alibaba/spring-ai-alibaba · error · IllegalArgumentException

Failed to parse result

Error message

Failed to parse result

What it means

TemplateTransformer.extractResultStrFromResponse applies a regex to the model response and throws IllegalArgumentException('Failed to parse result') when no match is found. It expects the LLM output to contain a specific template marker/structure (e.g. fenced or tagged result section) and the response deviated from it.

Source

Thrown at spring-boot-starters/spring-ai-alibaba-starter-builtin-nodes/src/main/java/com/alibaba/cloud/ai/graph/node/code/TemplateTransformer.java:65

	public Map<String, Object> transformResponse(String response) throws Exception {
		String resultStr = extractResultStrFromResponse(response);
		ObjectMapper mapper = new ObjectMapper();
		return mapper.readValue(resultStr,
				mapper.getTypeFactory().constructMapType(Map.class, String.class, Object.class));
	}

	public abstract String getRunnerScript(CodeStyle style);

	private String extractResultStrFromResponse(String response) {
		Pattern pattern = Pattern.compile(RESULT_TAG + "(.*?)" + RESULT_TAG, Pattern.DOTALL);
		Matcher matcher = pattern.matcher(response);

		if (matcher.find()) {
			return matcher.group(1).trim();
		}
		else {
			throw new IllegalArgumentException("Failed to parse result");
		}
	}

	private String serializeInputs(Map<String, Object> inputs) throws Exception {
		ObjectMapper mapper = new ObjectMapper();
		String inputsJsonStr = mapper.writeValueAsString(inputs);
		return Base64.getEncoder().encodeToString(inputsJsonStr.getBytes(StandardCharsets.UTF_8));
	}

	private String assembleRunnerScript(String code, Map<String, Object> inputs, CodeStyle style) throws Exception {
		String script = getRunnerScript(style);
		script = script.replace(CODE_PLACEHOLDER, code);
		script = script.replace(INPUTS_PLACEHOLDER, serializeInputs(inputs));
		return script;
	}

	private String getPreloadScript() {
		return "";

View on GitHub (pinned to f82da0b50f)

Solutions

1. Log the raw response to see how it deviates from the expected template.
2. Strengthen the prompt to require the exact output format with delimiters.
3. Lower temperature or increase max_tokens so the response is deterministic and complete.
4. Catch IllegalArgumentException and add a fallback parse (e.g. treat whole response as result).

Example fix

// before
String result = transformer.extractResultStrFromResponse(response); // IllegalArgumentException
// after
try {
    result = transformer.extractResultStrFromResponse(response);
} catch (IllegalArgumentException e) {
    log.warn("Template parse failed, using raw response");
    result = response.trim();
}

Pattern p = Pattern.compile(templateRegex);
boolean parseable = p.matcher(response).find();
if (!parseable) log.warn("Response missing expected template marker, length={}", response.length());

boolean hasTemplateMarker(String response, Pattern pattern) {
    return response != null && pattern.matcher(response).find();
}

try {
    return transformer.extractResultStrFromResponse(response);
} catch (IllegalArgumentException e) {
    log.warn("Result template parse failed; raw response: {}", response);
    return response.trim(); // fallback
}

When it happens

Trigger: Calling extractResultStrFromResponse with a response that lacks the expected template pattern — model returned plain text, refused, output in a different format, or truncated the marker.

Common situations: Prompt drift causing the model to skip the required delimiters; model answering in a different language/style; temperature too high producing malformed output; response truncated by max_tokens cutting off the closing marker.

Understand the failure class

• Parsing and encoding errors: unexpected token, malformed input — why parsers reject input and how to find the real culprit.

Related errors

• ChatClient successfully returned, but the returned json is i
• INVALID_PARAMS
• Invalid data format
• regex cannot be empty
• Invalid memory parameter

AI-assisted analysis of alibaba/spring-ai-alibaba@f82da0b50f (2026-09-09). Data as JSON: /api/errors/d1e0bec430c438f3. Report an issue: GitHub.