Imprint

Your codebase,
remembered.

The shared library knows how everyone builds. Imprint knows how you do. It scans your project, extracts the patterns you already trust, and registers them as private recalls — composable alongside the public library.

See how it works →

Bytes uploaded anywhere

strata — imprint scan

$ strata imprint scan ./src

✓ Scanning src/ ············ 1,847 files

✓ Extracting exports ·········· 312 functions found

✓ Identifying patterns ········· 89 recall candidates

✓ Deduplicating against library ···· 14 overlap(s) skipped

→ Writing 75 recalls to recalls/local/

auth.user.validate.local.v1

api.response.format.local.v1

db.pool.config.local.v1

+ 72 more

✓ Library index rebuilt ········· 7,900 total recalls

How Imprint works

Four steps from scan to recall.

Scan

Imprint walks your src/ tree. Every JavaScript and TypeScript file is parsed for exported functions, classes, and utilities.

Extract

Signatures, JSDoc, and call examples are pulled from your source. Patterns are matched against Strata's recall schema — inputs, outputs, domain, tags.

Register

Each valid pattern becomes a local recall in recalls/local/. The library index is rebuilt in-place. Nothing is transmitted.

Compose

Your local recalls enter the scoring pool alongside the public library. Strata retrieves them exactly as it retrieves library recalls — no special handling.

What gets captured

Your conventions. Not someone else's.

The generic library covers the ecosystem. Imprint covers you — your project's specific patterns that no npm package will ever know about.

Shared Utilities

Project-specific helpers that every module imports but nobody documents. Formatters, validators, converters — the internal stdlib of your codebase.

Domain Services

Business logic shaped to your data models and API conventions. Your auth flow, your response envelope, your error taxonomy — none of these exist in any library.

Config & Middleware

Environment wiring, database pool setup, shared Express middleware — the scaffolding you copy between services. Imprint captures the pattern once, recalls it everywhere.

Privacy model

Your code stays
on your machine.

Imprint is a local CLI command, not a service. It reads your source files and writes recall metadata to your project directory — the same directory Strata already reads from.

There is no Imprint server. No account required. No telemetry. The scan runs offline, the recalls are stored locally, and the retrieval is local too.

Source files are read, never transmitted. Imprint processes files in memory and writes only metadata (function names, tags, signatures).

Local recalls are scoped to your workspace. They are written to your project's recalls/local/ and never synced to the public library.

No internet connection required. Scan, register, and retrieve all run without any network call.

Add recalls/local/ to .gitignore to keep them entirely off version control, or commit them to share with your team.

Your source files

src/ — read-only, processed in memory

Imprint CLI

Runs locally — no server, no account

recalls/local/

Metadata only — no source code stored

Internet

Not contacted during scan or retrieval

The difference

Library alone vs. Imprinted.

The same query. The same scoring. But with Imprint, your own patterns surface ahead of generic ones when they're a better match.

Library only

the public library

capability: "validate user input against project schema"

validation.schema.zod.v1

Generic — Zod schema validation

validation.schema.joi.v1

Generic — Joi schema validation

validation.schema.ajv.v1

Generic — AJV JSON schema

With Imprint

the library + your recalls