Content Summarization

A student opens GamiBot Chat in a course, chooses Summarize Subject, and asks for a summary. The Chat block sends the request to the Moodle Manager plugin, which selects the standard Summarize Langflow purpose and supplies the Moodle domain and course context.

The current hosted language model is Qwen3.5 served by vLLM. Course material is prepared by the ingestion service with Docling, embedded with the nomic-text model through Ollama, and indexed in Qdrant. How the Summarize flow selects and presents that material is controlled by the deployed Langflow flow.

Summaries depend on material being available in the remote index. Moodle's global ingestion queue sends file updates gradually, and an HTTP acknowledgment does not mean the receiver has finished indexing. Check Data ingestion if a new or changed document does not appear in answers.

The Moodle plugin maintains the Summarize endpoint and input IDs. Site administrators enter one shared Langflow API key for all three purposes; they do not configure model parameters in the Moodle setup form. See Langflow connections and Moodle setup.

The exact prompts, retrieval thresholds, response format and latency depend on the hosted flow and are not fixed by the Moodle plugin. Review generated summaries against the course material before relying on them.