adk_ge_datastore_connector

Name: adk_ge_datastore_connector
Author: VeerMuchandi

15stars1forks

How to install adk_ge_datastore_connector?

veermuchandi/rad-skills/adk_ge_datastore_connector

$npx -y skills add VeerMuchandi/rad-skills --skill adk_ge_datastore_connector

Installs into the current project.

Summary

Skill for connecting ADK agents to Gemini Enterprise connected data sources with secure user-level access control.

Files · 1

View on GitHub

SKILL.md

1	# Instructions
2
3	This skill provides guidelines and patterns for connecting ADK agents to Gemini Enterprise connected data sources. The primary focus is on secure user-level access control (ACLs) by propagating the user's OAuth token.
4
5	## Core Concepts
6
7	### 1. User Identity Propagation
8	When searching an enterprise datastore (e.g., Google Drive, Jira), the agent must respect the calling user's permissions. This is achieved by passing the user's OAuth token directly to the Datastore API (Discovery Engine API).
9
10	### 2. Token Retrieval
11	In the Gemini Enterprise runtime, the user's access token is injected into the agent's session state.
12	Retrieve the token from the `ToolContext` using the key defined by the environment variable `AUTH_NAME`:
13
14	```python
15	from google.adk.tools import ToolContext
16
17	def my_tool(query: str, tool_context: ToolContext):
18	auth_name = os.getenv("AUTH_NAME")
19	access_token = tool_context.state.get(auth_name)
20	# Use access_token in API calls
21	```
22
23	### 3. Local Development Fallback
24	For local testing where the Gemini Enterprise runtime is not available, fall back to Application Default Credentials (ADC).
25
26	```python
27	from google.auth import default
28	from google.auth import transport
29
30	if access_token:
31	# Use retrieved token
32	else:
33	# Fallback to ADC
34	creds, project_id = default()
35	auth_req = transport.requests.Request()
36	creds.refresh(auth_req)
37	access_token = creds.token
38	```
39
40	## Implementation Patterns
41
42	### Generic Datastore Client
43	Use a generic client to handle API communication. Do not hardcode specific datastore IDs or project IDs; use environment variables.
44
45	See [examples/tools_template.py](./examples/tools_template.py) for a reusable template.
46
47	### Agent Integration
48	Focus this skill on integrating the datastore tool. Use the `ADK Developer` skill for general agent building and orchestration.
49
50	See [examples/agent_usage.py](./examples/agent_usage.py) for a minimal usage example.
51
52	## Design Considerations
53
54	### Documenting Testing Differences
55	When designing an agent that uses this skill, the architect MUST include a section in the architecture document detailing the differences between local testing and testing on Gemini Enterprise. This is crucial because the authentication mechanisms differ significantly (ADC vs. Session-injected OAuth tokens), affecting how user permissions (ACLs) are enforced.
56
57	Include a table or summary similar to the following in the design:
58
59	\| Feature \| Local Testing \| Gemini Enterprise \|
60	\| :--- \| :--- \| :--- \|
61	\| Auth Method \| Application Default Credentials (ADC) \| User OAuth Token (Session Injected) \|
62	\| Identity Used \| Developer's identity \| Calling End-User's identity \|
63	\| ACL Enforcement \| Based on Developer's access \| Based on End-User's access \|
64
65	### Environment Variables Fallbacks
66	When retrieving configuration like `PROJECT_ID` and `LOCATION` in your tools, use fallbacks to support both local development and Agent Engine runtime:
67	```python
68	project_id = os.getenv("PROJECT_ID") or os.getenv("GOOGLE_CLOUD_PROJECT")
69	location = os.getenv("LOCATION") or os.getenv("GOOGLE_CLOUD_LOCATION")
70	```
71	Note that `GOOGLE_CLOUD_PROJECT` is reserved in Agent Engine and cannot be set manually in the environment variables block of the deployment config.
72
73	### Prompt Engineering for Search
74	When instructing the agent to use the datastore search tool, ensure it handles relative or temporal queries (like 'last', 'latest', 'recent') gracefully. Keyword search engines might not understand these terms directly. Instruct the agent to:
75	1. Try searching with a broad query or keywords related to the topic instead of refusing to answer immediately.
76	2. Analyze the results to identify the most relevant or recent documents if date information is available.
77	3. Avoid early refusal when specific keywords are not provided.
78
79	## Deployment Considerations
80
81	### Terraform Deployment (Preferred)
82	For production deployments, it is recommended to use the generalized Terraform approach documented in the `Agent Engine Terraform Deployer` skill. This ensures reproducible deployments and proper management of infrastructure state.
83
84	### Gemini Enterprise Registration
85	Use the `Gemini Enterprise Agent Registrar` skill to perform the registration and handle OAuth setup. When constructing the registration payload, ensure that:
86	1. An Authorization Resource is created for the OAuth flow to propagate the user's token.
87	2. The `authorizationConfig` in the registration payload uses the numeric `PROJECT_NUMBER` instead of the `PROJECT_ID`.
88	3. Verify if the data source requires additional OAuth scopes beyond `cloud-platform` (e.g., for real-time federation to third-party systems like Jira or Workspace apps like Google Drive).

