researchforge-autorun

Drive the research loop

researchforge autorun runs this loop by itself when an API key is configured. Without one, you are the intelligence and the engine is still the lab: it chooses where to search, validates what you write, runs the benchmark, and records the result. Your job is the one part it cannot do without a provider — writing the plan and the patches.

Never guess at the next move. The engine's search knows things a results table does not show: which ideas were already tried at which node, what is in each node's ancestry, and how much budget is left.

1. Ask where to go next

researchforge autorun --dry-run --json

This spends nothing — no AI call, no experiment runs. It returns the node the loop would expand, the hypotheses it would try there in ranked order, and the exact command to take that step:

• node — the experiment to build on, or null for the baseline;
• hypotheses — ranked, best candidate first;
• command — run this next;
• retreat: true — the only moves left are on branches that gained nothing. Tell the user: this measures an idea without the gains already banked, which is an ablation rather than progress;
• needs_resynthesis: true — every hypothesis has been tried everywhere it can apply. Go to step 5.

2. Plan at that node

Run the command from step 1. With a parent it looks like:

researchforge experiment plan hyp-002 --parent exp-008 --json

The exported context's repository section shows the files as exp-008 leaves them, and applied lists the experiments already baked into what you are reading. Write the change against those contents: do not re-apply the parent's change, and do not expect the baseline's values.

Then follow the researchforge-plan skill to write plan.yaml and the patches. Set parent: exp-008 on every entry — on this path nothing sets it for you, and a patch written against the parent but imported without it will be applied to the baseline and fail.