Lease Lens: a local Gemma 4 reads the lease, plain code does the math

Description and use

Lease Lens uses a local Gemma 4 E2B model to read residential leases, validates extracted data with rules, and computes key dates and late‑fee tables, keeping tenant data on the user’s machine.

Problem

Small landlords and property managers need to extract key dates, late‑fee tables, and state compliance checks from residential leases without manual effort or uploading tenant data.

Outcome

Lease Lens processes a lease in ~29 s on an 8‑core CPU, outputs structured data, and flags potential compliance issues such as excessive late fees. It keeps tenant data local and offers a free CLI with paid add‑ons.

Workflow

• 1. Load lease text into the CLI.
• 2. Send the text to a local llama.cpp server running Gemma 4 E2B with a JSON schema request.
• 3. Receive structured JSON containing {value, quote} for each field.
• 4. Verify each quote matches the lease text and that numbers appear in their own quote; re‑parse dates and fee wording.
• 5. Compute grace periods, late‑fee tables, renewal notice dates, and deposit‑return deadlines in plain Python code.
• 6. Output .ics calendar, leases.csv row, and audit report with ok/FIXED/CHECK markers.

Evidence

Lease Lens takes a pile of residential leases and gives you, for each one: the key dates, a late‑fee table, a state check of the late‑fee terms for Colorado, New York and Texas; an .ics calendar; a leases.csv row; a field‑by‑field audit.

Source

Real run on an 8-core CPU with no GPU: about 29 s per lease.

Source

The model reads; tested code does every date and dollar. Nothing is uploaded anywhere: tenant names, addresses and rent stay on your machine.

Source

We wrote 7 synthetic leases with hand‑made answer keys, 15 scored fields each: ... The model returned a 5% fee and quoted a sentence with no percentage in it, and the first version let that through.

Source