EPIC Goal

To engineer and deliver the "Notebooks" functional area as a separate lightspeed-notebook dynamic plugin. This Epic focuses on implementing the backend RAG pipeline, document ingestion logic, and the coordination of specific frontend views for Notebook management with the UI team.

Background/Feature Origin{}

Currently, Developer Lightspeed offers general chat capabilities. Users require answers grounded strictly in their own private project documentation (architecture diagrams, internal policies, logs). We are introducing "Notebooks" to allow users to isolate data into workspaces so AI answers are precise and citation-backed. This work builds upon research spikes RHDHPLAN-245.

Why is this important?

User Scenarios

Dependencies (internal and external)

• Internal (Critical): Coordination with the "Lightspeed Chat UI Enhancements" task. The Notebooks feature relies on the new Tabbed Navigation and Layout structure being delivered by the UI team.

• Upstream/Architecture:

• • Vector Database setup (pgvector, Milvus, or FAISS) and Embedding Model implementation.

• • Validation of supported LLMs (Granite 7B, Llama 3 8B+).

• Infrastructure: Resource guardrails implementation (max 10MB upload, 100k tokens).

Acceptance Criteria

• Plugin Architecture: The feature is wrapped in a lightspeed-notebook dynamic plugin, distinct from the core chat code.

• Backend RAG Pipeline:

• • Ingestion engine processes allowed files (PDF, URL, MD, TXT, DOCX) and rejects unsupported types.

• • Vector DB stores embeddings with strict isolation per notebook.

• Integration: The Notebooks plugin renders correctly within the new Tabbed interface provided by the UI team.

• Configuration: Feature visibility is controlled via lightspeed.notebooks flag and RBAC policies.