables—Webhook receives Retell tool call; Set Variables loads calendarId=HwJD5Ujgck1FJERzBZVw, time_zone, API key, user_id HwJD5Ujgck1FJERzBZVw f9d1c743-5530-4ac0-9770-63aab57dac06
2. Switch -> 5 routes on body.name—Switch routes on: check_calendar_availability (out0), book_appointment (out1), check_calendar_slots (out2), update_appointment (out3), get_appointment (out4) check_calendar_availability book_appointment check_calendar_slots update_appointment get_appointment
3. Route 0: check_calendar_availability—Code3 -> date math (startTime/addDay/endDate) -> GET /calendars/free-slots (timezone from Set Variables) -> Code formats voice-ready slot list -> respondToWebhook /calendars/free-slots time_zone
4. Route 2: check_calendar_slots—Code2 -> date math -> GET /calendars/free-slots (America/Chicago hardcoded) -> Code1 formats grouped time ranges -> respondToWebhook /calendars/free-slots America/Chicago
5. Route 1: book_appointment—setContactIdFound -> date format + Merge -> calls appointment_booking_sub_workflow (mjLv4YHPv7CtDCES) -> on success: writes event_id to contact, removes stale tags, adds booked or ai-booked based on lead_stage -> respondToWebhook success or failure message mjLv4YHPv7CtDCES event_id booked ai-booked
6. Route 3: update_appointment—setContactIdFound1 -> date format + Merge1 -> calls Update_Appointment_SubWorkFlow (D0E0AxALdJmYcfLx) -> on success: writes new event_id, updates tags -> respondToWebhook D0E0AxALdJmYcfLx event_id
7. Route 4: get_appointment—HTTP GET /contacts/{id}/appointments -> Split Out -> Filter (confirmed + future, America/Chicago) -> Sort desc -> Limit 1 -> format fields -> respondToWebhook with summary or no-appointment message /contacts/{id}/appointments America/Chicago confirmed

Sub-workflows used: stoneBriar_appointment_booking_sub_workflow StonBriar_Update_Appointment_SubWorkFlow

Post-call processing

Outcome After every call the homeowner is routed correctly: opt-outs are set to Do-Not-Call and stopped; everyone else gets a written call summary, the right pipeline stage, and a callback scheduled if they asked. This is what keeps the CRM truthful after each call.

n8n · StoneBriar_voice_post_call — actual build

28 nodes · dnc_check -> [DNC: Set DND + tags, stop] | [pass: config -> Switch -> [call_started: conversation state] | [call_analyzed: 3-way parallel -> outcome routing -> pipeline move]]Lead Journey

Trigger

HTTP POST to webhook path 12d1db5a-4bc9-4c17-88b3-9ca5f5b20872 from Retell's post-call webhook. Fires for every call event. body.event is either call_started (call connected) or call_analyzed (call ended and analyzed). First node is always the DNC gate.

Module flow

1. dnc_check—IF node: checks custom_analysis_data.opt_out_requested -- if true, sets contact DND=true and adds tags DNC + dnc-honored, then stops all further processing opt_out_requested DND DNC dnc-honored
2. config + Switch—Set node loads GHL and Retell creds; Switch routes on body.event: call_started to out0, call_analyzed to out1 call_started call_analyzed
3. Call_Started -> Change_conversation_State—On call_started event: IF verifies call_status=ongoing, then PUTs GHL conversation to active state for real-time CRM visibility during the call /conversations/{id} active
4. HTTP Request + If1 (appointments fetch)—Fetches contact appointments to inform outcome routing; If1 gates entry into the call_analyzed branch /contacts/{id}/appointments
5. Call_Back_Requested IF -> schedule_call_back—Branches on lead_qualification_status=Qualified - Callback Scheduled; TRUE path parses callback time and POSTs Retell create-batch-call with trigger_timestamp Qualified - Callback Scheduled /create-batch-call trigger_timestamp
6. Generate Memory Summary1/2 + Update_contact—OpenAI generates a call memory summary; writes it back to GHL contact along with call disposition and lead_stage fields OpenAI ai_interaction_summary previous_call_summary
7. Pick_Stage code node—Maps call outcome to GHL pipeline stage IDs: booked->Inspection Booked, callback/voicemail/hang_up/no_answer->Contact Attempted, wrong_number->No-Show/Cancel, opted_out->Dead-Max Attempts booked callback_scheduled no_answer wrong_number opted_out fMFWsnKweG7UwwUTnDJ2
8. Search_Opportunity -> update_pipeline_stage—Searches GHL for the contact's open opportunity then PUTs the new stage ID to move the pipeline card /opportunities/search /opportunities/{id}
9. Call save_voice_logs_ghl—Executes shared sub-workflow (wpxsVyOfCMMedkDs) to create a call_logs custom object record in GHL linked to the contact wpxsVyOfCMMedkDs call_logs

