feat (2/n): Agent lib using resume_turn#158
Merged
yanxi0830 merged 12 commits intoagents_resumefrom Feb 21, 2025
Merged
Conversation
yanxi0830
requested review from
SLR722,
ashwinb,
dineshyv,
dltn,
ehhuang,
hardikjshah,
raghotham and
vladimirivic
as code owners
yanxi0830
force-pushed
the
agents_resume
branch
from
835a964 to
250676f
Compare
ehhuang
reviewed
Feb 21, 2025
ehhuang
reviewed
Feb 21, 2025
yanxi0830
added a commit
to llamastack/llama-stack
that referenced
this pull request
Feb 21, 2025
# What does this PR do? - #1178 - #1187 - #1194 **client changes** - llamastack/llama-stack-client-python#157 - llamastack/llama-stack-client-python#158 ## Test Plan ``` LLAMA_STACK_BASE_URL=http://localhost:8321 pytest -v tests/client-sdk/agents/test_agents.py --inference-model meta-llama/Llama-3.1-8B-Instruct ``` ``` LLAMA_STACK_CONFIG=fireworks pytest -v tests/client-sdk/agents/test_agents.py --inference-model meta-llama/Llama-3.1-8B-Instruct ``` <img width="1080" alt="image" src="https://github.com/user-attachments/assets/6c48e062-5d00-4fed-98de-ecca627e5195" /> **llama-stack-apps** ``` python -m examples.agents.react_agent localhost 8321 ``` - Test with script: https://gist.github.com/yanxi0830/f2e407527f468998a700cd29fd271b15 **Output Before**: we have 2 `turn_id` with 2 turns - https://gist.github.com/yanxi0830/9fbd7a80fcddc784a28c59d4a9c1d943 **Output After**: we have 1 `turn_id`, the final turn have all 3 steps - https://gist.github.com/yanxi0830/17754d56d08ccbeaec419b693137500c <img width="622" alt="image" src="https://github.com/user-attachments/assets/ed7f42ca-41c8-4ee9-b6fa-2299901cafb4" /> **Telemetry** <img width="1389" alt="image" src="https://github.com/user-attachments/assets/031bf6ab-959c-46a7-aa85-70a511cba296" /> [//]: # (## Documentation)
ashwinb
approved these changes
Feb 21, 2025
ehhuang
reviewed
Feb 21, 2025
Comment on lines
+181
to
+182
| if n_iter == 0: | ||
| yield chunk |
Contributor
There was a problem hiding this comment.
why yield when n_iter == 0?
Contributor
Author
There was a problem hiding this comment.
b/c when n_iter == 0, we will get back the actual Turn with turn_awaiting_input. e.g. https://gist.github.com/yanxi0830/17754d56d08ccbeaec419b693137500c
AgentTurnResponseStreamChunk(
│ event=TurnResponseEvent(
│ │ payload=AgentTurnResponseTurnAwaitingInputPayload(
│ │ │ event_type='turn_awaiting_input',
│ │ │ turn=Turn(
│ │ │ │ input_messages=[
│ │ │ │ │ UserMessage(
│ │ │ │ │ │ content='What is the boiling point of polyjuice in Celcius?',
│ │ │ │ │ │ role='user',
│ │ │ │ │ │ context=None
│ │ │ │ │ )
│ │ │ │ ],
│ │ │ │ output_message=CompletionMessage(
│ │ │ │ │ content='',
│ │ │ │ │ role='assistant',
│ │ │ │ │ stop_reason='end_of_turn',
│ │ │ │ │ tool_calls=[
│ │ │ │ │ │ ToolCall(
│ │ │ │ │ │ │ arguments={'liquid_name': 'polyjuice', 'celcius': 'True'},
│ │ │ │ │ │ │ call_id='782f2ba1-976d-45a5-874d-791c79ececf6',
│ │ │ │ │ │ │ tool_name='get_boiling_point'
│ │ │ │ │ │ )
│ │ │ │ │ ]
│ │ │ │ ),
│ │ │ │ session_id='8221e2f3-c4d3-44e7-ab7e-e74fdecb10e3',
│ │ │ │ started_at=datetime.datetime(2025, 2, 20, 21, 38, 8, 13477),
│ │ │ │ steps=[
│ │ │ │ │ InferenceStep(
│ │ │ │ │ │ api_model_response=CompletionMessage(
│ │ │ │ │ │ │ content='',
│ │ │ │ │ │ │ role='assistant',
│ │ │ │ │ │ │ stop_reason='end_of_turn',
│ │ │ │ │ │ │ tool_calls=[
│ │ │ │ │ │ │ │ ToolCall(
│ │ │ │ │ │ │ │ │ arguments={'liquid_name': 'polyjuice', 'celcius': 'True'},
│ │ │ │ │ │ │ │ │ call_id='782f2ba1-976d-45a5-874d-791c79ececf6',
│ │ │ │ │ │ │ │ │ tool_name='get_boiling_point'
│ │ │ │ │ │ │ │ )
│ │ │ │ │ │ │ ]
│ │ │ │ │ │ ),
│ │ │ │ │ │ step_id='c8a34ebd-2902-4da8-a78e-a549009880bb',
│ │ │ │ │ │ step_type='inference',
│ │ │ │ │ │ turn_id='38c0cc58-5c3b-4d8e-adf6-cb1c1a7bf0ee',
│ │ │ │ │ │ completed_at=datetime.datetime(2025, 2, 20, 21, 38, 8, 427067),
│ │ │ │ │ │ started_at=datetime.datetime(2025, 2, 20, 21, 38, 8, 24697)
│ │ │ │ │ )
│ │ │ │ ],
│ │ │ │ turn_id='38c0cc58-5c3b-4d8e-adf6-cb1c1a7bf0ee',
│ │ │ │ completed_at=datetime.datetime(2025, 2, 20, 21, 38, 8, 461543),
│ │ │ │ output_attachments=[]
│ │ │ )
│ │ )
│ )
)
Contributor
There was a problem hiding this comment.
Just for my understanding: sorry am still confused. What about subsequent iteration? Why dont we need to yield? Might be good to add a comment here.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
feat (1/n): agents resume turn (Sync updates from stainless branch: yanxi0830/dev) #157
Server change: feat(3/n): agent resume_turn llama-stack#1194
Test Plan