run-llama/llama_index · error · ValueError

Failed to convert output to JSON: {output!r}

Error message

Failed to convert output to JSON: {output!r}

What it means

After successfully parsing the LLM string as JSON/YAML, SelectionOutputParser normalizes a dict to a one-element list; if the parsed object is anything else (a bare string, number, or boolean), it raises ValueError('Failed to convert output to JSON: {output!r}'). The selection schema requires an array (or single object) of choice entries.

Source

Thrown at llama-index-core/llama_index/core/output_parsers/selection.py:97

                # NOTE: parsing again with pyyaml
                #       pyyaml is less strict, and allows for trailing commas
                #       right now we rely on this since guidance program generates
                #       trailing commas
                json_obj = yaml.safe_load(json_string)
            except yaml.YAMLError as e_yaml:
                raise OutputParserException(
                    f"Got invalid JSON object. Error: {e_json} {e_yaml}. "
                    f"Got JSON string: {json_string}"
                )
            except NameError as exc:
                raise ImportError("Please pip install PyYAML.") from exc

        if isinstance(json_obj, dict):
            json_obj = [json_obj]

        if not isinstance(json_obj, list):
            raise ValueError(f"Failed to convert output to JSON: {output!r}")

        json_output = self._format_output(json_obj)
        answers = [Answer.from_dict(json_dict) for json_dict in json_output]
        return StructuredOutput(raw_output=output, parsed_output=answers)

    def format(self, prompt_template: str) -> str:
        return prompt_template + "\n\n" + _escape_curly_braces(FORMAT_STR)

View on GitHub (pinned to afd0fef371)

Solutions

1. Restore/keep the original selection prompt format string so the model emits the documented JSON array
2. Pre-validate parsed output yourself if wrapping the parser: accept only dict/list, re-prompt otherwise
3. Use a model with reliable instruction following for selection/routing
4. Catch the ValueError (and OutputParserException) around the query call and retry with an explicit schema reminder

Example fix

# before
# model output: 'yes'  -> yaml parses to bool True -> ValueError: Failed to convert output to JSON
response = selector.select(options)

# after
from llama_index.core.output_parsers import OutputParserException
try:
    response = selector.select(options)
except (OutputParserException, ValueError):
    response = selector.select(options, prompt_extra="Respond ONLY as a JSON array per the schema.")

import json
obj = json.loads(llm_output)  # or yaml.safe_load
if not isinstance(obj, (dict, list)):
    raise ValueError(f"selection output is a scalar ({obj!r}); re-prompt for JSON array")

def is_selection_shape(obj) -> bool:
    return isinstance(obj, (dict, list)) and (not isinstance(obj, list) or all(isinstance(i, dict) for i in obj))

try:
    result = selector.select(prompt)
except ValueError as e:
    if "Failed to convert output to JSON" in str(e):
        result = selector.select(prompt, prompt_extra="Return ONLY the JSON array per the schema.")
    else:
        raise

When it happens

Trigger: Model returns a plain string like '"Paris"' or a number (e.g. via yaml fallback parsing 'yes'/'42'), or a YAML scalar ('answer' parses as the string 'answer') instead of the expected [{answer: ..., score: ...}] structure; also markdown-fenced content where the outer parse yields a scalar.

Common situations: Prompts where the model replies with just the choice text instead of the JSON structure; selection templates modified/customized so the FORMAT_STR contract is broken; very small models echoing the query.

Related errors

• Got invalid JSON object. Error: {e_json} {e_yaml}. Got JSON
• Failed to parse pydantic object from guidance program. Proba
• Got empty streaming response
• Expected ActionReasoningStep, got {reasoning_step}
• LLM only supports text inputs

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/766980eaf6f53562. Report an issue: GitHub.