google / google/adk-python

Global Session Storage

Offen
#1,793 3 Kommentare 1 Reaktion 1 zugewiesene Person Beansprucht von @hangfei Auf GitHub ansehen
needs review services
Vorherrschende Sprache
Python
Sterne
21.5k
Forks
4k
Ø Merge
1 T. 14 Std.
Gemergte PRs (30 T.)
37

Beschreibung

**Description**:
Currently, Google ADK provides session storage that is scoped only to individual sessions. This means that when multiple users interact with the same ADK agent tool (for example, a test case generation tool built for our company), each user’s context is isolated and not shared.
**Problem:**
In our use case, test cases for different features can be interrelated, and having access to a broader, company-wide context would significantly improve the quality and relevance of generated test cases. For example, if a user provides context for a particular feature, it would be beneficial for the LLM to be aware of related features, previously generated test cases, and company-specific requirements—even if those were provided in a different session or by a different user.
**Request:**
Please add support for a global or shared context storage mechanism in Google ADK, in addition to the current session-scoped storage. This would allow:

1. Storing and retrieving context that is accessible across all sessions and users for a given agent or tool.
2. Enabling the LLM to leverage company-wide knowledge, previously generated artifacts, and inter-feature relationships.
3. Improving the ability to generate more comprehensive, consistent, and context-aware outputs (such as test cases) that reflect the broader organizational context.

**Example Scenario:**

1. User A generates test cases for Feature X and provides detailed company context.
2. User B, in a different session, generates test cases for Feature Y, which is related to Feature X.
3. With global context storage, the LLM can reference the context and test cases from Feature X when generating cases for Feature Y, leading to better coverage and consistency.

**Benefits:**

1. Enhanced LLM understanding of company-specific requirements and inter-feature dependencies.
2. More robust and interconnected outputs for complex, enterprise-scale use cases.
3. Reduced duplication of context and improved collaboration across teams.

Beitragsleitfaden

Beitragsleitfaden öffnen

Bewertung

Dieses Issue wurde noch nicht bewertet.

Neue Issues direkt in Ihr Postfach

Eine kurze Übersicht über anfängerfreundliche GitHub-Issues.