PitchPilot — a field-sales companion with a RAG copilot (case study)

A RAG copilot that turns a 40-page sales playbook into a grounded one-paragraph answer, with a button to the doc it came from — at the doorstep, in real time.

Source is private. This is the architecture and the thinking behind it — no product code. The retrieval technique is shown, runnable, in a separate repo: rag-demo.

PitchPilot is a white-label app for door-to-door / field sales reps: capture a lead, run the pitch, handle objections, and track commission — all on a phone, at the door. Its core is Wingman, a retrieval-augmented copilot that answers a rep's question ("what do I say when they mention price?") from the product's own knowledge base, in real time. The answer is grounded — the system prompt hard-limits the model to the retrieved chunks — and it can hand the rep a button straight to the source doc. The prose itself is deliberately clean of inline citations: a rep reading off a doorstep does not want [chunk 3] in the middle of a sentence.

Where it stands: this is a live, working demo — branded "VoltLine" for a fictional energy-sector vertical — not yet deployed to a real sales team. Everything below (the architecture, the debugging, the fixes) is real and running in that demo today; the next milestone is a first paying field-sales client.

The problem

A field rep can't read a 40-page playbook on a doorstep. When a prospect pushes back, the rep has seconds to answer, accurately, in the product's own words. Wingman turns the whole playbook — pricing, objections, scripts — into something you ask in natural language and get a one-paragraph, on-message answer from.

Architecture

flowchart LR
    subgraph Ingest [once, offline]
        D[product docs<br/>pricing · objections · scripts] -->|chunk + embed| KB[(pgvector KB)]
    end

    subgraph Ask [at pitch time]
        R[rep question] --> EF[copilot-ask<br/>edge function]
        EF -->|embed + match| KB
        KB -->|top-k chunks| EF
        EF -->|context-limited prompt| G[Gemini]
        G -->|grounded answer + actions| R
    end

Loading

• Knowledge base — the playbook is chunked, embedded, and stored in Supabase pgvector, partitioned by product so a white-label tenant only ever retrieves its own content.
• Retrieval RPC — a Postgres match function returns the nearest chunks; this is the exact shape shown in rag-demo.
• Grounded generation — a Supabase Edge Function builds a context-limited prompt and calls Gemini. The system prompt hard-limits the model to the retrieved chunks and forbids inventing pricing or promises — accuracy over hype, because a wrong answer at a doorstep costs a sale.
• White-label — one codebase rebrands for any industry in under a day; a public demo runs as "VoltLine", a fictional energy-sector tenant, not a client.

What made it actually work (the debugging that mattered)

• Retrieval was silently returning nothing. An IVFFlat index over a tiny corpus collapsed recall to zero — the copilot kept deflecting to "check the FAQ." Dropping to exact search fixed it instantly. Lesson: approximate indexes need enough rows to be approximate over.
• Truncated mid-sentence answers. The model is a "thinking" variant; reasoning tokens were eating the output budget. Setting the thinking budget to zero restored full answers.
• Cost under load. The free tier dried up during a 50-question QA run, so generation falls back through a cost-ordered chain of models to stay within quota.

What a competitive check found and fixed

A 2026-07-04 pass benchmarked PitchPilot's positioning against established sales-enablement categories — Gong/Chorus (conversation intelligence), Klue/Crayon (competitive intel), Highspot/Seismic (enablement platforms). The "rebrand any industry in under a day" claim held up for speed: new content, new brand, no code change. It did not hold up for concurrency. The knowledge base used a hardcoded two-value product set (luckycat and a single template slot), and the indexer deleted-and-reinserted that one template slot on every rebrand — so reindexing a second industry vertical silently wiped out whatever demo was already using it. Two white-label demos could never run at the same time; the claim was only ever true one vertical at a time.

Fix: replaced the hardcoded slot with a real multi-tenant KB registry (migration 0096_pitcher_kb_tenants) — each vertical now gets its own registered slug, its own isolated chunks, and a foreign-key constraint that stops any unregistered slug from writing content at all. Multiple industry demos can now coexist without one erasing another.

Isolating the demo (a security note)

The public demo and the real product shared one API key, so demo traffic burned the production quota. I added product-aware key selection in the edge function — the demo tenant uses its own key — so a curious visitor can never exhaust or bill the real one.

Screenshots

Screenshots are from the public white-label demo (fictional "VoltLine" energy tenant), not a client deployment.

Wingman — answering a live price objection	Pitch kit — tools, demos & key numbers

Stack

Supabase (pgvector · Edge Functions) · Gemini · RAG · TypeScript / React · Cloudflare Pages

Live demo & the technique

• Live demo: https://lucky-cat.pages.dev/pitcher-template/
• The retrieval pipeline, runnable end-to-end: github.com/augbastos/rag-demo

More work: augustobastos.pages.dev