Is your feature request related to a problem? Please describe.
The Python ADK currently lacks a straightforward, serverless, and "happy path" solution for persistent session management in a Google Cloud environment. This presents a challenge for developers moving agents from local testing to production.
The current core options have limitations for this specific use case:
InMemorySessionService: Excellent for local dev, but not viable for production as state is lost on restart and cannot be shared across instances.VertexAiSessionService: While powerful, it introduces a high barrier to entry (requires deploying a "Reasoning Engine" and complex setup).DatabaseSessionService: Requires managing a SQL database (PostgreSQL/MySQL), which adds infrastructure overhead compared to a serverless NoSQL solution.
There is currently no native, serverless option in the core library that allows for immediate, low-config persistence on Google Cloud.
Describe the solution you'd like
I request the addition of a native FirestoreSessionService implementation as a core part of google.adk.sessions.
Why this solution is needed:
- A Clear Production Path: Firestore is a scalable, serverless, and widely used Google Cloud database. It provides a simple "happy path" for developers needing persistent sessions without managing infrastructure.
- Ease of Use: Enabling Firestore and configuring it is a well-understood process, offering a much lower barrier to entry than the Reasoning Engine or a full SQL database setup.
- Parity & Demand: This aligns with similar requests in other ADK languages (e.g., Java ADK Issue #497) and addresses a common developer need for simple cloud persistence.
Describe alternatives you've considered
- Using
DatabaseSessionService: This works but forces the developer to provision and manage a SQL database instance, which is often overkill for simple session state storage.
- Using
VertexAiSessionService: Considered, but the setup complexity for Reasoning Engine is too high for simple persistence needs.
- Third-party extensions: I am aware of
adk-ext, but this functionality is fundamental enough to the Google Cloud ecosystem that it warrants inclusion in the core google/adk-python library to ensure standard support and maintenance.
Additional context
This request mirrors the "FirestoreSessionService/VertexAiSessionService gap" issue currently open in the Java ADK repository: google/adk-java#497
As noted in that issue, the demand for this is high enough that AI models often hallucinate the existence of FirestoreSessionService because it feels like a missing piece of the standard library.
Guidance on Repository Placement:
I would appreciate guidance on whether this implementation belongs in this core google/adk-python repository or in google/adk-python-community.
- My reading of the Contributing Guide suggests that integrations often go to the community repo.
- However, I noticed that
FirestoreSessionService was recently considered for the Java core library.
Given that Firestore is a fundamental "happy path" for GCP deployments, I believe it fits best in core to ensure feature parity and standard support, but I defer to the maintainers' judgment.
Is your feature request related to a problem? Please describe.
The Python ADK currently lacks a straightforward, serverless, and "happy path" solution for persistent session management in a Google Cloud environment. This presents a challenge for developers moving agents from local testing to production.
The current core options have limitations for this specific use case:
InMemorySessionService: Excellent for local dev, but not viable for production as state is lost on restart and cannot be shared across instances.VertexAiSessionService: While powerful, it introduces a high barrier to entry (requires deploying a "Reasoning Engine" and complex setup).DatabaseSessionService: Requires managing a SQL database (PostgreSQL/MySQL), which adds infrastructure overhead compared to a serverless NoSQL solution.There is currently no native, serverless option in the core library that allows for immediate, low-config persistence on Google Cloud.
Describe the solution you'd like
I request the addition of a native
FirestoreSessionServiceimplementation as a core part ofgoogle.adk.sessions.Why this solution is needed:
Describe alternatives you've considered
DatabaseSessionService: This works but forces the developer to provision and manage a SQL database instance, which is often overkill for simple session state storage.VertexAiSessionService: Considered, but the setup complexity for Reasoning Engine is too high for simple persistence needs.adk-ext, but this functionality is fundamental enough to the Google Cloud ecosystem that it warrants inclusion in the coregoogle/adk-pythonlibrary to ensure standard support and maintenance.Additional context
This request mirrors the "FirestoreSessionService/VertexAiSessionService gap" issue currently open in the Java ADK repository: google/adk-java#497
As noted in that issue, the demand for this is high enough that AI models often hallucinate the existence of
FirestoreSessionServicebecause it feels like a missing piece of the standard library.Guidance on Repository Placement:
I would appreciate guidance on whether this implementation belongs in this core
google/adk-pythonrepository or ingoogle/adk-python-community.FirestoreSessionServicewas recently considered for the Java core library.Given that Firestore is a fundamental "happy path" for GCP deployments, I believe it fits best in core to ensure feature parity and standard support, but I defer to the maintainers' judgment.