feat(configs): connectome C1 foraging config (Stage 2 step 2a)

[chrisjz]

Summary

The connectome arm's C1 (foraging-only) config — Stage 2 step 2a — completing C1 for all 5 MUST arms. Authored connectomeppo_small_continuous2d_fick_adaptive_klinotaxis.yml: the Cook-2019 hermaphrodite connectome (gap junctions + strict chemical mask + klinotaxis; projection wiring unchanged from T4-grid — validate-don't-rebuild) on the Fick+adaptive continuous substrate with the shared C1 recipe.

Recipe (shared with the generic arms)

max_steps 800 + forage-to-survive satiety (300/0.8/20) + gamma 0.99 + lr 0.0003 + entropy 0.05 + max_turn_rad 0.5, on the metric-coherent reward. Substrate: Fick 8 / adaptive ε 0.1 / capture 2.0 / foods 5 / target 10.

Convergence (n=4, corrected reward)

100 / 100 / 100 / 100 per-seed success, eaten 10.0, holding across all quarters.

The connectome was the weakest arm at T4 and needed entropy 0.10 / lr 0.001 on the buggy reward — it needs no special tuning on the metric-coherent reward (the bug-era knobs were value-explosion compensation). So C1 is now complete for all 5 MUST arms: MLP ~99 / Transformer 100 / LSTM 100 / CfC ~91 / connectome 100.

Scope

• New connectome C1 config (config + eval only; brain/projections/substrate already exist).
• phase6-tracking Stage 2 step 2a recorded.
• Remaining Stage 2: 2b isolated projection validation (predator-only + thermo-only), 2c connectome C2 (foraging + predator).

Full pre-commit run -a green.

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Added a new continuous-action foraging scenario with adaptive chemical sensing on a static Fick-shaped gradient environment, using the C1 connectome setup and tuned PPO settings. Reported success is 100% across all validated runs.

• Bug Fixes

  • Improved the foraging reward metric consistency across steps to prevent earlier recurrent-arm collapse and ensure stable convergence.

• Documentation

  • Updated phase tracking to reflect completion of Stage 2 for the C1 step, including validation results and the finalized long-episode recipe.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 60dba950-5da8-4176-8899-2761e28cd4e4

📥 Commits

Reviewing files that changed from the base of the PR and between aca96de and cd2aa15.

📒 Files selected for processing (1)

• openspec/changes/phase6-tracking/tasks.md

✅ Files skipped from review due to trivial changes (1)

• openspec/changes/phase6-tracking/tasks.md

---

📝 Walkthrough

Walkthrough

Adds a new YAML scenario configuration for ConnectomePPO continuous-action klinotaxis foraging using the Cook-2019 hermaphrodite connectome with a Fick gradient and adaptive chemosensor. Updates the Phase 6 tracking document to mark Connectome C1 Stage 2 step 2a as complete with 100/100 per-seed success.

Changes

ConnectomePPO C1 Klinotaxis Config and Phase 6 Tracking

Layer / File(s)	Summary
ConnectomePPO C1 klinotaxis scenario config configs/scenarios/foraging/connectomeppo_small_continuous2d_fick_adaptive_klinotaxis.yml	Introduces a 97-line YAML scenario for ConnectomePPO C1 with Cook-2019 connectome topology, strict chemical masking, gap junctions, Gaussian action head, PPO hyperparameters (entropy_coef: 0.05, learning_rate: 0.0003), continuous_2d environment, adaptive fold-change chemosensor, and static Fick gradient geometry.
Phase 6 tracking: C1 completion and context openspec/changes/phase6-tracking/tasks.md	Marks T7.*.c1_foraging complete for all 5 PPO arms and documents the root cause of earlier recurrent-arm collapse: reward-metric inconsistency (Manhattan vs. Euclidean). Records the fix (metric-coherent potential term) and validation results showing all generic arms at ~10/10. Marks Connectome C1 Stage 2 step 2a complete with 100/100 per-seed success using the Fick+adaptive continuous config with shared C1 recipe hyperparameters.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• SyntheticBrains/nematode#242: Finalized the same "simplified C1" continuous-2D klinotaxis recipe parameters (entropy_coef: 0.05, learning_rate: 0.0003, constrained turning/satiety) that this PR adopts for ConnectomePPO.
• SyntheticBrains/nematode#227: Calibrated the continuous 2D gradient_field_mode: fick and adaptive chemosensor parameters that this PR's new config consumes.
• SyntheticBrains/nematode#230: Shares the same Manhattan-vs-Euclidean reward-distance/potential-metric correction that this PR documents as the root cause and fix for T7 continuous-substrate stability.

Poem

🐇 Hop, hop, through the gradient field I go,
The Fick-shaped signal tells me where the food will flow.
With epsilon and alpha tuned just right,
My connectome charts a klinotaxis flight.
100 seeds bloom, all munching well —
Stage 2a done! ✅ What a tale to tell!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'feat(configs): connectome C1 foraging config (Stage 2 step 2a)' directly and clearly describes the main change: adding a new connectome C1 foraging configuration file that completes a specific stage/step in the project roadmap.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch openspec/t7-connectome-c1-foraging

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@openspec/changes/phase6-tracking/tasks.md`:
- Around line 387-390: The T7.*.c1_foraging task checkbox is unchecked (- [ ])
but the resolution notes confirm C1 is complete for all 5 MUST arms (MLP,
Transformer, LSTM, CfC, and connectome). Change the checkbox from - [ ]
T7.*.c1_foraging to - [x] T7.*.c1_foraging to mark the task as complete and keep
the tracker state consistent with the documented completion status.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 7230df23-1069-4ff8-b9ca-e198af07d6ed

📥 Commits

Reviewing files that changed from the base of the PR and between 1343644 and aca96de.

📒 Files selected for processing (2)

• configs/scenarios/foraging/connectomeppo_small_continuous2d_fick_adaptive_klinotaxis.yml
• openspec/changes/phase6-tracking/tasks.md

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!