Skip to content

Latest commit

 

History

History
53 lines (46 loc) · 3.72 KB

File metadata and controls

53 lines (46 loc) · 3.72 KB

Trinity - Autonomous Agent Orchestration Platform - Requirements

SINGLE SOURCE OF TRUTH — all development must trace back to these documents. Update the relevant area file BEFORE implementing any new feature.

This document is the index for Trinity's requirements. The requirements themselves live in per-area files under requirements/ — the monolith was split for navigability and to remove merge-conflict pressure (#1406). This file stays the stable entry point, so existing links to requirements.md still resolve.

Start with requirements/overview.md for the platform Vision, Status Labels, Non-Functional Requirements, and Out-of-Scope list.

Area files

Area file Covers
overview.md Vision, Status Labels, Non-Functional Requirements, Out of Scope
core-agent.md Core agent management, template system, chat & terminal, activity monitoring, agent collaboration
auth.md Authentication & authorization (email/admin login, roles, JWT)
credentials.md Credential management (file injection, encrypted exports)
mcp.md MCP server, A2A discoverability, per-agent MCP exposure
scheduling.md Scheduling & autonomy, agent-defined pipelines, schedule timeout validation, MCP chat timeout recovery, sequential agent loops
github.md GitHub integration (sync, PAT, recovery)
infrastructure.md Infrastructure, platform operations, CLI tool, canary invariant harness, enterprise edition architecture, build info surface
content-files.md Content & file management, image generation, avatars, agent runtime data
runtimes.md Multi-runtime support, OpenAI Codex, voice chat, VoIP telephony
public-access.md Public access, Nevermined payments, mobile admin PWA
security.md Security & compliance, operator queue / Operating Room, agent guardrails
skills.md Skills management (GitHub-based), playbooks tab
lifecycle-observability.md Agent soft-delete & retention, compatibility validation, first-run operator profile, agent-reported structured reports
roadmap.md Advanced features, planned features, process engine, future vision

How requirements are organized

  • One area per file. Each file groups a cross-cutting product area. A single requirement lives in exactly one area file.
  • Where does a new requirement go? Add it to the area file that owns its product surface (use the table above). If a new requirement spans areas, put it in the file matching its primary surface and cross-link the others with a relative link (e.g. [MCP exposure](mcp.md)). If nothing fits, it belongs in roadmap.md until it earns its own area — do not re-grow this index into a monolith.
  • Identify by feature tag, not ordinal. Existing sections keep their historical ## N. numbers (verbatim, no renumbering), but that global ordinal sequence is retired — it already collided (two ## 35.). New requirements get a short heading with their issue/feature tag (e.g. ## Fleet Snapshots (#1234) or a FOO-001-style ID), not the next number.
  • This index is thin by contract. It carries no requirement text — only the map and these rules. Keep it that way (a CI guard enforces the size cap and flags dangling requirements/ links).