fixes one off errors and auto water

[micahwoodard]

resolves #207

• requires a run length strictly greater than min_consecutive_stable_trials

resolves #208

• instead of always picking left, lefts randomly select a side

resolves #203

• stops when max_trial <= len(history)

[Copilot]

🟡 Changes recommended

An unresolved critical autowater bug and missing exact-boundary regression tests must be addressed.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR fixes trial-count boundaries, stability thresholds, and equal-probability autowater selection.

Changes:

• Stops sessions at max_trial.
• Requires stability runs to exceed min_consecutive_stable_trials.
• Randomizes autowater side selection when probabilities tie.

File summaries

File	Summary	Findings
src/aind_behavior_dynamic_foraging/task_logic/trial_generators/uncoupled_trial_gnerator.py	Corrects maximum-trial termination.	Moderate (3 votes): Add exact-boundary regression tests for max_trial and max_trial - 1.
src/aind_behavior_dynamic_foraging/task_logic/trial_generators/coupled_trial_generators/coupled_trial_generator.py	Corrects trial termination and stability thresholds.	Moderate (3 votes): Add exact-boundary regression tests for max_trial and max_trial - 1.
src/aind_behavior_dynamic_foraging/task_logic/trial_generators/block_based_trial_generator.py	Randomizes equal-probability autowater selection.	Critical (3 votes): Convert the sampled numpy.bool_ to a Python bool before identity checks.

Review details

Suppressed comments (1)

src/aind_behavior_dynamic_foraging/task_logic/trial_generators/coupled_trial_generators/coupled_trial_generator.py:235

• This strict comparison invalidates existing stability tests: with the default min_stable=5 and kernel_size=2, test_update_block_switches_after_conditions_met supplies six choices and produces only five stable windows, so the block no longer switches; the anytime fixture can similarly fail when the high-reward side is left. Update those fixtures to provide min_stable + kernel_size stable choices and retain an exact-boundary negative case.

                if run_len > min_stable:

• Files reviewed: 3/3 changed files
• Comments generated: 3
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.