dosubot[bot]
commented
1 year ago Answer generated by a 🤖
Answer
I understand that you're encountering a JSONDecodeError and an OutputParserException when using the Self Query Retriever with Google Flan T5 models in the LangChain library. The error seems to occur when the library attempts to parse JSON output from the language model.
The error message "Expecting value: line 1 column 1 (char 0)" typically indicates that the JSON decoder is unable to find any content to parse at the beginning of the string. This could be due to an empty string or a string that doesn't start with a valid JSON character (like '{' for an object or '[' for an array).
In your case, it appears that the output from the language model (LLM) is not in the expected JSON format. The JsonOutputFunctionsParser class in the LangChain library is trying to parse the output from the LLM as JSON, but it's failing because the output is not valid JSON.
The parse_result method in the JsonOutputFunctionsParser class is where the JSON parsing is happening. It's trying to load the function arguments as a JSON object using json.loads(func). If the JSON loading fails, it raises an OutputParserException.
Here's the relevant code from the JsonOutputFunctionsParser class:
class JsonOutputFunctionsParser(OutputFunctionsParser):
def parse_result(self, result: List[Generation]) -> Any:
func = super().parse_result(result)
if self.args_only:
return json.loads(func)
func["arguments"] = json.loads(func["arguments"])
return funcTo address this issue, you need to ensure that the output from the LLM is in the correct format. The output should be a string that includes a JSON object. If the output from the LLM is not in this format, the JsonOutputFunctionsParser will not be able to parse it, and you'll get the error you're seeing.
Please note that this is one potential solution based on the information provided. If this doesn't resolve your issue or if your experience contradicts this solution, I recommend checking if you're using the latest version of the LangChain library, as the issue might have been addressed in a more recent update.
If you believe this is a bug that could impact other users and you're comfortable doing so, you're encouraged to make a pull request with a proposed fix. Your contribution would be greatly appreciated by the LangChain community.
Sources
- langchain/agents/chat/output_parser.py
- langchain/agents/structured_chat/output_parser.py
- langchain/output_parsers/openai_functions.py
This response is meant to be useful, save you time, and share context. It is not meant to be a precise solution, but rather a starting point for your own research.
Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.
abdobeid
System Info
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/output_parsers/json.py:32 │ │ in parse_and_check_json_markdown │ │ │ │ 29 │ │ 30 def parse_and_check_json_markdown(text: str, expected_keys: List[str]) -> dict: │ │ 31 │ try: │ │ ❱ 32 │ │ json_obj = parse_json_markdown(text) │ │ 33 │ except json.JSONDecodeError as e: │ │ 34 │ │ raise OutputParserException(f"Got invalid JSON object. Error: {e}") │ │ 35 │ for key in expected_keys: │ │ │ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/output_parsers/json.py:25 │ │ in parse_json_markdown │ │ │ │ 22 │ json_str = json_str.strip() │ │ 23 │ │ │ 24 │ # Parse the JSON string into a Python dictionary │ │ ❱ 25 │ parsed = json.loads(json_str) │ │ 26 │ │ │ 27 │ return parsed │ │ 28 │ │ │ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/json/init.py:346 in loads │ │ │ │ 343 │ if (cls is None and object_hook is None and │ │ 344 │ │ │ parse_int is None and parse_float is None and │ │ 345 │ │ │ parse_constant is None and object_pairs_hook is None and not kw): │ │ ❱ 346 │ │ return _default_decoder.decode(s) │ │ 347 │ if cls is None: │ │ 348 │ │ cls = JSONDecoder │ │ 349 │ if object_hook is not None: │ │ │ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/json/decoder.py:337 in decode │ │ │ │ 334 │ │ containing a JSON document). │ │ 335 │ │ │ │ 336 │ │ """ │ │ ❱ 337 │ │ obj, end = self.raw_decode(s, idx=_w(s, 0).end()) │ │ 338 │ │ end = _w(s, end).end() │ │ 339 │ │ if end != len(s): │ │ 340 │ │ │ raise JSONDecodeError("Extra data", s, end) │ │ │ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/json/decoder.py:355 in raw_decode │ │ │ │ 352 │ │ try: │ │ 353 │ │ │ obj, end = self.scan_once(s, idx) │ │ 354 │ │ except StopIteration as err: │ │ ❱ 355 │ │ │ raise JSONDecodeError("Expecting value", s, err.value) from None │ │ 356 │ │ return obj, end │ │ 357 │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────╯ JSONDecodeError: Expecting value: line 1 column 1 (char 0)
During handling of the above exception, another exception occurred:
