What is GLChat DeepResearch?

GLChat DeepResearch is an experience-layer pipeline that defines product behavior for deep research in GLChat.

As an experience-layer pipeline, it is responsible for:

  • Interpreting user intent and research requirements

  • Applying product-level settings (profiles, configurations)

  • Preparing results for presentation in the chat UI

  • Handling user experience concerns (progress, citations, formatting)

What it does NOT do: Backend execution, infrastructure management, or reasoning design (these are delegated to lower layers).

Responsibilities

1. Pre-processing

  • Gathers and prepares context (attachments, conversation history, memory)

  • Interprets user intent and research requirements

  • Determines if deep research is needed or if normal generation suffices

2. Profile Selection and Resolution

Applies the user's selected profile and resolves to the appropriate backend implementation:

Profile
Backend
Characteristics

Essentials

OpenAI Deep Research

Fast, cost-effective, concise

Comprehensive

Parallel AI

Thorough, expensive, comprehensive

Internal

OpenAI + Google Drive

Internal data access

3. Research Execution

Invokes the appropriate DeepResearcher component (SDK layer) with prepared inputs.

4. Post-processing

  • Structures research results for chat presentation

  • Formats citations and references

  • Applies product-level policies before delivery

Pipeline Flow

GLChat DeepResearch depends on two other layers in the deep research architecture. This section provides a brief overview; see their respective documentation for details.

GL SDK DeepResearcher

GL SDK DeepResearcher is a library-level SDK abstraction that provides a unified, backend-agnostic interface for executing deep research.

What it does:

  • Exposes consistent research() interface across all providers

  • Handles streaming, events, and result structuring

  • Hides backend differences from pipelines

Implementations used by GLChat DeepResearch:

  • OpenAIDeepResearcher — For Essentials and Internal profiles

  • ParallelDeepResearcher — For Comprehensive profile

📚 For more details: GL SDK DeepResearcher Documentationarrow-up-right

Last updated