Sub-workflows used: save_voice_logs_ghl

GHL tag bridge

Outcome The GoHighLevel “Jenny” SMS bot can trigger any booking action just by tagging a contact — every few minutes this turns that tag into the matching workflow (book, reschedule, cancel, schedule-check, or call-me-back).

n8n · Stonebriar_Tag_Bridge_Poller — actual build

2 nodes (trigger + single code node handling all 5 bridges) · Every 3 min -> Tag Bridge PollerSMS Booking & Ops

Trigger

n8n Schedule trigger: every 3 minutes. No external webhook -- pure poller. GHL Conversation AI (Jenny) sets tags on contacts based on homeowner intent; this workflow is the bridge that converts those tags into n8n engine calls.

Module flow

1. Every 3 min—Schedule trigger fires every 3 minutes around the clock minutesInterval: 3
2. Bridge: cancel-requested—GHL search by tag cancel-requested -> strip tag first (GR-008 dedupe) -> POST stonebriar/cancel-appt {contact_id} -> if webhook fails, re-add tag (self-heal) cancel-requested stonebriar/cancel-appt
3. Bridge: schedule-check—GHL search by tag schedule-check -> strip tag -> POST stonebriar/schedule-check {contact_id} -> self-heal on failure schedule-check stonebriar/schedule-check
4. Bridge: booking-requested—GHL search by tag booking-requested -> strip tag -> POST stonebriar/sms-book {contact_id, action:book} -> self-heal on failure booking-requested stonebriar/sms-book action:book
5. Bridge: reschedule-requested—GHL search by tag reschedule-requested -> strip tag -> POST stonebriar/sms-book {contact_id, action:reschedule} -> self-heal on failure reschedule-requested action:reschedule
6. Bridge: callback-requested—GHL search by tag callback-requested -> strip tag -> POST voice_trigger_call webhook 9f45335d {contact_id, phone} -> self-heal on failure (P21 pattern, added v2.0 2026-06-15) callback-requested 9f45335d-da8b-47c4-9465-6e9e6b2c571b

SMS booking engine

Outcome The appointment is actually booked or changed on the CRM calendar and a truthful confirmation is written back to the contact — which Jenny reads out to the homeowner, so the text they get always matches what really happened.

n8n · Stonebriar_SMS_Booking_Engine — actual build

3 nodes · Webhook -> SMS_Booking_Engine code -> Respond_OK (booking_reply and available_slots written inside the code node)SMS Booking & Ops

Trigger

HTTP POST to stonebriar/sms-book from Tag_Bridge_Poller with {contact_id, action}. Action values: book, reschedule, cancel, check, prefetch. Cloned from Beem n8n_beem_book.js v6 (2026-06-16), adapted for GHL-native calendar.

Module flow

1. Webhook—Listens at stonebriar/sms-book; accepts {contact_id, action} stonebriar/sms-book contact_id action
2. SMS_Booking_Engine code node—Single code node implementing all 5 actions: book (POST /calendars/events/appointments), reschedule (PUT), cancel (DELETE), check (GET appointments + format), prefetch (GET free-slots + format for bot prompt). Timezone hardcoded America/Chicago throughout. America/Chicago /calendars/events/appointments /calendars/free-slots book reschedule cancel check prefetch
3. Write booking_reply field—Writes the natural-language outcome (e.g. 'Your inspection is booked for Tuesday June 24 at 10 AM') to GHL LARGE_TEXT field mj2pMnYp9cOsiAo9y0qT -- Jenny merges this via {{contact.booking_reply}} mj2pMnYp9cOsiAo9y0qT booking_reply
4. Write available_slots field—On prefetch/check actions: writes formatted slot list to field 0gK2qWEIrK94ysqGDzEX so Jenny's prompt includes real availability 0gK2qWEIrK94ysqGDzEX available_slots
5. Write appointment_time_display—On successful book/reschedule: writes human-readable appointment time to TEXT field BlD6F0J5yvDzcgUV7uWN for CRM display BlD6F0J5yvDzcgUV7uWN appointment_time_display
6. Respond_OK—Final respondToWebhook confirming engine completed

Appointment lookup

Outcome The homeowner is told their real upcoming inspection — by text, or read out by Sarah on a call — straight from the calendar, with no guessing.

n8n · Stonebriar_Schedule_Check — actual build

17 nodes · Webhook -> Config -> Get Appointments -> Filter -> [found: Write Display + Write booking_reply + CRM Note -> Channel Gate -> SMS or skip] | [not found: Write booking_reply + CRM Note -> Channel Gate] -> Respond OKSMS Booking & Ops

Trigger

HTTP POST to stonebriar/schedule-check from Tag_Bridge_Poller (tag=schedule-check) or directly from a Retell voice agent function tool call (with channel=voice in payload). Payload: {contact_id, channel?}.

Module flow

1. Webhook + Config—Webhook at stonebriar/schedule-check; Config set node loads GHL API creds and calendar ID HwJD5Ujgck1FJERzBZVw stonebriar/schedule-check HwJD5Ujgck1FJERzBZVw
2. Get_Contact_Appointments—GET /contacts/{id}/appointments to retrieve all appointments for this contact /contacts/{id}/appointments
3. Filter_And_Format—Code node: filters for confirmed future appointments on calendar HwJD5Ujgck1FJERzBZVw, sorts by start time, formats in Central Time (America/Chicago) confirmed America/Chicago
4. Appointment_Found? branches—IF: appointment found -> Write_Appt_Display_Field then Write_Booking_Reply_Found then CRM_Note_Found; not found -> Write_Booking_Reply_None then CRM_Note_None Write_Booking_Reply_Found Write_Booking_Reply_None
5. Channel_Is_Voice_Found / None gates—IF on each path: if channel=voice, goes directly to Respond_OK (skips SMS); otherwise sends SMS confirmation via GHL conversations API channel=voice /conversations/messages
6. CRM_Note_SMS_Failed—continueErrorOutput on the SMS node: if SMS send fails, writes a CRM note and still responds OK -- booking_reply field is already written CRM_Note_SMS_Failed
7. Respond_OK—respondToWebhook terminates the workflow; Retell reads booking_reply from GHL field if calling as a function tool

Appointment cancel

Outcome The inspection is cancelled on the calendar, the booked tag is removed, and the homeowner gets a confirmation — keeping the calendar and the CRM in sync.

n8n · Stonebriar_Appt_Cancel — actual build

22 nodes · Webhook -> Config -> Get Appointments -> Find Upcoming -> [not found: Note + exit] | [found: Cancel -> [fail: Note + exit] | [success: Remove Tag + Clear Field + Write booking_reply + Tag -> Channel Gate -> SMS or skip]] -> Respond OKSMS Booking & Ops

Trigger

HTTP POST to stonebriar/cancel-appt from Tag_Bridge_Poller (tag=cancel-requested) or from a Retell voice agent function tool call (with channel=voice). Payload: {contact_id, channel?}.

