PHPackages egroupware/rag - PHPackages - PHPackages [Skip to content](#main-content)[PHPackages](/)[Directory](/)[Categories](/categories)[Trending](/trending)[Leaderboard](/leaderboard)[Changelog](/changelog)[Analyze](/analyze)[Collections](/collections)[Log in](/login)[Sign up](/register) 1. [Directory](/) 2. / 3. [Utility & Helpers](/categories/utility) 4. / 5. egroupware/rag ActiveEgroupware-app[Utility & Helpers](/categories/utility) egroupware/rag ============== A RAG system for EGroupware. 26.6.20260619(2w ago)01.4k1[2 PRs](https://github.com/EGroupware/rag/pulls)1PHP Since Jan 6Pushed 2d ago1 watchersCompare [ Source](https://github.com/EGroupware/rag)[ Packagist](https://packagist.org/packages/egroupware/rag)[ RSS](/packages/egroupware-rag/feed)WikiDiscussions master Synced 2d ago READMEChangelogDependenciesVersions (21)Used By (1) RAG (Retrieval-Augmented Generation) for EGroupware =================================================== [](#rag-retrieval-augmented-generation-for-egroupware) Features (planned or already implemented) ----------------------------------------- [](#features-planned-or-already-implemented) - stores embeddings for all supported applications in table `egw_rag` - stores a fulltext index for all supported applications in table `egw_rag_fulltext` - create AND keep the above indexes up-to-date when entries are added, updated or deleted - provides search primitives applications can use to provide: - `Rag\Embeddings::searchEmbeddings()` a semantic search by given search-pattern - `Rag\Embeddings::searchFulltext()` a fulltext search either in [natural language mode](https://mariadb.com/docs/server/ha-and-performance/optimization-and-tuning/optimization-and-indexes/full-text-indexes/full-text-index-overview#in-natural-language-mode) or in [boolean mode](https://mariadb.com/docs/server/ha-and-performance/optimization-and-tuning/optimization-and-indexes/full-text-indexes/full-text-index-overview#in-boolean-mode), if operators are used (the default preference to add an `*` after each word, also swiches automatically to boolean mode!) - `Rag\Embeddings::search()` hybrid search combining semantic and fulltext search - the primitives return an array with ID => distance or relevance score of the embeddings or fulltext index - UI for an application independent search - show a (bigger) search-field - show a (multiple) application selection, default all supported apps - show a NM list with link-title, app-name and ID and onclick action to edit the entry - replace old EPL search in avatar-menu with RAG, if available - new RAG preference what to use in default search, if RAG is available (implementation in Api\\Storage): - fulltext search (default) - hybrid search (if RAG/Embedding is configured) - semantic search / RAG only - legacy search of the apps - search will still support # to find the ID - apps can prefer/overwrite the above, e.g. Addressbook to use its own phone-number search - fulltext search supports [boolean search operators](https://mariadb.com/docs/server/ha-and-performance/optimization-and-tuning/optimization-and-indexes/full-text-indexes/full-text-index-overview#in-boolean-mode) - preference to automatically append an asterisk (\*) after each search pattern, which is not enclosed in quotes ("), to also match words beginning with the pattern > Please note: > > the fulltext index does NOT find patterns within a word, only complete words or words starting with the pattern, if the above preference is set (by default), or the `*` operator is added manually after the pattern e.g. `some*` will also find `something`, but not `awesome`! - preference to keep sort-order from apps (default), or use relevance from RAG/fulltext search, when search in the apps - UI of RAG allows sorting the search result by distance&relevance (default), modification date, distance or relevance alone - change search in filterbox, to include: - selection which search-type to use incl. legacy search - fulltext operator help (like in RAG UI) - cache embeddings of the search-pattern to not always have to request them again - store sha256 of chunks, to not regenerate embeddings of unchanged chunks - support other embedding models AND modify schema to support its number of dimensions (the latter is not yet implemented!) - make the results of the above searches available to AI agents / integrate with them Supported applications ---------------------- [](#supported-applications) - InfoLog - Tracker - allow app-plugins in apps namespace `EGroupware\\Rag` instead directly in RAG's `EGroupware\Rag\Embeddings\` - Addressbook, Calendar, Timesheet, ProjectManager and Phpbrain (aka KnowledgeBase) - other egroupware applications: