Dashboard/Intelligence-Research-OS/wiki/requirements.md
Intelligence-Research-OS

wiki/requirements.md

Intelligence Research OS — Requirements

Current Product Goal

Build a local-first intelligence workbench on top of ResearchModule that supports mission creation, source-backed research, signal triage, evidence validation, multi-POV synthesis, output packs, and deltas over time.

Implemented Requirements

  • Python package research_module with file-gated pipeline.
  • CLI for scoping, continuing, mission create/run/status/show output.
  • Local web app command: uv run research app --workspace-parent <dir> --port 8765.
  • Local browser UI for Mission Control, Mission Builder, Source Radar, Signal Inbox, Evidence Board, Synthesis Studio, Output Workspace, and Delta Timeline.
  • FastAPI endpoints for health, workspace, missions, sources, signals, evidence, synthesis, and outputs.
  • Strict citation/source/evidence validation primitives.
  • Six-screen GPT Image 2 UX mockups and IA diagram.
  • Twelve mission templates with explicit support levels.
  • Competitive Intelligence deterministic demo workflow with source-backed signals, evidence review, multi-POV synthesis display, output export, and delta timeline.

Acceptance Rules

  • No accepted claim or finding without source/evidence refs.
  • No final output can be considered complete unless citation audit passes.
  • Local artifacts remain the source of truth; no parallel UI database in v1.
  • Source family readiness must never reveal secret values.
  • Every UI status must show a next action or clear empty state.

Next Required Functionality

  • Wire UI Run Mission into live Scoper/AAR/Fan-out/Synthesis execution.
  • Generate signals.json and signal_clusters.json from dossier artifacts.
  • Implement source-backed Competitive Intelligence battlecards/digests.
  • Add delta memory across repeated mission runs.
  • Add dedicated source adapters after provider-search/web-fetch v1 is stable.

Latest Acceptance Target

The next acceptance target is mission-to-engine continuity: creating and running a mission from the UI should produce or continue real ResearchModule run artifacts, persist the generated research run id on the mission report, and expose the live phase/next-action state back in the workbench.