Documentation outline¶
This outline is a starting map for discussions about how to organize AI agent system design documentation. Section titles are provisional; the goal is coverage and traceability to sources, not final taxonomy.
1. Problem framing¶
- What problem the agent solves and what “done” means for a run.
- Harness versus runtime: prompts, tools, and skills versus durable execution, tenancy, and observability.
- Non-functional requirements: latency, cost, quality, reliability, and compliance.
2. Architecture views¶
- Control plane versus data plane for agent workloads.
- Single-agent loop versus multi-agent orchestration.
- Synchronous chat versus background jobs, cron, and proactive agents.
- Integration surfaces: MCP, A2A, webhooks, and bespoke APIs.
3. Runtime capabilities (production checklist)¶
Use Production requirements and runtime capabilities as the working matrix. Draft deep dives: runtime-capabilities/README.md (one page per row).
4. Data, memory, and state¶
- Short-term thread state versus long-term store semantics.
- Checkpointing, replay, and time travel for debugging.
- Retention, migration, and model/provider changes without losing organizational memory.
5. Safety, guardrails, and human oversight¶
- Middleware and deterministic policy enforcement.
- Human-in-the-loop patterns: approval gates, draft review, clarifying questions.
- Prompt injection, sandbox boundaries, and credential handling.
6. Operations¶
Draft: 05-operations.md: tracing, evals, streaming, DR, versioning, and cost.
7. Economics and platform constraints¶
- Inference scaling, caching, and cost optimization.
- Multi-tenancy, RBAC, and operator access models.
8. Ethics, compliance, and product risk¶
Draft: 06-product-and-compliance.md.
9. Reference library¶
- Primary deep reference: LangChain runtime article.
- Data-agent domain reference: Databricks Genie article.
- Case studies: Databricks Genie data agents (maps the Genie article onto this outline).
- Social and secondary framing: Sources.
10. Harness versus runtime and program fit¶
Draft: 03-harness-vs-runtime.md, 07-transmute-data-fit.md, 08-decision-log.md.
Open questions for adaptation¶
- Whether “harness” and “runtime” remain the top-level split or merge into a single “platform” view.
- How much LangSmith or LangChain-specific terminology to keep versus neutral labels.
- Which checklist rows become first-class chapters versus appendices.
- Where Transmute-Data-specific conventions (eval harnesses,
tdCLI, Coder workspaces) attach without turning this into an implementation guide too early.