Invoices, Resources, Records, ViDoTeach, ... Requirements and installation ----------------------------- [](#requirements-and-installation) - MariaDB 11.8 for the required vector type to store and search for embeddings. Add the following to your docker-compose.override.yml to use MariaDB 11.8 (also works with EGroupware 23.1) ``` # /etc/egroupware-docker/docker-compose.override.yml service: db: image: mariadb:11.8 environment: - MARIADB_AUTO_UPGRADE=true ``` > The above will happen automatic with egroupware-docker-26 package (once it's released). - [bge-m3](https://ollama.com/library/bge-m3) Embedding Model via either: - an [Ollama installation](https://ollama.com/blog/ollama-is-now-available-as-an-official-docker-image) with bge-m3 installed: `ollama pull bge-m3:latest` - or an OpenAI compatible endpoint and API key to access bge-m3 e.g. via IONOS AI Hub - Make sure ollama is [reachable](https://docs.ollama.com/faq#how-can-i-expose-ollama-on-my-network) from the RAG container, by default it binds to 127.0.0.1:11434 - RAG application repo must be cloned into the EGroupware source directory (`/usr/share/egroupware` in an on-premise installation) > RAG application will be delivered with regular EGroupware 26.x container images from next pre-release on. - the RAG app then needs to be configured: `Admin > Applications > RAG > App Configuration`, by default it only does the fulltext index App Configuration ----------------- [](#app-configuration) ### General Configuration [](#general-configuration) - URL and optional API key for the [OpenAI compatible endpoint](https://docs.ollama.com/api/openai-compatibility) - Add "/v1/" to your Ollama URL to access the OpenAI compatible endpoint - (optional, without RAG just uses/creates the fulltext index) > Ollama: you can NOT use the default localhost in an other container, use e.g. the docker0 address 172.17.0.1! ### Embedding Configuration [](#embedding-configuration) - name of an embedding model to use, defaults to `bge-m3` > Currently 1024 dimensions are hard-coded in the schema, you need to change the schema manually for a different value! - chunk-size and overlap for chunking the texts of the application entries (default is 500 and 50 chars) - setting to minimize chunks by concatenating all texts (before splitting them into chunks), instead of chunk-splitting them all separate > Calculating and storing Embeddings, and to much lower extend also the Fulltext index, costs a not to diminishing amount of storage for the tokens! Usage for the supported apps ---------------------------- [](#usage-for-the-supported-apps) - RAG search over all (supported) application is now shown in the avatar menu under `search` - supported apps use by default the fulltext index for searching - the above can be changed permanently via a preference or - through the small RAG user-interface show in the filterbox ### Health Score 47 — FairBetter than 93% of packages Maintenance98 Actively maintained with recent releases Popularity22 Limited adoption so far Community16 Small or concentrated contributor base Maturity44 Maturing project, gaining track record Bus Factor1 Top contributor holds 97.6% of commits — single point of failure How is this calculated?**Maintenance (25%)** — Last commit recency, latest release date, and issue-to-star ratio. Uses a 2-year decay window. **Popularity (30%)** — Total and monthly downloads, GitHub stars, and forks. Logarithmic scaling prevents top-heavy scores. **Community (15%)** — Contributors, dependents, forks, watchers, and maintainers. Measures real ecosystem engagement. **Maturity (30%)** — Project age, version count, PHP version support, and release stability. ### Release Activity Cadence Every ~8 days Total 20 Last Release 19d ago ### Community Maintainers ![](https://www.gravatar.com/avatar/85a8407098a44193698887e739a263c1623a169548b3867103015d2b3c9eb78d?d=identicon)[ralfbecker](/maintainers/ralfbecker) --- Top Contributors [![ralfbecker](https://avatars.githubusercontent.com/u/972180?v=4)](https://github.com/ralfbecker "ralfbecker (81 commits)")[![nathangray](https://avatars.githubusercontent.com/u/10407710?v=4)](https://github.com/nathangray "nathangray (2 commits)") ### Embed Badge ![Health badge](/badges/egroupware-rag/health.svg) ``` [![Health](https://phpackages.com/badges/egroupware-rag/health.svg)](https://phpackages.com/packages/egroupware-rag) ``` ### Alternatives [adamwojs/php-cs-fixer-phpdoc-force-fqcn 181.2M63](/packages/adamwojs-php-cs-fixer-phpdoc-force-fqcn)[dantleech/fink Checks Links 2266.9k2](/packages/dantleech-fink) PHPackages © 2026 [Directory](/)[Categories](/categories)[Trending](/trending)[Changelog](/changelog)[Analyze](/analyze)