Module flow

1. Webhook + Config—Webhook at stonebriar/cancel-appt; Config loads GHL creds and calendar ID HwJD5Ujgck1FJERzBZVw stonebriar/cancel-appt HwJD5Ujgck1FJERzBZVw
2. Get_Contact_Appointments + Find_Upcoming_Appt—GET appointments; code node finds the soonest confirmed future event and extracts its event_id for the DELETE call /contacts/{id}/appointments event_id
3. Appointment_Found? -> Cancel_Event—IF: if appointment found, DELETE /calendars/events/{eventId}; if not found, branches to CRM_Note_Not_Found and exits /calendars/events/{eventId} DELETE
4. Cancel_Succeeded? -> CRM_Note_Cancelled—IF: on successful delete, writes CRM note; on failure writes CRM_Note_Cancel_Failed and exits CRM_Note_Cancelled CRM_Note_Cancel_Failed
5. Remove_Booked_Tag + Clear_Appt_Time_Display—Removes booked tag from contact; clears appointment_time_display field BlD6F0J5yvDzcgUV7uWN booked BlD6F0J5yvDzcgUV7uWN /contacts/{id}/tags
6. Write_Booking_Reply_Cancelled—Writes cancellation confirmation text to booking_reply field mj2pMnYp9cOsiAo9y0qT for Jenny to merge mj2pMnYp9cOsiAo9y0qT booking_reply
7. Add_Cancelled_Tag_Fallback + Channel_Is_Voice_Cancelled—Adds a cancelled tag for funnel tracking; Channel IF: voice channel goes directly to Respond_OK; SMS channel sends confirmation via GHL conversations API channel=voice /conversations/messages
8. CRM Note branches (all paths)—Every exit path (cancelled, failed, not found, SMS failed) writes a CRM note -- GR-008 visible artifact on every branch CRM_Note_SMS_Failed CRM_Note_Not_Found

Inbound caller lookup

Outcome When a homeowner calls the main line, Sarah instantly knows who they are and their next inspection, so she greets them by name and helps right away instead of “let me look you up.”

n8n · Stonebriar_inbound_lookup — actual build

11 nodes · Webhook -> Config -> Normalize -> Has Phone? -> [no: No_Match] | [yes: GHL Search -> Pick Contact -> Contact Found? -> [no: No_Match] | [yes: Get Appointments -> Format Found]] -> RespondSMS Booking & Ops

Trigger

HTTP POST to stonebriar/inbound-lookup called by the Retell inbound agent as a lookup_caller function tool when a homeowner calls the Stonebriar main line. Payload contains either body.args.phone or body.call.from_number (two shapes are handled after a bug fix).

Module flow

1. Webhook + Config—Webhook at stonebriar/inbound-lookup; Config loads GHL API key and location ID stonebriar/inbound-lookup
2. Normalize_Phone—Code node: handles both body.args.phone and body.call.from_number input shapes; normalises to E.164 (+1XXXXXXXXXX) body.args.phone body.call.from_number E.164
3. Has_Phone? -> GET_Contacts_By_Phone—IF: if normalised phone is valid, search GHL /contacts/ by phone; if not, go directly to No_Match response /contacts/ phone
4. Pick_Contact—Code node: selects the best GHL contact match from search results
5. Contact_Found? -> Get_Contact_Appointments—IF: if contact found, GET /contacts/{id}/appointments; if not found, go to No_Match with empty response /contacts/{id}/appointments
6. Format_Found—Code node: builds structured response with contact_id, first_name, email, has_appointment (bool), appointment_summary (plain text), event_id contact_id first_name has_appointment appointment_summary event_id
7. No_Match + Respond—No_Match code returns empty-contact response; single Respond respondToWebhook node handles both found and not-found paths

The email AI assistant

Outcome Inbound emails get an on-brand, context-aware reply in seconds. The right specialist agent handles each one -- book, answer, handle an objection, or escalate -- and every exchange is logged back to the CRM.

n8n · StoneBriar_Email_Bot-Multi-Agent — actual build

30 nodes -> reads context -> multi-agent reply -> Send_Reply -> Log ConversationEmail Assistant

Trigger

Fires on an inbound email. It loads the contact's full GoHighLevel record, custom fields, campaign values, appointments, and the prior email thread before the AI replies.

