Use case
Receipts accumulate across folders and devices, making reliable lookup, normalization, and archival slow and repetitive.
Receipt Finder
Receipts accumulate across folders and devices, making reliable lookup, normalization, and archival slow and repetitive.
How it solves it
A local-first tool searches receipt files, extracts useful metadata, reduces image size, and places organized copies into a predictable destination.
Architecture
The design separates inputs, orchestration, AI assistance, and human verification. Secrets and environment-specific identifiers belong in runtime configuration, never in this page or source control.
Software stack
Python, Pillow, filesystem scanning, OCR-ready processing boundaries, macOS packaging scripts, Firebase-ready static assets.
Source-controlledTestableObservableLeast privilege
AI and AgenticOps
AI can help classify receipt fields only after local redaction; AgenticOps can schedule indexing or reporting without uploading private receipts by default.
Technical operating model
Ingest
Accept structured inputs and normalize them.
Accept structured inputs and normalize them.
Reason
Retrieve approved context and produce a bounded plan.
Retrieve approved context and produce a bounded plan.
Verify
Apply guardrails, human review, and evidence checks.
Apply guardrails, human review, and evidence checks.
The privacy boundary is local processing first, with explicit user control over any later synchronization or sharing.