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/query_constructor/b │ │ ase.py:37 in parse │ │ │ │ 34 │ │ try: │ │ 35 │ │ │ expected_keys = ["query", "filter"] │ │ 36 │ │ │ allowed_keys = ["query", "filter", "limit"] │ │ ❱ 37 │ │ │ parsed = parse_and_check_json_markdown(text, expected_keys) │ │ 38 │ │ │ if len(parsed["query"]) == 0: │ │ 39 │ │ │ │ parsed["query"] = " " │ │ 40 │ │ │ if parsed["filter"] == "NO_FILTER" or not parsed["filter"]: │ │ │ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/output_parsers/json.py:34 │ │ in parse_and_check_json_markdown │ │ │ │ 31 │ try: │ │ 32 │ │ json_obj = parse_json_markdown(text) │ │ 33 │ except json.JSONDecodeError as e: │ │ ❱ 34 │ │ raise OutputParserException(f"Got invalid JSON object. Error: {e}") │ │ 35 │ for key in expected_keys: │ │ 36 │ │ if key not in json_obj: │ │ 37 │ │ │ raise OutputParserException( │ ╰──────────────────────────────────────────────────────────────────────────────────────────────────╯ OutputParserException: Got invalid JSON object. Error: Expecting value: line 1 column 1 (char 0)
During handling of the above exception, another exception occurred:
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/IPython/core/magics/execution.py:132 │ │ 5 in time │ │ │ │ 1322 │ │ else: │ │ 1323 │ │ │ st = clock2() │ │ 1324 │ │ │ try: │ │ ❱ 1325 │ │ │ │ exec(code, glob, local_ns) │ │ 1326 │ │ │ │ out=None │ │ 1327 │ │ │ │ # multi-line %%time case │ │ 1328 │ │ │ │ if expr_val is not None: │ │ in:1 │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/base.py:149 in │
│ call │
│ │
│ 146 │ │ │ ) │
│ 147 │ │ except (KeyboardInterrupt, Exception) as e: │
│ 148 │ │ │ run_manager.on_chain_error(e) │
│ ❱ 149 │ │ │ raise e │
│ 150 │ │ run_manager.on_chain_end(outputs) │
│ 151 │ │ final_outputs: Dict[str, Any] = self.prep_outputs( │
│ 152 │ │ │ inputs, outputs, return_only_outputs │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/base.py:143 in │
│ call │
│ │
│ 140 │ │ ) │
│ 141 │ │ try: │
│ 142 │ │ │ outputs = ( │
│ ❱ 143 │ │ │ │ self._call(inputs, run_manager=run_manager) │
│ 144 │ │ │ │ if new_arg_supported │
│ 145 │ │ │ │ else self._call(inputs) │
│ 146 │ │ │ ) │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/conversational_retr │
│ ieval/base.py:110 in _call │
│ │
│ 107 │ │ │ ) │
│ 108 │ │ else: │
│ 109 │ │ │ new_question = question │
│ ❱ 110 │ │ docs = self._get_docs(new_question, inputs) │
│ 111 │ │ new_inputs = inputs.copy() │
│ 112 │ │ new_inputs["question"] = new_question │
│ 113 │ │ new_inputs["chat_history"] = chat_history_str │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/conversational_retr │
│ ieval/base.py:191 in _get_docs │
│ │
│ 188 │ │ return docs[:num_docs] │
│ 189 │ │
│ 190 │ def _get_docs(self, question: str, inputs: Dict[str, Any]) -> List[Document]: │
│ ❱ 191 │ │ docs = self.retriever.get_relevant_documents(question) │
│ 192 │ │ return self._reduce_tokens_below_limit(docs) │
│ 193 │ │
│ 194 │ async def _aget_docs(self, question: str, inputs: Dict[str, Any]) -> List[Document]: │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/retrievers/self_query/base │
│ .py:96 in get_relevant_documents │
│ │
│ 93 │ │ inputs = self.llm_chain.prep_inputs({"query": query}) │
│ 94 │ │ structured_query = cast( │
│ 95 │ │ │ StructuredQuery, │
│ ❱ 96 │ │ │ self.llm_chain.predict_and_parse(callbacks=callbacks, inputs), │
│ 97 │ │ ) │
│ 98 │ │ if self.verbose: │
│ 99 │ │ │ print(structured_query) │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/llm.py:281 in │
│ predict_and_parse │
│ │
│ 278 │ │ ) │
│ 279 │ │ result = self.predict(callbacks=callbacks, kwargs) │
│ 280 │ │ if self.prompt.output_parser is not None: │
│ ❱ 281 │ │ │ return self.prompt.output_parser.parse(result) │
│ 282 │ │ else: │
│ 283 │ │ │ return result │
│ 284 │
│ │
│ /home/cloud/anaconda3/envs/mir/lib/python3.10/site-packages/langchain/chains/query_constructor/b │
│ ase.py:50 in parse │
│ │
│ 47 │ │ │ │ **{k: v for k, v in parsed.items() if k in allowed_keys} │
│ 48 │ │ │ ) │
│ 49 │ │ except Exception as e: │
│ ❱ 50 │ │ │ raise OutputParserException( │
│ 51 │ │ │ │ f"Parsing text\n{text}\n raised following error:\n{e}" │
│ 52 │ │ │ ) │
│ 53 │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
OutputParserException: Parsing text
json "query": "patient medical notes", "ID": "11542052" "old": ""
raised following error:
Got invalid JSON object. Error: Expecting value: line 1 column 1 (char 0)
Who can help?
@hwchase17 @ag
Information
Related Components
Reproduction
model_id = 'google/flan-t5-xl' tokenizer = AutoTokenizer.from_pretrained(model_id) pipe = pipeline( model=model_id, tokenizer=tokenizer, max_length=2048, temperature=0.1, top_p=0.95, repetition_penalty=1. )
llm = HuggingFacePipeline(pipeline=pipe) document_content_description = "Patient medical notes" metadata_field_info = [ AttributeInfo( name="ID", description="The unique identifier 'ID' of the patient", type="string", ), AttributeInfo( name="source", description="source of the document", type="string", ), ] retriever = SelfQueryRetriever.from_llm( llm, db, document_content_description, metadata_field_info, verbose = True ) memory=ConversationBufferMemory(memory_key="chat_history",output_key='answer') chain = ConversationalRetrievalChain.from_llm( llm = llm, retriever=retriever, memory=memory, get_chat_history=lambda h :h)
Expected behavior
Expected behavior should be something as attached