fix(observer): auto-scale max_turns by analysis batch size

[zanni098]

Fixes #2035.

Problem

observer-loop.sh hardcodes max_turns=20 as the default for the Claude analysis subprocess, while MAX_ANALYSIS_LINES defaults to 500. The two defaults are mismatched: Claude exhausts its turn budget before it can finish reading all observations and writing all qualifying instinct files.

Error: Reached max turns (20)

Fix

When ECC_OBSERVER_MAX_TURNS is not explicitly set, auto-scale max_turns proportionally to the actual analysis batch size (analysis_count, which is already computed just before this code):

max_turns = clamp(analysis_count / 10, floor=20, cap=100)

Results:

analysis_count	max_turns
< 200 lines	20 (existing floor, unchanged)
500 lines	50 ← resolves the reported failure
800 lines	80
≥ 1000 lines	100 (cap)

Setting ECC_OBSERVER_MAX_TURNS explicitly still overrides the auto-scaled value, preserving the existing escape hatch.

Why not just raise the hardcoded default to 50?

Auto-scaling is strictly better: it handles any value of ECC_OBSERVER_MAX_ANALYSIS_LINES correctly, whereas a new hardcoded default of 50 would still fail if a user cranks MAX_ANALYSIS_LINES beyond 500.

Files changed

• skills/continuous-learning-v2/agents/observer-loop.sh — replace 3-line max_turns assignment with an 8-line auto-scaling block

Testing

Manual verification of the logic (pure shell arithmetic, no external deps):

for lines in 10 50 199 200 500 800 1000 1500; do
  turns=$(( lines / 10 ))
  [ "$turns" -lt 20 ] && turns=20
  [ "$turns" -gt 100 ] && turns=100
  echo "$lines lines → $turns turns"
done
# 10 lines → 20 turns
# 50 lines → 20 turns
# 199 lines → 20 turns
# 200 lines → 20 turns
# 500 lines → 50 turns  ✓
# 800 lines → 80 turns
# 1000 lines → 100 turns
# 1500 lines → 100 turns (cap)

---

Summary by cubic

Auto-scales the observer’s max_turns based on analysis_count so Claude doesn’t hit the turn limit before processing the default 500-line batch and writing all instinct files.

• Bug Fixes
  • Compute max_turns = clamp(analysis_count / 10, 20, 100) when ECC_OBSERVER_MAX_TURNS is not set.
  • Preserve manual override via ECC_OBSERVER_MAX_TURNS.
  • Examples: 500 lines → 50 turns; 800 → 80; <200 → 20.

^{Written for commit 8dff671. Summary will update on new commits. Review in cubic}

Summary by CodeRabbit

• Improvements
  • Observer loop now automatically scales its iteration depth based on analysis size instead of using a fixed default, optimizing performance without requiring manual configuration adjustments.

[coderabbitai]

📝 Walkthrough

Walkthrough

The observer loop's max_turns default parameter is now calculated dynamically based on the input analysis_count when the ECC_OBSERVER_MAX_TURNS environment variable is not set. The scaling formula is 1 turn per 10 lines of analysis, with hard bounds enforcing a minimum of 20 and maximum of 100 turns.

Changes

Observer Max Turns Auto-scaling

Layer / File(s)	Summary
Observer max_turns auto-scaling logic skills/continuous-learning-v2/agents/observer-loop.sh	Default max_turns assignment replaces the fixed 20 value with a scaled calculation from analysis_count, dividing by 10 and enforcing minimum and maximum bounds.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 A scaling tale of turns and lines,
No fixed bounds need bind these designs,
Ten lines per turn, a ratio fair,
The observer leaps through code with care!

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: auto-scaling max_turns based on analysis batch size in the observer component.
Linked Issues check	✅ Passed	The PR implements auto-scaling of max_turns based on analysis_count as proposed in issue #2035, addressing the insufficient turn limit for the configured batch size.
Out of Scope Changes check	✅ Passed	All changes are directly related to fixing the observer's max_turns scaling issue; no out-of-scope modifications detected.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

skills/continuous-learning-v2/agents/observer-loop.sh (1)

221-229: 💤 Low value

Consider adding a comment explaining the validation block's purpose.

The validation block serves as a safety net for invalid ECC_OBSERVER_MAX_TURNS input (e.g., non-numeric or too-low values). While the auto-scaled path always produces valid values ≥20, this block is still necessary for the explicit-override case.

