| # | Document | Description |
|---|---|---|
| 01 | Vision & Overview | What GearTrain is, why it exists, core principles, target users, success criteria |
| 02 | Architecture | Five-layer architecture: agents, workflows, teams, engine, channels. Foundation: memory & integrations |
| 03 | Core Concepts | Glossary of all key terms and their relationships. Configuration hierarchy |
| 04 | Memory & Knowledge System | Memory vs knowledge base distinction, four memory scopes (agent-instance, workflow, workspace, agent-level), dual-format storage, guardrails, MVP vs future |
| 05 | Example: Software Development | Full software development workflow with all sub-pipelines. Dogfooding plan |
| 06 | MVP Scope & Plan | What's in/out of MVP, key tradeoffs, tech stack, phased implementation plan, risk register, open questions |
| 07 | Design Notes | Extended concepts: self-bootstrapping, agent as an abstraction (cli + langchain), user-controlled IDE/CLI agents, task assignment, PM agent, integration patterns, non-software workflows, memory future-proofing, dreaming, git strategy, serverless |
| 08 | MVP Definition Contracts | Validation-ready contracts for workspace, engine, agent, workflow, memory, node types, and plain text outputs |
| 09 | UI and Observability Spec | Post-MVP UI target for runs, workflows, checkpoints, memory, tool calls, errors, timings, and observability dashboards |
For a new team member: 01 → 03 → 02 → 04 → 05 → 06 → 08 → 09 → 07
For a quick overview: 01 → 06 → 08
These documents represent the research and ideation stage output. They are intended to be reviewed, debated, and refined by the team before development begins. Key areas marked [to be defined] need team input.
Edit the markdown files directly. Keep the YAML examples consistent with the Pydantic model definitions once implementation begins. Add new documents with the next sequential number.