HKUDS/DeepTutor · error · GenerationFailure

generated code does not parse: {syntax_error}

Error message

generated code does not parse: {syntax_error}

What it means

The generated snippet was syntax-checked with `_syntax_error(code, language)` and failed to parse, so the block refuses to publish broken/truncated code and raises GenerationFailure carrying the parser's message, letting the compiler's retry path try again.

Source

Thrown at deeptutor/book/blocks/code.py:104

            system_prompt=get_book_prompt(prompts, "system"),
            max_tokens=900,
            temperature=0.3,
            language=ctx.language,
        )

        code = str(data.get("code") or "").strip()
        if not code:
            raise GenerationFailure("LLM did not return any code.")
        if "<think" in code.lower() or "</think" in code.lower():
            raise GenerationFailure("prompt leak detected in generated code.")

        code_language = str(data.get("language") or language).strip() or language
        syntax_error = _syntax_error(code, code_language)
        if syntax_error:
            # A truncated or malformed snippet is worse than none: the reader
            # copies it, it fails, and nothing said it was never checked. Fail
            # the block so the compiler's retry path gets a second attempt.
            raise GenerationFailure(f"generated code does not parse: {syntax_error}")

        metadata = data.get("_metadata") if isinstance(data.get("_metadata"), dict) else {}
        return (
            {
                "language": code_language,
                "code": code,
                "explanation": str(data.get("explanation") or "").strip(),
                "intent": intent,
            },
            [],
            {**metadata, "syntax_checked": _CHECKABLE.get(code_language.lower()) is not None},
        )


__all__ = ["CodeGenerator"]

View on GitHub (pinned to 3e82f13042)

Solutions

1. Retry the block (the compiler's retry path exists precisely for this)
2. Increase max_tokens so snippets aren't truncated mid-statement
3. Verify the `language` value in the response matches the actual snippet
4. Use a stronger model or lower temperature for long code blocks

from deeptutor.book.blocks.code import _syntax_error
if _syntax_error(snippet, language):
    regenerate_or_reject(snippet)

def parses_clean(code: str, language: str) -> bool:
    return _syntax_error(code, language) is None

try:
    result = await code_gen._generate(ctx)
except GenerationFailure as e:
    if "does not parse" in str(e):
        result = await code_gen._generate(ctx)
    else:
        raise

When it happens

Trigger: The LLM returns code with a syntax error in the declared language — truncation mid-statement, mismatched fences, wrong language tag (e.g. Python snippet labeled 'bash' so it's checked with the wrong parser), or genuine model mistakes.

Common situations: Long snippets cut off by max_tokens; language field defaulting incorrectly; models mixing syntax across languages; unavailable parser causing fallback strictness.

Related errors

• Model not configured for agent {self.agent_name}. Please act
• Render failed because local LaTeX is missing. Please avoid T
• Document not found
• Unsupported import source: {value!r}
• Invalid folder path

AI-assisted analysis of HKUDS/DeepTutor@3e82f13042 (2026-08-27). Data as JSON: /api/errors/e548cb3fda655254. Report an issue: GitHub.