ChatFlowComms Automation · In production

ChatFlow — AI email & CRM automation

Drafts, refines, and learns from approved replies — IMAP in, SMTP out.

RoleArchitect & BuilderDraft engine, sync, training loop
LoopAuto-trainlearns from approved replies
StackLLM · IMAP/SMTPNeon · Drizzle · SSE

The context

Customer email is high-volume and repetitive, but every reply still needs judgement and brand voice.

ChatFlow handles the loop: IMAP sync ingests inbound, an LLM drafts and refines replies over streaming SSE, the operator approves or edits, and the system auto-trains on approved replies to improve. It matches Shopify orders, tracks deal stages, and rate-limits the AI endpoints. The human stays in the loop; the drudgery does not.

What I built
Draft GenerationReply RefinementAuto-training LoopIMAP/SMTPShopify MatchingDeal StagesRate Limiting
Stack
LLMIMAPSMTP / NodemailerNeonDrizzleSSEReact

The loop

human-in-the-loop
   inbound email ──IMAP──▶ ┌──────────────────┐
                           │   ChatFlow        │
                           │  draft ─▶ refine  │──SSE──▶ operator
                           └──────────────────┘            │
                                  ▲                approve / edit
                                  │                        │
                            auto-train ◀───────────────┘
                                  │
                            SMTP send ─▶ customer
Stream
SSE draft + refine — replies write themselves, you approve.
Learns
Auto-trains on approved replies to sharpen voice.
CRM
Matches Shopify orders, tracks deal stages inline.