Add language guards for Python-only endpoints

[mohammedahmed18]

Issue #9: Python-Only Endpoints Called Without Language Guards

Problem

The process_review() method calls get_new_explanation() and get_optimization_review() without checking the language. These are Python-only endpoints that would fail or return incorrect results for JavaScript/TypeScript.

Root Cause

File: /opt/codeflash/codeflash/languages/function_optimizer.py

Line 2575: get_new_explanation() called without language check
Line 2638: get_optimization_review() called without language check

This is the same pattern as Issue #8 (adaptive_optimize) which was already fixed in PR #1995.

Impact

• Latent bug (not yet triggered - will manifest when JS/TS optimizations start succeeding)
• Severity: MEDIUM (not blocking current work, but will block future success)
• All current JS/TS optimizations fail at baseline (Docker/timeouts), so these endpoints are never reached
• Once baseline issues are resolved, JS/TS optimizations would fail when trying to generate explanations or reviews

Fix

Line 2575 - get_new_explanation:

# Before (broken):
new_explanation_raw_str = self.aiservice_client.get_new_explanation(...)

# After (fixed):
new_explanation_raw_str = None
if self.function_to_optimize.language == "python":
    new_explanation_raw_str = self.aiservice_client.get_new_explanation(...)

Line 2638 - get_optimization_review:

# Before (broken):
try:
    opt_review_result = self.aiservice_client.get_optimization_review(...)
except Exception as e:
    logger.debug(...)

# After (fixed):
if self.function_to_optimize.language == "python":
    try:
        opt_review_result = self.aiservice_client.get_optimization_review(...)
    except Exception as e:
        logger.debug(...)

For non-Python languages:

• get_new_explanation returns None → uses original explanation from AI service
• get_optimization_review returns empty review → PR proceeds without review rating

Testing

✅ Existing function_optimizer tests pass (5 tests)
✅ No linting/type errors (uv run prek)
✅ No regressions in optimization flow

Related Issues

• Issue end to end test with prod aiserver #8 (Fixed in PR Fix: Add language check before calling adaptive_optimize for JS/TS #1995): adaptive_optimize called without language guard
• Same pattern, same fix location (process_review + adaptive_optimize both in function_optimizer.py)

Category

Latent bug (will reproduce when JS/TS optimizations succeed)

Files Changed

• codeflash/languages/function_optimizer.py - Added language guards

[claude]

Claude finished @mohammedahmed18's task in 3m 22s —— View job

---

PR Review Summary

• Triage PR size — SMALL (1 file, +39/-30)
• Run lint/typecheck (prek) — ✅ Pass
• Resolve stale review threads — none found
• Review diff
• Duplicate detection
• Post summary

---

Prek Checks

✅ All ruff/format checks pass. No issues.

---

Code Review

Correctness: The guards are logically correct. get_new_explanation() and get_optimization_review() are Python-specific AI service endpoints and should not be called for JS/TS/Java. The FunctionToOptimize.language field stores str(Language.X) values (e.g. "javascript", "python"), so == "python" comparisons work correctly.

⚠️ Overlap with open PR #1995

The change at function_optimizer.py:1268 (adding and self.function_to_optimize.language == "python" to the adaptive_optimize guard) is identical to the change in PR #1995, which is currently open. Both PRs modify the same line to the same content. Whichever merges second will have a trivial merge conflict on that line. Coordinate with or supersede PR #1995 to avoid churn.

⚠️ Language guard pattern goes against migration direction

Per .claude/rules/language-patterns.md:

Prefer LanguageSupport protocol dispatch over is_python()/is_javascript() guards — remaining guards are being migrated to protocol methods.

These new == "python" string comparisons add three more sites that will eventually need to be migrated. For the two new guards (get_new_explanation, get_optimization_review), consider whether the AI service client itself should handle this (e.g., a supports_explanation_generation() protocol method), rather than adding more inline language checks. This isn't a blocker given the stated latent-bug urgency, but worth tracking.

⚠️ No tests for the new guards

CLAUDE.md: "Everything that can be tested should have tests." PR #1995 added 195 lines of tests for the adaptive_optimize guard (same pattern, same file). This PR adds no tests for the get_new_explanation and get_optimization_review guards. Tests should be added to cover the non-Python path returning None/empty-review correctly. Fix this →

---

Duplicate Detection

No duplicates detected. The guards follow the same pattern already used in this file.

---

Last updated: 2026-04-04