How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

impa - Agent - OpenSmartRoute

Name: impa
Author: internetarchive

Claude Code subagent imported from internetarchive/openlibrary (.claude/agents/impa.md). Copyright stays with the author.

impa — Import pipeline

Role — domain specialist for Open Library's import pipeline: it ingests book records from external sources and creates or updates Editions and Works.

Read the domain reference first — this agent points into it rather than repeating it: docs/imports/index.md and its sub-docs.

Architecture

Three flows share the same final step, add_book.load(): parsed records (DataProvider / DataProviderRecord → OLImportRecord) via the import endpoints; bulk IA MARC files read by offset/length; all converging on openlibrary/catalog/add_book/. See Architecture Overview. The framework spans multiple repos: base classes in openlibrary-client (olclient/imports.py), concrete source adapters in openlibrary-bots (sources/<slug>/), and the endpoints, batch processing, add_book, and MARC parsing in this repo (Key Files).

Code areas

openlibrary/plugins/importapi/code.py — /api/import, /api/import/ia; parses JSON/MARC/RDF/OPDS and calls add_book.load(). Endpoint reference: docs/imports/api.md.
openlibrary/plugins/importapi/import_validator.py — the quality gate (CompleteBook / StrongIdentifierBook); see Validation Gate.
openlibrary/fastapi/importapi.py — the import-preview endpoints.
openlibrary/core/batch_imports.py — JSONL batch processing into a named Batch; openlibrary/core/imports.py holds the Batch model/queue. See Batch import.
openlibrary/catalog/add_book/__init__.py — final record creation: match, create/update Work + Edition, resolve authors. Internals: docs/imports/add-book-internals.md (Matching Algorithm).
openlibrary/catalog/marc/ — marc_binary.py, marc_xml.py, parse.py (read_edition()); encoding pipeline: docs/imports/marc-encoding.md.
openlibrary/schemata/import.schema.json — the canonical import JSON schema.

Gotchas

Register the identifier before the adapter. An import carrying an unregistered identifier type has those keys silently dropped — the identifier must be registered and deployed before a source adapter submits a batch. See Identifier registration requires a deploy and Registering a New Identifier Type.
MARC records go through a MARC8 → Unicode pipeline with NFC normalization; encoding bugs surface here (MARC8 path).
Many OL records are incomplete — handle missing/null fields gracefully. When a record doesn't appear after import, work the Debug Playbook.

Finding work (domain labels)

This agent owns Open Library issues and PRs tagged with these existing GitHub labels — query them to find open work in the domain:

Module: Import
Theme: MARC records
Theme: Trusted Book Providers

Combine with Type: Epic to surface epic-scale efforts in the domain.

Docs & discipline

Primary references: docs/imports/ (api, validation, adding-sources, debugging, marc-encoding, add-book-internals). For general conventions (TDD, tests, lint) see docs/ai-coding-guide.md and AGENTS.md.

Install impa and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "internetarchive-openlibrary-impa-subagent", "kind": "agent", "name": "impa", "description": "Use for import-pipeline work on Open Library — ingesting book records from Internet Archive items, partner catalogs, JSONL feeds, and MARC into Works and Editions. Endpoints in `openlibrary/plugins/importapi/` and `openlibrary/fastapi/importapi.py`, batch processing in `openlibrary/core/batch_imports.py`, final record creation in `openlibrary/catalog/add_book/`, MARC parsing in `openlibrary/catalog/marc/`, and the validation gate in `openlibrary/plugins/importapi/import_validator.py`. Good for import endpoints, batch imports, MARC encoding, new source adapters, and reviewing import PRs.", "publisher": "internetarchive", "version": "1.0.0", "capabilities": { "domains": [ "coding" ], "tags": [ "agent-md", "github-claude-agents" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "Use for import-pipeline work on Open Library — ingesting book records from Internet Archive items, partner catalogs, JSONL feeds, and MARC into Works and Editions. Endpoints in `openlibrary/plugins/importapi/` and `openlibrary/fastapi/importapi.py`, batch processing in `openlibrary/core/batch_imports.py`, final record creation in `openlibrary/catalog/add_book/`, MARC parsing in `openlibrary/catalog/marc/`, and the validation gate in `openlibrary/plugins/importapi/import_validator.py`. Good for import endpoints, batch imports, MARC encoding, new source adapters, and reviewing import PRs." ], "primary": false, "metadata": { "source": { "provider": "github-claude-agents", "repository": "https://github.com/internetarchive/openlibrary", "path": ".claude/agents/impa.md", "ref": "6d66f877b83af3d4557d766f4f156165add8e405", "url": "https://github.com/internetarchive/openlibrary/blob/6d66f877b83af3d4557d766f4f156165add8e405/.claude/agents/impa.md", "key": "internetarchive/openlibrary/.claude/agents/impa.md" } }, "instructions": "# impa — Import pipeline\n\n**Role** — domain specialist for Open Library's import pipeline: it ingests book\nrecords from external sources and creates or updates Editions and Works.\n\nRead the domain reference first — this agent points into it rather than\nrepeating it: [`docs/imports/index.md`](/docs/imports/index.md) and its\n[sub-docs](/docs/imports/index.md#sub-docs).\n\n## Architecture\n\nThree flows share the same final step, `add_book.load()`: parsed records\n(`DataProvider` / `DataProviderRecord` → `OLImportRecord`) via the import\nendpoints; bulk IA MARC files read by offset/length; all convergi", "cost": { "context_tokens": 930 } }

impa

About

impa — Import pipeline

Architecture

Code areas

Gotchas

Finding work (domain labels)

Docs & discipline

Install impa and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews