Open Questions

Tracks unresolved questions, uncertainties, and items requiring follow-up or user input.

Format

[Q-XXX] <question>
  Status: OPEN | RESOLVED | DEFERRED
  Context: <where this came up>
  Resolution: <answer, if resolved>

Questions

Scope & Priority

[Q-001] What is the target deployment environment for the agent harness (local-first, cloud, hybrid)? Status: OPEN Context: Impacts architecture decisions across all research areas Resolution: Research recommends a hybrid approach — local-first (SQLite) with cloud scaling path (PostgreSQL + Redis). See 18_architecture_recommendations/README.md.

[Q-002] What is the priority ordering of research topics? Should all 19 areas be researched in parallel or sequentially? Status: RESOLVED Context: Project initialization Resolution: All 19 areas were researched comprehensively. Sequential deep-dives proved most effective for cross-referencing.

[Q-003] What LLM providers must be supported for "model-agnostic" to be satisfied (OpenAI, Anthropic, Google, local/Ollama, etc.)? Status: RESOLVED Context: Impacts 13_model_agnostic_harness_architecture Resolution: 25 models across 11 providers profiled. Backends: Ollama (local), OpenRouter (hosted), optional LiteLLM proxy, or direct APIs — all via OpenAI-compat clients. See 13_model_agnostic_harness_architecture/model_landscape_june_2026.md.

[Q-004] Is there a preferred programming language/runtime for the harness backend (Python, TypeScript/Node, Rust, etc.)? Status: OPEN Context: Impacts 14_backend_agent_stack_engineering Resolution: Research recommends Python for agent core + TypeScript for gateway/frontend. See 18_architecture_recommendations/README.md.

[Q-005] What is the intended end-user profile — developers building agents, or end-users interacting with agents? Status: OPEN Context: Impacts 15_frontend_react_vite_agent_stack