Skip to content

Remove the submit tools and the solve and learn prompts - #191

Merged
yichao-liang merged 1 commit into
masterfrom
empiric-submit-tools-cleanup
Sep 28, 2026
Merged

yichao-liang merged 1 commit into
masterfrom
empiric-submit-tools-cleanup

Conversation

@yichao-liang

Copy link
Copy Markdown
Collaborator

Summary

Stacked on #190 (base empiric-phased-cleanup); review and merge after it.

The phased solve sessions delivered plans through submit_plan and submit_policy.
Their capture gate re-ran each plan before accepting it: repeat rollouts, physics and rule-parameter margins, a necessity check and a robot-clearance probe.
The continual arms never had these tools; they rehearse plans with the sim probe inside run_python.
This removes the tools and what only they used: 79 files, about 8.7k lines.

Deleted:

  • tools/testing.py, tools/capture.py, tools/tasks.py and tools/clearance.py;
  • sketch_prompts.py, learn_prompts.py, the solve_* and learn_* templates, and their goldens and tests;
  • the attempt deadline and the capture fields on ToolContext;
  • the standalone arm's particle override scope;
  • the journal's strategy document and test-phase rollback helpers;
  • the skills' contact_objects hook, which only the clearance probe read;
  • 14 settings flags without a reader, including agent_plan_validation_rule_param_margin and agent_plan_validation_physics_margin, which the no-uncertainty and real-to-sim arms only checked were off.

Moved: _policy_source_path into tools/verdicts.py, and render_physical_params_section into play_prompts.py.

No benchmark arm changes behavior.
The rule-parameter ensemble used to be built when info-seeking or the margin gate was on; its only remaining reader is info-seeking, and every benchmark arm sets the two flags alike.
The empiric configs stop setting the three removed flags they named.

Left for a separate change: tool descriptions and probe docstrings that the agent can read still mention submit_plan (tools/exploration.py, tools/synthesis.py, notices and docstrings in belief_probe.py, the evaluate_trajectory helper).
Editing them changes the agent's prompts, so that should be its own decision.

Test plan

  • yapf, isort 5.10.1, docformatter, mypy and pylint on this commit (sbatch replay of CI's static jobs).
  • The 8 CI shards in the CI container.

🤖 Generated with Claude Code

The phased solve sessions delivered plans through submit_plan and
submit_policy, whose capture gate re-ran each plan (repeat rollouts,
physics and rule-parameter margins, a necessity check, a robot
clearance probe) before accepting it. The continual arms never had
these tools. This removes them together with:

- the solve and learn prompt templates, their goldens and tests
- the attempt deadline and the capture fields on the tool context
- the standalone arm's particle override scope
- the journal's strategy document and test-phase rollback helpers
- the skills' contact_objects hook, which only the clearance probe read
- 14 settings flags that lost their last reader

The empiric configs stop setting the three removed flags they named;
none of them changed behavior. Tool descriptions and probe docstrings
the agent can read still mention submit_plan and are left for a
separate change, since editing them changes the agent's prompts.
@yichao-liang
yichao-liang changed the base branch from empiric-phased-cleanup to master September 28, 2026 06:00
@yichao-liang
yichao-liang force-pushed the empiric-submit-tools-cleanup branch from a3b4c01 to d7c9bac Compare September 28, 2026 06:00
@yichao-liang
yichao-liang merged commit 1b02e6f into master Sep 28, 2026
14 checks passed
@yichao-liang
yichao-liang deleted the empiric-submit-tools-cleanup branch September 28, 2026 09:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant