Engine (backend/engine)

LLM-facing agent logic: prompt assembly, tool definitions, and orchestration of model turns before actions are handed to backend/orchestration/ and executed in backend/execution/.

Layout

• orchestrator.py — engine-side coordination for model calls and tool intent in the session loop.
• tools/ — built-in tools (terminal, file edits, search, debugger, etc.) the model can invoke.
• prompts/ — system prompt partials and related prompt assets.

Risk limits, retries, stuck detection, and task validation are implemented in backend/orchestration/, not in this package.

References

• Architecture — end-to-end topology
• Developer guide — repository layout and workflows