4623659974
**Issue #8: CLI incorrectly calls adaptive_optimize for JavaScript/TypeScript** **Problem:** When a refined candidate (source=REFINE) succeeds for JS/TS, the next iteration calls adaptive_optimize (Python-only endpoint) instead of optimize_code_refinement (all languages). This results in "422 - Invalid code generated" from the AI service because adaptive_optimize tries to parse JS/TS code using libcst (Python AST parser). **Root Cause:** File: codeflash/languages/function_optimizer.py (line 1266) The code checked if a REFINE candidate existed but did not check the language before calling adaptive_optimize. **Evidence:** - Trace ID: 1417a6da-796c-4a38-8c44-00401dbab6c7 - Function: formatBytes (TypeScript) - Error: "POST /ai/adaptive_optimize HTTP/1.1" 422 36 - AI service logs: "adaptive_optimize invalid code" **Fix:** Added language check at line 1266: ```python if is_candidate_refined_before and self.function_to_optimize.language == "python": # Call adaptive_optimize (Python-only) else: # Call optimize_code_refinement (all languages) ``` **Testing:** - Added 4 regression tests in test_adaptive_optimize_language_bug.py - All tests pass - No linting errors from `uv run prek` **Impact:** - Fixes systematic bug affecting JS/TS optimizations with successful REFINE candidates - Allows second refinement iteration to proceed for JS/TS - Python behavior unchanged (still uses adaptive_optimize after REFINE) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> |
||
|---|---|---|
| .claude | ||
| .codex | ||
| .gemini | ||
| .github | ||
| code_to_optimize | ||
| codeflash | ||
| codeflash-benchmark | ||
| codeflash-java-runtime | ||
| docs | ||
| experiments | ||
| packages/codeflash | ||
| tests | ||
| .gitignore | ||
| .mcp.json | ||
| .pre-commit-config.yaml | ||
| CLAUDE.md | ||
| codeflash.code-workspace | ||
| LICENSE | ||
| mypy_allowlist.txt | ||
| pyproject.toml | ||
| README.md | ||
| SECURITY.md | ||
| tessl.json | ||
| uv.lock | ||
Codeflash is a general purpose optimizer for Python that helps you improve the performance of your Python code while maintaining its correctness. It uses advanced LLMs to generate multiple optimization ideas for your code, tests them to be correct and benchmarks them for performance. It then creates merge-ready pull requests containing the best optimization found, which you can review and merge.
How to use Codeflash -
- Optimize an entire existing codebase by running
codeflash --all - Automate optimizing all future code you will write by installing Codeflash as a GitHub action.
- Optimize a Python workflow
python myscript.pyend-to-end by runningcodeflash optimize myscript.py
Codeflash is used by top engineering teams at Pydantic (PRs Merged), Roboflow (PRs Merged 1, PRs Merged 2), Unstructured (PRs Merged 1, PRs Merged 2), Langflow (PRs Merged) and many others to ship performant, expert level code.
Codeflash is great at optimizing AI Agents, Computer Vision algorithms, PyTorch code, numerical code, backend code or anything else you might write with Python.
Installation
To install Codeflash, run:
pip install codeflash
Add codeflash as a development time dependency if you are using package managers like uv or poetry.
Quick Start
-
To configure Codeflash for a project, at the root directory of your project where the pyproject.toml file is located, run:
codeflash init- It will ask you a few questions about your project like the location of your code and tests
- Ask you to generate an API Key to access Codeflash's LLMs
- Install a GitHub app to open Pull Requests on GitHub.
- Ask if you want to setup a GitHub actions which will optimize all your future code.
- The codeflash config is then saved in the pyproject.toml file.
-
Optimize your entire codebase:
codeflash --allThis can take a while to run for a large codebase, but it will keep opening PRs as it finds optimizations.
-
Optimize a script:
codeflash optimize myscript.py
Documentation
For detailed installation and usage instructions, visit our documentation at docs.codeflash.ai
Demo
- Optimizing the performance of new code for a Pull Request through GitHub Actions. This lets you ship code quickly while ensuring it remains performant.
https://github.com/user-attachments/assets/38f44f4e-be1c-4f84-8db9-63d5ee3e61e5
- Optiming a workflow end to end automatically with
codeflash optimize
https://github.com/user-attachments/assets/355ba295-eb5a-453a-8968-7fb35c70d16c
Support
Join our community for support and discussions. If you have any questions, feel free to reach out to us using one of the following methods:
License
Codeflash is licensed under the BSL-1.1 License. See the LICENSE file for details.