Module flow

1. config + Search_contact—Loads settings and pulls the emailer's full GoHighLevel record.
2. Custom fields + campaign values—Maps the contact's fields and campaign data into context. GHL
3. Get appointments + conversation thread—Pulls existing appointments and the prior email thread so the AI has the full picture.
4. Build Universal Agent Context—Assembles everything into the variables the AI reads.
5. Ultimate Assistant (multi-agent)—A LangChain agent with memory plus 5 specialist tools: booking, simple-response, information, objection-handling, escalation. OpenAI memory
6. Send_Reply + Log Conversation—Sends the email reply and logs the exchange back to GoHighLevel. visible artifact

Rep SMS reply poller

Outcome A field rep just texts 1, 2 or 3 after an inspection and it's picked up automatically and handed to the outcome catcher — no app, no form to fill in.

n8n · Stonebriar_Rep_Reply_Poller — actual build

2 nodes · Every 1 min -> Poll rep replies (single code node, both staff numbers)Supporting

Trigger

n8n Schedule trigger: every 1 minute. Polls GHL conversations for two hardcoded staff phone numbers: Darian (+14693436120) and a second rep (+18325852875). Uses n8n staticData to track the last-seen message ID per rep so only genuinely new messages are forwarded.

Module flow

1. Every 1 min—Schedule trigger fires every 60 seconds minutesInterval: 1
2. Poll rep replies -> catcher—Code node: queries GHL conversations for Darian (+14693436120) and a second rep (+18325852875); uses staticData to track last-seen message ID per number; for each new inbound message, POSTs {from_phone, body, contact_id} to stonebriar/rep-outcome; bootstrap guard on first run seeds seen state without replaying history +14693436120 +18325852875 stonebriar/rep-outcome staticData pit-b4407fda-bbf6-44c4-9007-bee80f590d00

Rep outcome recorder

Outcome The result the rep texted is written onto the right customer's record and the rep gets a confirmation — so the pipeline reflects what actually happened in the field.

n8n · Stonebriar_Rep_Outcome_Catcher — actual build

3 nodes · Webhook -> Catcher code node -> Respond 200Supporting

Trigger

HTTP POST to stonebriar/rep-outcome from Stonebriar_Rep_Reply_Poller. Payload: {from_phone, body, contact_id}. The body field contains the rep's reply digit (1=completed, 2=no-show, 3=reschedule).

Module flow

1. Webhook—Listens at stonebriar/rep-outcome; receives {from_phone, body, contact_id} from Rep_Reply_Poller stonebriar/rep-outcome from_phone body
2. Catcher code node—Validates from_phone is a known staff number; maps body digit (1/2/3) to appointment outcome text; searches GHL for contacts tagged outcome-pending; sets appointment outcome field (27quKIEu2BIBFCyszE0E) on the matching customer; strips outcome-pending tag; texts the rep a confirmation; handles disambiguation if multiple pending customers exist 27quKIEu2BIBFCyszE0E outcome-pending appointment outcome +14693436120 +18325852875
3. Respond 200—respondToWebhook terminates the workflow after outcome is recorded

Appointment safety net

Outcome No inspection slips through the cracks — if a rep forgets to report, this auto-marks a no-show after a few hours and flags overdue ones, keeping the pipeline honest on its own, every hour.

n8n · Stonebriar_Appt_Safety_Net — actual build

2 nodes · Schedule Trigger -> Outcome Poller (all logic in one code node)Supporting

Trigger

n8n Schedule trigger: every 1 hour. Scans GHL calendar HwJD5Ujgck1FJERzBZVw for all appointments that ended within the last 26 hours. Uses n8n staticData to act at most once per appointment ID.

