# ADR-007: Research Workspace

- **Status:** Approved
- **Decision:** Aggregate research documents into company and topic workspaces
- **Implemented:** Sprint 1 Phase 3

## Context

The Research Engine can normalize public website and RSS/Atom content into `ResearchDocument` objects. Users need a stable workspace abstraction that gathers related evidence by company or topic without adding AI summarization, trading signals, or external financial APIs.

## Decision

Axiom will introduce `ResearchWorkspace` in `packages/research/workspace/`.

A workspace contains:

- a stable workspace id;
- a company or topic subject;
- matching `ResearchDocument` objects;
- distinct source names;
- timeline entries generated from document timestamps;
- a workspace refresh timestamp.

The workspace is built by `ResearchService` from normalized document storage. It does not own connector fetching, storage persistence, search ranking, or AI reasoning.

## Consequences

- Company and topic views can share one model.
- Workspace generation remains deterministic for a given store and clock.
- UI and future API layers can consume one workspace contract.
- Refreshing a workspace reloads current normalized documents rather than mutating external sources.

## Boundaries

- No AI summarization is performed.
- No SEC or PDF connector behavior is introduced.
- No trading decisions, rankings, or signals are produced.