Prompt file imported from vkondi/knowledge-onboarding-agent (
.github/prompts/architecture-discussion.prompt.md). Copyright stays with the author.
--- mode: ask description: Start an architecture discussion for Knowledge Onboarding Agent. Loads project context and frames the decision correctly.
Architecture Discussion - Knowledge Onboarding Agent
Context
I am working on Knowledge Onboarding Agent, a local-first AI knowledge system.
Before we discuss, please acknowledge these constraints (from docs/constraints/runtime-constraints.md):
- Entirely local - no cloud AI APIs
- 16GB RAM laptop, CPU-only inference
- Ollama for all LLM and embedding inference
- Python 3.11+, LlamaIndex, ChromaDB
And these architecture rules (from .github/copilot-instructions.md):
- Pipeline stages are isolated (no cross-stage imports)
- Interfaces are Protocols, not ABCs
- No hardcoded values - everything from
config/settings.yaml
Decision to Discuss
Decision: [What needs to be decided?]
Context: [Why does this decision need to be made now? What triggered it?]
Constraints that apply:
- [List any specific constraints from the runtime-constraints doc or project conventions]
Request
Please:
- Identify 2–3 options for this decision
- For each option, list pros and cons relative to our constraints
- Recommend an option with reasoning
- Identify any information I would need to gather before finalizing the decision (e.g., benchmarks, library docs)
- Draft the outline of an ADR for this decision
Do not generate implementation code yet.