survos/bookworm-bundle

Index versioned Markdown documentation for source-grounded Symfony AI assistants.

2.26.10 2026-08-31 12:01 UTC

Requires

• php: ^8.4
• ecourty/text-chunker: ^1.2
• survos/kit-bundle: ^2.5
• symfony/ai-store: ^0.13
• symfony/console: ^8.1
• symfony/filesystem: ^8.1
• symfony/finder: ^8.1
• symfony/framework-bundle: ^8.1
• symfony/messenger: ^8.1
• symfony/string: ^8.1
• symfony/uid: ^8.1

Requires (Dev)

• knplabs/github-api: ^3.19
• phpunit/phpunit: ^12.0|^13.0

Suggests

• ext-zip: Extracts repositories downloaded by bookworm:fetch.
• knplabs/github-api: Enables bookworm:fetch for GitHub repositories.
• symfony/ai-postgres-store: Stores document embeddings in PostgreSQL with pgvector.

MIT

• symfony
• documentation
• markdown
• ai
• rag
• vector-search

Published on Packagist.org on 2026-08-31 12:02 UTC

README

Bookworm turns versioned Markdown collections into source-grounded Symfony AI corpora. It discovers Markdown files, preserves heading hierarchy while chunking, adds stable source metadata, and sends documents through an app-configured indexer.

The input contract is Markdown: survos/doc-bundle may generate it and Bookworm may consume it without coupling the two bundles.

survos_bookworm:
    corpora:
        civicrm_user:
            directory: '%kernel.project_dir%/var/bookworm/civicrm/user-en'
            indexer: 'ai.indexer.civicrm_user'
            repository: 'civicrm/documentation/docs/user-en'
            canonical_base_url: 'https://docs.civicrm.org/user/en/latest/'
            max_chunk_size: 4000

bin/console bookworm:index civicrm_user          # dispatch to Messenger
bin/console bookworm:index civicrm_user --sync   # index in this process
bin/console bookworm:status

Route Survos\BookwormBundle\Message\IndexCorpus to the desired Messenger transport. Bookworm deliberately does not select an embedding model or vector store; the configured Symfony AI indexer owns those decisions.

Optional GitHub fetching

composer require knplabs/github-api symfony/http-client nyholm/psr7
bin/console bookworm:fetch github.com/symfony/symfony-docs

The destination defaults to var/bookworm/<owner>/<repository>. GitLab and other source hosts can be added behind the source boundary without changing indexing.