The llm-long-running-tasks branch is our main work spike for LLM agent infrastructure on Endo. It includes a llamadrome package (multi-backend LLM agent with Anthropic/Ollama), conversation persistence across daemon restarts, and a set of new daemon verbs for capability-separated evaluation: define, endow, form, requestEvaluation, and storeValue. The idea is that an LLM guest can propose code with named slots, the host endows those slots with capabilities, and the code runs in a sandbox with host approval.
I've looked at the daemon-reply and daemon-collector branches and want to merge that work into the LLM branch so we're building on the latest daemon internals rather than diverging further.
Bring in the formula GC (collection, deincarnation, reincarnation), reply support, and all the daemon refactoring as the new foundation.
- The entire
packages/llamadrome/package -- agent lifecycle, backends, system prompt, conversation store - New CLI commands:
define,endow,form,approve-eval,respond-form help-text.jsfor method-level documentation
- Durable messages / bigint parsing (already merged via PR #3074)
- All the WIP commits that originated from Kris's branches
After the merge, re-wire the define/endow/form/requestEvaluation/storeValue plumbing in mail.js, guest.js, host.js, daemon.js, types.d.ts, and interfaces.js to fit the refactored daemon internals.
I started adding threadId propagation through message formulas and a threads CLI command. The daemon-reply branch adds reply support. Are these the same concept or complementary? Happy to drop the thread work and build on the reply mechanism if that's the right foundation.
Is daemon-collector ready to merge into the LLM branch now, or are there pending changes that should land first?