People who liked this also viewed

6 picks

Type

azure-storageAzure Storage Services including Blob Storage, File Shares, Queue Storage, Table Storage, and Data Lake. Answers questions about storage access tiers (hot, cool, cold, archive), when to use each tier,SkillsJul 2026471k1.3k
supabase-postgres-best-practicesPostgres performance optimization and best practices from Supabase. Use this skill when writing, reviewing, or optimizing Postgres queries, schema designs, or database configurations.SkillsJul 2026302k2.4k
supabaseUse when doing ANY task involving Supabase. Triggers: Supabase products (Database, Auth, Edge Functions, Realtime, Storage, Vectors, Cron, Queues); client libraries and SSR integrations (supabase-js, SkillsJul 2026179k2.4k
convex-quickstartCreates or adds Convex to an app. Use for new Convex projects, npm create convex@latest, frontend setup, env vars, or the first npx convex dev run.SkillsJul 202690k39
convex-performance-auditAudits Convex performance for reads, subscriptions, write contention, and function limits. Use for slow features, insights findings, OCC conflicts, or read amplification.SkillsJul 202689k39
convex-setup-authSets up Convex auth, identity mapping, and access control. Use for login, auth providers, users tables, protected functions, or roles in a Convex app.SkillsJul 202689k39

adk_ge_datastore_connector

15stars1forks

View on GitHub

How to install adk_ge_datastore_connector?

veermuchandi/rad-skills/adk_ge_datastore_connector

$npx -y skills add VeerMuchandi/rad-skills --skill adk_ge_datastore_connector

Installs into the current project.

Summary

Skill for connecting ADK agents to Gemini Enterprise connected data sources with secure user-level access control.

