Comment by beefsack
12 hours ago
I feel like so many of these memory solutions are incredibly over-engineered too.
You can work around a lot of the memory issues for large and complex tasks just by making the agent keep work logs. Critical context to keep throughout large pieces of work include decisions, conversations, investigations, plans and implementations - a normal developer should be tracking these and it's sensible to have the agent track them too in a way that survives compaction.
Yes. I have (as part of Claude output) a
- `FEATURE_IMPL_PLAN.md` (master plan; or `NEXT_FEATURES_LIST.md` or somesuch)
- `FEATURE_IMPL_PROMPT_TEMPLATE.md` (where I replace placeholders with next feature to be implemented; prompt includes various points about being thorough, making sure to validate and loop until full test pipeline works, to git version tag upon user confirmation, etc.)
- `feature-impl-plans/` directory where Claude is to keep per-feature detailed docs (with current status) up to date - this is esp. useful for complex features which may require multiple sessions for example
- also instruct it to keep main impl plan doc up to date, but that one is limited in size/depth/scope on purpose, not to overwhelm it
- CLAUDE.md has summary of important code references (paths / modules / classes etc.) for lookup, but is also restricted in size. But it includes full (up-to-date) inventory of all doc files, for itself
- If I end up expanding CLAUDE.md for some reason or temporarily (before I offload some content to separate docs), I will say as part of prompt template to "make sure to read in the whole @CLAUDE.md without skipping any content"
I agree. Plan files and I use git for my work logs. Have been successful.
Great advise. For large plans I tell the agent to write to an “implementation_log.md” and make note of it during compaction. Additionally the agent can also just reference the original session logs.
The problem with this approach, is that the model may forget to update the log... It usually happens when the context window >50% filled
I found this happen less often if the task is a part of the plan. It typically gets in a cycle habit of editing code and updating the doc
.. and not only those, but the baseline as well aka CLAUDE.md.. I've countless of times told it basics, in the same session without compacting etc etc
Yep. I just have my agents write out key details to a markdown file. Doesn’t have to be perfect. Just enough to reorient itself to a problem.