Update blogs/graphreader /graphreader_langgraph.ipynb (possibly to NOT use ollama_functions.with_structured_output?) #41
Description
Trying to run the graph reader agent notebook on my Windows 10, Python 3.11, latest version of Ollama and LangChain/Graph, I get the below traceback (I added in some extra printouts to help see the values of variables and what functions are being called when in order to help):
langgraph.invoke({"question":"Did Joan of Arc lose any battles?"})
--------------------
Step: rational_plan
Rational plan: To answer this question, we first need to find information about Joan of Arc's military campaigns and her battles against the English. We will look for details about specific battles where she was defeated or lost control, such as the Siege of Orléans, and compare them to any victories she had during that time.
message: content='' additional_kwargs={} response_metadata={} id='run-71658f84-2a46-456f-9ebe-51c4838994ca-0' tool_calls=[{'name': 'InitialNodes', 'args': {'initial_nodes': [{'key_element': 'Siege of Orléans', 'score': 80}, {'key_element': 'Joan in her only major film role', 'score': 60}, {'key_element': 'capture by Burgundians', 'score': 70}, {'key_element': 'Loire Campaign', 'score': 50}, {'key_element': 'Burgundian troops', 'score': 40}, {'key_element': 'visions from Michael, Margaret, Catherine', 'score': 30}, {'key_element': 'lead the French in battle', 'score': 90}, {'key_element': 'success in routing the English', 'score': 80}, {'key_element': "attempted to restore Dreyer's final cut", 'score': 60}, {'key_element': "Joan's trial", 'score': 70}]}, 'id': 'call_60a160c65f8b46df907ef071ee7acd6d', 'type': 'tool_call'}]
kwargs: {}
tool_calls: [{'name': 'InitialNodes', 'args': {'initial_nodes': [{'key_element': 'Siege of Orléans', 'score': 80}, {'key_element': 'Joan in her only major film role', 'score': 60}, {'key_element': 'capture by Burgundians', 'score': 70}, {'key_element': 'Loire Campaign', 'score': 50}, {'key_element': 'Burgundian troops', 'score': 40}, {'key_element': 'visions from Michael, Margaret, Catherine', 'score': 30}, {'key_element': 'lead the French in battle', 'score': 90}, {'key_element': 'success in routing the English', 'score': 80}, {'key_element': "attempted to restore Dreyer's final cut", 'score': 60}, {'key_element': "Joan's trial", 'score': 70}]}, 'id': 'call_60a160c65f8b46df907ef071ee7acd6d', 'type': 'tool_call'}]
--------------------
Step: atomic_fact_check
Reading atomic facts about: ['lead the French in battle', 'Siege of Orléans', 'success in routing the English', 'capture by Burgundians', "Joan's trial"]
message: content='' additional_kwargs={} response_metadata={} id='run-32a90a8a-772a-4336-a6b9-61a1c6357622-0' tool_calls=[{'name': 'AtomicFactOutput', 'args': {'updated_notebook': '', 'rational_next_action': "read_chunk(List['82f69cf57d252cb552a2076a7217b3a5'])", 'chosen_action': ''}, 'id': 'call_b37fa27ffbac4799980188360fe3c722', 'type': 'tool_call'}]
kwargs: {}
tool_calls: [{'name': 'AtomicFactOutput', 'args': {'updated_notebook': '', 'rational_next_action': "read_chunk(List['82f69cf57d252cb552a2076a7217b3a5'])", 'chosen_action': ''}, 'id': 'call_b37fa27ffbac4799980188360fe3c722', 'type': 'tool_call'}]
Rational for next action after atomic check: read_chunk(List['82f69cf57d252cb552a2076a7217b3a5'])
parse_function method started...
input_str:
pattern: (\w+)(?:\((.*)\))?
match: None
No match found...
Ending parse_function...
to_return: None
Chosen action: None
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
Cell In[32], line 1
----> 1 langgraph.invoke({"question":"Did Joan of Arc lose any battles?"})
File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\langgraph\pregel\__init__.py:1545, in Pregel.invoke(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, **kwargs)
1543 else:
1544 chunks = []
-> 1545 for chunk in self.stream(
1546 input,
1547 config,
1548 stream_mode=stream_mode,
1549 output_keys=output_keys,
1550 interrupt_before=interrupt_before,
1551 interrupt_after=interrupt_after,
1552 debug=debug,
1553 **kwargs,
1554 ):
1555 if stream_mode == "values":
1556 latest = chunk
File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\langgraph\pregel\__init__.py:1278, in Pregel.stream(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, subgraphs)
1267 # Similarly to Bulk Synchronous Parallel / Pregel model
1268 # computation proceeds in steps, while there are channel updates
1269 # channel updates from step N are only visible in step N+1
1270 # channels are guaranteed to be immutable for the duration of the step,
1271 # with channel updates applied only at the transition between steps
1272 while loop.tick(
1273 input_keys=self.input_channels,
1274 interrupt_before=interrupt_before_,
1275 interrupt_after=interrupt_after_,
1276 manager=run_manager,
1277 ):
-> 1278 for _ in runner.tick(
1279 loop.tasks.values(),
1280 timeout=self.step_timeout,
1281 retry_policy=self.retry_policy,
1282 get_waiter=get_waiter,
1283 ):
1284 # emit output
1285 yield from output()
1286 # emit output
File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\langgraph\pregel\runner.py:52, in PregelRunner.tick(self, tasks, reraise, timeout, retry_policy, get_waiter)
50 t = tasks[0]
51 try:
---> 52 run_with_retry(t, retry_policy)
53 self.commit(t, None)
54 except Exception as exc:
File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\langgraph\pregel\retry.py:29, in run_with_retry(task, retry_policy)
27 task.writes.clear()
28 # run the task
---> 29 task.proc.invoke(task.input, config)
30 # if successful, end
31 break
File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\langgraph\utils\runnable.py:385, in RunnableSeq.invoke(self, input, config, **kwargs)
383 context.run(_set_config_context, config)
384 if i == 0:
--> 385 input = context.run(step.invoke, input, config, **kwargs)
386 else:
387 input = context.run(step.invoke, input, config)
File ~\AppData\Local\Programs\Python\Python311\Lib\site-packages\langgraph\utils\runnable.py:167, in RunnableCallable.invoke(self, input, config, **kwargs)
165 else:
166 context.run(_set_config_context, config)
--> 167 ret = context.run(self.func, input, **kwargs)
168 if isinstance(ret, Runnable) and self.recurse:
169 return ret.invoke(input, config)
Cell In[20], line 45, in atomic_fact_check(state)
41 chosen_action = parse_function(atomic_facts_results.chosen_action)
42 print(f"Chosen action: {chosen_action}")
43 response = {
44 "notebook": notebook,
---> 45 "chosen_action": chosen_action.get("function_name"),
46 "check_atomic_facts_queue": [],
47 "previous_actions": [
48 f"atomic_fact_check({state.get('check_atomic_facts_queue')})"
49 ],
50 }
51 if chosen_action.get("function_name") == "stop_and_read_neighbor":
52 neighbors = get_neighbors_by_key_element(
53 state.get("check_atomic_facts_queue")
54 )
AttributeError: 'NoneType' object has no attribute 'get'
It appears that the "next action" is not being set properly. I noticed that the notebook makes use of the with_structured_output method which has been deprecated in LangChain v0.2, however the:
from langchain_community.chat_models import ChatOllama
# OR
from langchain_ollama import ChatOllama
from those latest versions don't offer the with_structured_output anymore. So the code would need to be updated to mitigate using this altogether, but just something I'm running into trying to run the code right now.