HomeTruth
Resources
Document Ingestion & Structuring
Auto-typed · User-confirmed · Dual-stack
Smart document intake

Every upload becomes structured intelligence—fast.

HomeTruth scans any upload, builds a JSON view, matches against known schemas, and asks you to confirm the type. Once confirmed, data is saved as semi-structured facts and pushed into both private and (anonymized) public knowledge stacks.

Auto-detect

AI compares extracted JSON to known schemas to pick the right document class (e.g., EPC, insurance policy, electricity bill).

User confirmation

You approve the detected type before it’s stored, preventing misfiled documents.

Structured facts

Dates, parties, categories, and totals are captured as fields—ready for search, checklists, and alerts.

Dual-stack

Data lands in your private namespace; anonymized learnings can enrich the public stack for market context.

Flow snapshot
1

Upload: Drag/drop any PDF or image.

2

Scan & JSON: AI extracts text, tables, and metadata into a draft JSON.

3

Schema match: JSON compared to doc schemas (EPC, bill, survey, policy, quote, invoice).

4

User confirm: “Is this your electricity bill?” You confirm or correct.

5

Structured save: Key fields normalized; file + JSON linked.

6

Knowledge stacks: Private namespace updated; anonymized signals may flow to public stack.

Why this matters

Instant retrieval

Search “EPC expiry” or “latest electrician invoice” and get answers from structured fields, not just full-text.

Compliance & valuation

EPCs, warranties, and surveys roll into checklists and valuation signals automatically.

Contextual AI

Chat responses reference your documents and the public market, side by side.

Example extraction (electricity bill)

Detected fields

  • Document type: Electricity bill
  • Supplier: NorthGrid Energy
  • Service address: 14 Deacons Walk, Bristol
  • Billing period: 01 Feb 2025 – 28 Feb 2025
  • Amount due: £128.40
  • Due date: 12 Mar 2025
  • Tariff: FlexSaver 2025

Stored JSON (private namespace)

{
  "doc_id": "DOC-8821",
  "doc_type": "utility_bill_electricity",
  "source_file": "bill-feb-2025.pdf",
  "billing_period": { "start": "2025-02-01", "end": "2025-02-28" },
  "amount_due_gbp": 128.40,
  "due_date": "2025-03-12",
  "supplier": "NorthGrid Energy",
  "address": "14 Deacons Walk, Bristol",
  "tariff": "FlexSaver 2025",
  "namespace": "user/lujain",
  "ingested_at": "2025-03-01T10:18:00Z"
}

Schema matching logic

Compare JSON to known schemas

Field presence, regex patterns, totals, and date ranges score against EPC, bill, survey, policy, quote, or invoice schemas.

User-in-the-loop

If the top score is low, we prompt for confirmation or re-typing to avoid misclassification.

Linked assets

Original file, parsed text, and structured JSON are bound for auditability.

Knowledge stacks

Private stack (per user)

Namespaced: user/{id}. Drives personal chat answers, tasks, renewals, and valuations.

Public stack (anonymized)

Aggregated signals (e.g., tariff trends, survey issues) enrich market-level responses without exposing PII.

Cross-talk

Users can ask about “my EPC vs market averages” because both stacks are queryable with guardrails.

What the user sees

Upload → detected type

“We think this is an electricity bill — confirm?” with type options and a preview.

Structured summary

Key fields rendered as chips; warnings if due dates are near or totals jump abnormally.

Search & chat

Search filters by doc type/date; chat references both your docs and public context.