Module flow

1. Schedule Trigger—Fires every 60 minutes to scan for unresolved inspection outcomes hoursInterval: 1
2. Stonebriar Outcome Poller code node—Queries GHL calendar events for appointments ended in last 26h; for each: checks outcome field (27quKIEu2BIBFCyszE0E) and outcome-set tag -- skips if already handled; if ended >3h and no outcome: writes CRM note FIRST (GR-008), moves pipeline to No-Show stage (bc717599), adds no-show + outcome-no-show tags, removes booked; if ended >4h and not flagged: writes CRM note, creates GHL task for rep, adds outcome-overdue tag; within 3h grace window: skip. staticData deduplicates across runs. 27quKIEu2BIBFCyszE0E outcome-set bc717599-e7e6-4133-bbb7-c8306495e1ca no-show outcome-no-show outcome-overdue HwJD5Ujgck1FJERzBZVw staticData

Bulk reactivation push

Outcome A batch of dormant leads is dropped back into a re-engagement cadence in one go — turning the 2,000–10,000-lead database back into booked inspections.

n8n · Stonebriar_Reactivation_Bulk_Push — actual build

5 nodes · Webhook -> Respond Accepted (async split) -> Config + Parse -> Process Leads -> Batch SummarySupporting

Trigger

HTTP POST to stonebriar/reactivation-bulk-push with {leads:[{contact_id,...}], mode, dry_run?, entry_tag?}. Mode=reactivation tags with 60-day-sunset (GHL WF-6). Mode=fresh tags with non-booked-funnel (GHL WF-1). Supports dry_run=true for testing without writing tags.

Module flow

1. Webhook—Listens at stonebriar/reactivation-bulk-push; receives the leads batch with mode and optional dry_run stonebriar/reactivation-bulk-push leads[] mode dry_run
2. Respond Accepted—Sends immediate 200 to the caller before the batch begins (prevents HTTP timeout on large batches)
3. Config + Parse—Code node: validates body.leads array, determines tag to apply (60-day-sunset for mode=reactivation, non-booked-funnel for mode=fresh), sets throttle delay between contacts 60-day-sunset non-booked-funnel mode
4. Process Leads—Code node: iterates the leads array sequentially with throttle; for each valid contact_id, applies the tag to GHL /contacts/{id}/tags; dry_run skips all GHL writes; tracks processed/skipped/error counts /contacts/{id}/tags dry_run throttle
5. Batch Summary—Code node: assembles and logs batch results (total, processed, skipped, errors) for operator review processed skipped errors

Referral lead intake

Outcome A rep's quick “my neighbor has hail damage” text becomes a real new lead in the CRM (and the referrer gets credited) — so word-of-mouth enters the same machine as every other lead.

n8n · Stonebriar_Referral_Intake — actual build

21 nodes · Webhook -> Respond Accepted -> Get Referrer -> Parse Referral -> [no phone: Tag Pending + Note] | [phone: Upsert -> New/Existing -> Read Back -> Update Referrer -> Note + Tag Captured]Supporting

Trigger

HTTP POST to stonebriar/referral_intake from a GHL Workflow triggered when a rep tags a contact with stonebriar-referral. Payload: {referrer_contact_id, referral_text}. The referral_text is the rep's raw SMS message containing a name and/or phone number.