A brief comment would help future maintainers understand why this validation remains after the auto-scaling change.

📝 Suggested clarifying comment

   fi
   exit_code=0
 
+  # Validate max_turns in case ECC_OBSERVER_MAX_TURNS was set to invalid value
   case "$max_turns" in
     ''|*[!0-9]*)
       max_turns=20

🤖 Prompt for 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.

In `@skills/continuous-learning-v2/agents/observer-loop.sh` around lines 221 -
229, Add a brief explanatory comment above the validation block that guards the
max_turns value: explain that this checks the user-provided
ECC_OBSERVER_MAX_TURNS (stored in max_turns) for non-numeric or too-small values
and forces a safe default of 20, and note that although auto-scaling produces
values ≥20, the check is needed for explicit overrides; locate the block
referencing max_turns and the numeric check (case "$max_turns" in ... and the if
[ "$max_turns" -lt 4 ] test) and insert the comment immediately before it.

🤖 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 `@skills/continuous-learning-v2/agents/observer-loop.sh`:
- Around line 208-218: The test assertion expecting the literal string
max_turns="${ECC_OBSERVER_MAX_TURNS:-20}" must be updated to reflect the new
auto-scaling logic in observer-loop.sh: change tests/hooks/hooks.test.js to no
longer look for the removed ":-20" default and instead assert that
observer-loop.sh references the ECC_OBSERVER_MAX_TURNS variable and implements
the fallback autoscale block (checks for ECC_OBSERVER_MAX_TURNS being non-empty,
otherwise computes max_turns from analysis_count and clamps to 20–100). Ensure
the test verifies presence of the ECC_OBSERVER_MAX_TURNS reference and the
autoscale/clamp logic (or matches a regex covering those lines) rather than the
old fixed default literal.

---

Nitpick comments:
In `@skills/continuous-learning-v2/agents/observer-loop.sh`:
- Around line 221-229: Add a brief explanatory comment above the validation
block that guards the max_turns value: explain that this checks the
user-provided ECC_OBSERVER_MAX_TURNS (stored in max_turns) for non-numeric or
too-small values and forces a safe default of 20, and note that although
auto-scaling produces values ≥20, the check is needed for explicit overrides;
locate the block referencing max_turns and the numeric check (case "$max_turns"
in ... and the if [ "$max_turns" -lt 4 ] test) and insert the comment
immediately before it.

🪄 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: cf9e366d-6d09-4467-a085-4444cfffc05f

📥 Commits

Reviewing files that changed from the base of the PR and between 928076c and 8dff671.

📒 Files selected for processing (1)

• skills/continuous-learning-v2/agents/observer-loop.sh

[coderabbitai]

⚠️ Potential issue | 🔴 Critical | ⚡ Quick win

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Check if the test still expects the old pattern

echo "=== Checking test expectations ==="
rg -n 'max_turns="\$\{ECC_OBSERVER_MAX_TURNS:-20\}"' tests/hooks/hooks.test.js

echo ""
echo "=== Checking if test was recently modified ==="
git log -1 --oneline -- tests/hooks/hooks.test.js

Repository: affaan-m/ECC

Length of output: 354

---

Fix failing hook test assertion (old :-20 string removed)

tests/hooks/hooks.test.js still asserts that observer-loop.sh contains the exact string max_turns="${ECC_OBSERVER_MAX_TURNS:-20}", but the new logic auto-scales max_turns (clamped to 20–100) instead of using that fixed :-20 default—so CI will fail when the literal is no longer present.

🤖 Prompt for 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.

In `@skills/continuous-learning-v2/agents/observer-loop.sh` around lines 208 -
218, The test assertion expecting the literal string
max_turns="${ECC_OBSERVER_MAX_TURNS:-20}" must be updated to reflect the new
auto-scaling logic in observer-loop.sh: change tests/hooks/hooks.test.js to no
longer look for the removed ":-20" default and instead assert that
observer-loop.sh references the ECC_OBSERVER_MAX_TURNS variable and implements
the fallback autoscale block (checks for ECC_OBSERVER_MAX_TURNS being non-empty,
otherwise computes max_turns from analysis_count and clamps to 20–100). Ensure
the test verifies presence of the ECC_OBSERVER_MAX_TURNS reference and the
autoscale/clamp logic (or matches a regex covering those lines) rather than the
old fixed default literal.

[cubic-dev-ai]

No issues found across 1 file

_{Re-trigger cubic}