search

Other Pipeline

hashtag
No-Op Pipeline

The No-Op (No Operation) pipeline skips retrieval entirely. The LLM receives only the conversation history and the query — no documents, no web search, no database queries.

When to use it:

  • The chatbot should rely purely on the model's own knowledge

  • Retrieval is not applicable

  • No knowledge base is configured for the chatbot

The No-Op pipeline still receives the preprocessing output and still passes through postprocessing — only the retrieval and knowledge-augmented generation steps are skipped.

hashtag
GLChat Live Pipeline

The GLChat Live pipeline enables real-time, streaming conversations using the Gemini Live model. Unlike the Standard RAG pipeline, it is designed for continuous, low-latency interaction rather than single-turn question and answer.

When it is used:

GLChat Live is used when user want an interactive voice and live conversation scenarios

hashtag
Deep Research Pipeline

The Deep Research pipeline is designed for queries that require broad synthesis across multiple sources rather than a single targeted answer. It runs a multi-step research analysis and formats web-sourced references for the response.

When it is used:

Deep Research is activated when the request's search_type is one of:

  • ESSENTIALS_DEEP_RESEARCH

  • COMPREHENSIVE_DEEP_RESEARCH

  • INTERNAL_DEEP_RESEARCH

It also requires the BACKEND_ESSENTIALS_DEEP_RESEARCH_PIPELINE_HANDLER environment variable to be configured. The routing stage directs the message to a dedicated internal research handler before the pipeline executes.

hashtag
External Pipeline

The External Pipeline routes a request to any external service that exposes an OpenAI-compatible API endpoint. The project configuration provides the webhook URL, API key, and model name — GLChat creates a compatible LM invoker at runtime and delegates generation to the external service entirely.

When it is used:

This pipeline is configured per chatbot via the external pipeline metadata. The chatbot must have webhook_url, api_key, and external_model_name set. No special routing flag is needed — if the chatbot is configured with this pipeline, it will always route here.

For more detailed information, refer to this page

hashtag
Datasaur External Pipeline

The Datasaur External pipeline integrates GLChat with Datasaur's hosted LLM platform. It is designed specifically for chatbots where Datasaur manages the model and knowledge base, providing native citation support and Datasaur-specific reference formatting.

When it is used:

This pipeline is configured per chatbot. It requires citation support to be enabled (citations_enabled: true) in the chatbot configuration. The routing stage directs the message to this pipeline based on the chatbot's pipeline type assignment.

PreviousStandard RAGNextCustom Pipeline Development Guide

Last updated