Module flow

1. Webhook + Respond Accepted—Webhook at stonebriar/referral_intake; Respond Accepted fires immediately for async processing stonebriar/referral_intake referrer_contact_id referral_text
2. Get Referrer—GET /contacts/{referrer_contact_id} to fetch the referring rep's full record /contacts/{id}
3. Parse Referral—Code node: extracts a US phone number (normalised to E.164) and name from free text using regex; three outcomes: phone+name found, name only, unparseable E.164 regex
4. Has Phone? -> Upsert Referral—IF: if phone found, POST /contacts/upsert by phone to create or find existing contact; if no phone, branches to manual-follow-up path /contacts/upsert
5. New Contact? -> Set New Fields + Tag New—IF: if new contact, sets fields (name, source=referral) and tags referred-lead + non-booked-funnel; if existing, tags referred-existing referred-lead non-booked-funnel referred-existing
6. Read Back Referral + Verify Write—Re-fetches the upserted contact and verifies the write succeeded (read-back confirmation pattern)
7. Update Referrer Fields + Remove Pending Tag—Increments referral count field on referrer; removes referral-intake-pending tag; adds referral-captured tag referral-intake-pending referral-captured
8. Write Receipt Note—Writes a CRM note on the referrer contact confirming the referral was captured (name, phone, new contact ID) /contacts/{id}/notes
9. Manual path (no phone found)—Tags referrer referral-intake-pending; writes a note asking rep to provide a phone number; Manual Receipt code builds the note body referral-intake-pending Note Manual

Error alert

Outcome If any workflow ever fails, Jay gets an email instantly with the details — so problems are caught and fixed fast instead of silently breaking the system.

n8n · Stonebriar_error_alert — actual build

2 nodes · Error Trigger -> Send Gmail (jjcavada1@gmail.com)Supporting

Trigger

n8n errorTrigger -- fires automatically when any workflow that has errorWorkflow set to WMmYytB6NH41OMHy encounters an unhandled error. No webhook or schedule. This is the global error handler for all Stonebriar workflows.

Module flow

1. Error Trigger—errorTrigger node: receives error context from any failing workflow -- workflow name, node name, error message, execution ID errorTrigger workflow.name node.name error.message executionId
2. Send a message (Gmail)—Sends email to jjcavada1@gmail.com with workflow name, failed node, error text, and a direct link to the n8n execution log jjcavada1@gmail.com Gmail execution URL

GoHighLevel · the CRM layer

GoHighLevel is the system of record: it holds every contact, runs the SMS bot, drives the pipeline, and fires the n8n workflows by adding tags. n8n does the heavy lifting; GHL is where the lead lives and where every action is written back so a human can always see it.

The tags that drive everything

Adding a tag in GoHighLevel is what fires an automation. This is the wiring between GHL and n8n.

GHL-side workflows & pipeline

Retell · the voice agent

Retell is the voice. A GPT-4.1 outbound agent calls new storm-damage leads and an inbound line answers the main number. Mid-call, the agent reaches into n8n through function tools to read and write the calendar in real time.

Functions — the tools the agent calls mid-call

Each is a live webhook into the n8n booking + CRM workflows. The agent uses them while talking.

Dynamic variables — context passed in at call time

Injected from GoHighLevel so the agent already knows the lead before it speaks:

Prompt structure

The agent prompt is engineered, not improvised. Its backbone:

Full workflow inventory

All 35 workflows with live status. firing = running now · sub = building block called by others · standby = built, off, ready · retired = unpublished 2026-06-18. The 19 firing entry points are detailed in full above; the sub-workflows below are the reusable building blocks they call.

Sub-workflows (in use, called by the flows above)

What triggers what

The master map: every source event, the workflow it hits, and what happens next.

Key IDs & config

Identifiers the workflows rely on. Secrets (keys/tokens) are redacted to their shape — real values live in the credential vault.

Retired / not in use

Verified against live execution history (2026-06-18). These are not firing — listed so the team knows exactly what's dormant and why. All reversible.

Retired — unpublished 2026-06-18 retired

The old n8n SMS chatbot subtree, superseded by the GoHighLevel "Jenny" bot + the tag bridge + the booking engines. Zero executions on record. Deactivated on 2026-06-18; can be switched back on if ever needed.

Standby / off (already inactive)

Open items · for the dev

Surfaced by the live audit, ranked by priority. None block the system running — these are the "before you fully trust it" checks.

Who to contact