Description
Summary I'm building an AI voice agent for a real estate business that answers inbound calls, qualifies leads, and books appointments directly into GoHighLevel — fully autonomous, no human needed to close the booking. I started this build with AI assistance (Manus + ChatGPT/Claude) but have hit a wall debugging cross-system issues without a coding background. I need an experienced developer to take over, diagnose what's broken, and finish the build to a genuinely working state. I'm not a developer. I need someone who can work independently, make good judgment calls, and communicate clearly about what's wrong and what they're doing to fix it — not just deliver a black box. ## Current State (Be Honest With Me About What You Find) - **Vapi.ai assistant** is partially configured with a system prompt but is still needed 4 tools(`ghl_get_contact`, `ghl_create_contact`, `ghl_check_availability`, `ghl_create_event`) intended to look up/create contacts and book appointments live during the call. - **Known symptom:** during a live test call, when the agent reaches the calendar availability check, the call hangs (goes silent/on hold) until it times out. This suggests the `ghl_check_availability` tool call is firing but never returning a valid response — could be a malformed request, wrong endpoint, auth/scope issue, or a date/timestamp formatting problem. - **GoHighLevel** has a Private Integration Token configured (scopes may not be fully correct — needs verification) and a calendar set up. - **Make.com** has a scenario that receives Vapi's end-of-call-report webhook and creates a contact in GHL + sends an SMS notification. This part has mostly worked, with two known bugs already identified and partially fixed: a `locationId` field that was incorrectly set to a full URL instead of the bare ID, and a custom field key that was referencing a broken variable instead of a literal string. - I do NOT have current confidence that any of the four Vapi tools work correctly in isolation — they have not been properly tested outside of live phone calls. - A separate Twilio A2P 10DLC SMS registration was previously rejected and is being resubmitted with a new number; this is being handled separately and is lower priority than getting the voice agent working. ## Scope of Work 1. **Audit everything that currently exists** — Vapi assistant config, all four tools, GHL token/scopes, Make.com scenario — and tell me honestly what's actually wired correctly vs. what needs to be rebuilt. 2. **Fix the calendar availability hang** as the first priority — diagnose why the call goes silent at that step and get it returning real available time slots. 3. **Get full live booking working end-to-end**: caller is qualified → contact is created/found in GHL → availability is checked → appointment is booked directly onto the GHL calendar → agent confirms verbally only after a successful booking. 4. **Verify the Make.com post-call layer works cleanly** alongside the live booking (lead notification SMS, any post-call data logging) without throwing errors. 5. **Test thoroughly** — both via Vapi's built-in tool tester (not just live calls) and with multiple real test calls covering different scenarios (caller books successfully, caller doesn't want to book, caller asks an off-script question). 6. **Document what was wrong and what you changed**, in plain language I can actually follow, so I'm not fully dependent on you for every future tweak. ## Nice to Have / Possible Follow-On Work - Help getting the Twilio A2P 10DLC SMS registration approved (privacy policy/consent flow review) - Building a second assistant for a legal intake use case (same architecture, different vertical) - Ongoing maintenance/support relationship after initial launch ## What I'm Looking For - Real, demonstrable experience with Vapi.ai (or similar voice AI platforms like Retell/Bland) AND GoHighLevel's API — this specific combination, not just general automation experience - Comfortable diagnosing existing broken systems, not just building from a blank slate - Clear, plain-language communicator — I will not understand technical jargon without explanation, and I need to actually follow what's happening - Honest about timelines and what's realistic — I've already lost about three weeks to a build process that didn't surface problems clearly ## Project Type & Budget Open to fixed-price for a clearly scoped first phase (get booking working end-to-end) followed by hourly or a small retainer for ongoing work. Please propose what scope/pricing structure makes sense once you've asked any clarifying questions. ## Timeline This has already taken much longer than expected. I'd like to see real progress within the first few days of engagement — even if the full fix takes longer, I want to see that someone competent is actually making headway. --- **To apply, please answer:** 1. Have you worked with this specific Vapi + GoHighLevel combination before? Please describe a similar project. 2. Based on the "calendar hang" symptom described above, what's your first instinct on what's likely wrong? 3. How do you typically communicate progress/blockers with non-technical clients? 4. What would you want from me (account access, existing configs, etc.) to get started?