alisha-1000 opened a new issue, #41299:
URL: https://github.com/apache/beam/issues/41299
### What happened?
`ADKAgentModelHandler` fails on the second call that uses the same
`session_id`, so the stateful multi-turn mode described in the class
docstring does not work.
**Cause**
In `_invoke_agent`
(sdks/python/apache_beam/ml/inference/agent_development_kit.py):
```python
try:
await runner.session_service.get_session(session_id)
except Exception as e:
await runner.session_service.create_session(
app_name=app_name, user_id=user_id, session_id=session_id)
```
1. ADK's `get_session` takes keyword-only arguments
(`*, app_name, user_id, session_id, config=None`), so the positional
call always raises `TypeError`, whether or not the session exists.
2. `except Exception` swallows it and calls `create_session`. For a new
session this happens to work, which is why stateless runs pass.
3. For an existing session, `create_session` raises
`AlreadyExistsError: Session with id <id> already exists.`
Fixing only the arguments is not enough: with correct arguments,
`InMemorySessionService.get_session` returns `None` for a missing
session instead of raising, so the `except` branch would never run.
**Reproduction**
Direct calls against `InMemorySessionService` (google-adk 2.14.1):
| Call | Result |
| --- | --- |
| `get_session("sid")` | `TypeError: ... takes 1 positional argument but 2
were given` |
| `get_session(app_name=..., user_id=..., session_id=...)`, missing session
| returns `None` |
| second `create_session` with the same id | `AlreadyExistsError` |
Through the handler: two `RunInference` runs with
`inference_args={"session_id": "fixed-session-1"}` and a shared
`InMemorySessionService` passed via `session_service_factory`.
Run 1 succeeds. Run 2 fails with the `TypeError` at the `get_session`
line followed by `AlreadyExistsError` from `create_session`.
The failure happens before the model is called, so it does not depend
on the model. I reproduced it with a local model (LiteLlm + Ollama).
**Environment**
- apache-beam 2.78.0.dev0 (master at 7ab792f471f), DirectRunner
- google-adk 2.14.1, Python 3.13.6, macOS arm64
The same code is present in the 2.75.0 release, and the same call is
unchanged in draft PR #39074 (line 463 there).
I would like to work on a fix.
### Issue Priority
Priority: 2 (default / most bugs should be filed as P2)
### Issue Components
- [x] Component: Python SDK
- [ ] Component: Java SDK
- [ ] Component: Go SDK
- [ ] Component: Typescript SDK
- [ ] Component: IO connector
- [ ] Component: Beam YAML
- [ ] Component: Beam examples
- [ ] Component: Beam playground
- [ ] Component: Beam katas
- [ ] Component: Website
- [ ] Component: Infrastructure
- [ ] Component: Spark Runner
- [ ] Component: Flink Runner
- [ ] Component: Prism Runner
- [ ] Component: Hazelcast Jet Runner
- [ ] Component: Google Cloud Dataflow Runner
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