Files · 1

View on GitHub

SKILL.md

1	# Instructions
2
3	This skill provides guidelines and patterns for connecting ADK agents to Gemini Enterprise connected data sources. The primary focus is on secure user-level access control (ACLs) by propagating the user's OAuth token.
4
5	## Core Concepts
6
7	### 1. User Identity Propagation
8	When searching an enterprise datastore (e.g., Google Drive, Jira), the agent must respect the calling user's permissions. This is achieved by passing the user's OAuth token directly to the Datastore API (Discovery Engine API).
9
10	### 2. Token Retrieval
11	In the Gemini Enterprise runtime, the user's access token is injected into the agent's session state.
12	Retrieve the token from the `ToolContext` using the key defined by the environment variable `AUTH_NAME`:
13
14	```python
15	from google.adk.tools import ToolContext
16
17	def my_tool(query: str, tool_context: ToolContext):
18	auth_name = os.getenv("AUTH_NAME")
19	access_token = tool_context.state.get(auth_name)
20	# Use access_token in API calls
21	```
22
23	### 3. Local Development Fallback
24	For local testing where the Gemini Enterprise runtime is not available, fall back to Application Default Credentials (ADC).
25
26	```python
27	from google.auth import default
28	from google.auth import transport
29
30	if access_token:
31	# Use retrieved token
32	else:
33	# Fallback to ADC
34	creds, project_id = default()
35	auth_req = transport.requests.Request()
36	creds.refresh(auth_req)
37	access_token = creds.token
38	```
39
40	## Implementation Patterns
41
42	### Generic Datastore Client
43	Use a generic client to handle API communication. Do not hardcode specific datastore IDs or project IDs; use environment variables.
44
45	See [examples/tools_template.py](./examples/tools_template.py) for a reusable template.
46
47	### Agent Integration
48	Focus this skill on integrating the datastore tool. Use the `ADK Developer` skill for general agent building and orchestration.
49
50	See [examples/agent_usage.py](./examples/agent_usage.py) for a minimal usage example.
51
52	## Design Considerations
53
54	### Documenting Testing Differences
55	When designing an agent that uses this skill, the architect MUST include a section in the architecture document detailing the differences between local testing and testing on Gemini Enterprise. This is crucial because the authentication mechanisms differ significantly (ADC vs. Session-injected OAuth tokens), affecting how user permissions (ACLs) are enforced.
56
57	Include a table or summary similar to the following in the design:
58
59	\| Feature \| Local Testing \| Gemini Enterprise \|
60	\| :--- \| :--- \| :--- \|
61	\| Auth Method \| Application Default Credentials (ADC) \| User OAuth Token (Session Injected) \|
62	\| Identity Used \| Developer's identity \| Calling End-User's identity \|
63	\| ACL Enforcement \| Based on Developer's access \| Based on End-User's access \|
64
65	### Environment Variables Fallbacks
66	When retrieving configuration like `PROJECT_ID` and `LOCATION` in your tools, use fallbacks to support both local development and Agent Engine runtime:
67	```python
68	project_id = os.getenv("PROJECT_ID") or os.getenv("GOOGLE_CLOUD_PROJECT")
69	location = os.getenv("LOCATION") or os.getenv("GOOGLE_CLOUD_LOCATION")
70	```
71	Note that `GOOGLE_CLOUD_PROJECT` is reserved in Agent Engine and cannot be set manually in the environment variables block of the deployment config.
72
73	### Prompt Engineering for Search
74	When instructing the agent to use the datastore search tool, ensure it handles relative or temporal queries (like 'last', 'latest', 'recent') gracefully. Keyword search engines might not understand these terms directly. Instruct the agent to:
75	1. Try searching with a broad query or keywords related to the topic instead of refusing to answer immediately.
76	2. Analyze the results to identify the most relevant or recent documents if date information is available.
77	3. Avoid early refusal when specific keywords are not provided.
78
79	## Deployment Considerations
80
81	### Terraform Deployment (Preferred)
82	For production deployments, it is recommended to use the generalized Terraform approach documented in the `Agent Engine Terraform Deployer` skill. This ensures reproducible deployments and proper management of infrastructure state.
83
84	### Gemini Enterprise Registration
85	Use the `Gemini Enterprise Agent Registrar` skill to perform the registration and handle OAuth setup. When constructing the registration payload, ensure that:
86	1. An Authorization Resource is created for the OAuth flow to propagate the user's token.
87	2. The `authorizationConfig` in the registration payload uses the numeric `PROJECT_NUMBER` instead of the `PROJECT_ID`.
88	3. Verify if the data source requires additional OAuth scopes beyond `cloud-platform` (e.g., for real-time federation to third-party systems like Jira or Workspace apps like Google Drive).

People who liked this also viewed

6 picks

Type