Prolo — Ingestion & RFQ Matching

Design deliverables for the quote-ingestion pipeline (Gmail → extracted quote in RDS) and the RFQ matching + submission tail. Each card renders a diagram or spec; yellow = AI-bearing throughout.

Ingestion — Stages 1–4

Gmail poll → fetch → classify → extract & persist to RDS.

F2 · Data flow

Data Flow

The logical journey of a quote (stages 1–4), plus the decomposed Stage-4 extraction sub-pipeline.

Open →
F1 · Architecture

AWS Architecture

Private VPC, scheduled-poll notifications, SQS between stages, Bedrock, and the CloudWatch → SNS → PagerDuty alerting.

Open →
F4 · AI strategy

AI Strategy

Prompt-engineering over Bedrock, versioned templates, schema-constrained output, embeddings, and evaluation.

Open →

RFQ Matching — Stages 5–6

ExtractedQuote → matched to a customer RFQ → submitted to the core system.

Data flow

Data Flow

Matching + submission (stages 5–6), plus the decomposed retrieve-then-rerank matching sub-pipeline.

Open →
Architecture

AWS Architecture

Tighter posture — no NAT, no internet egress; RFQ read-model sync, Fargate matching over pgvector, private submit.

Open →
AI strategy

AI Strategy

Three-model retrieve-then-rerank stack: embedding recall → cross-encoder precision → LLM disambiguation.

Open →

Cross-cutting

Shared data model and the API contract across both stages.

Data model

DB Schema

DDL for the Quotes side (ingestion) and the RFQ side (matching), plus the revision-as-new-version design.

Open →
Deliverable 4

API & Data Structures

REST endpoints, the SupplierQuoteObject schema + Pydantic models, and the FastAPI-on-Lambda deployment.

Open →