Conversation
oracle-contributor-agreement
bot
added
the
OCA Verified
|
Internal regression succeeded 🍏: Build ID #370 |
sonleoracle
force-pushed
the
feat-gemini-support
branch
from
71c2a66 to
dba744d
Compare
|
Internal regression succeeded 🍏: Build ID #371 |
sonleoracle
force-pushed
the
feat-gemini-support
branch
2 times, most recently
from
b8301b2 to
8972673
Compare
sonleoracle
requested review from
a team,
cesarebernardis,
dhilloulinoracle,
jschweiz and
paul-cayet
and removed request for
jschweiz
sonleoracle
force-pushed
the
feat-gemini-support
branch
3 times, most recently
from
366b14b to
0159d0c
Compare
|
Internal regression succeeded 🍏: Build ID #385 |
pyagentspec/src/pyagentspec/serialization/pydanticserializationplugin.py
Outdated
Show resolved
Hide resolved
pyagentspec/src/pyagentspec/serialization/pydanticserializationplugin.py
Outdated
Show resolved
Hide resolved
paul-cayet
left a comment
paul-cayet
left a comment
There was a problem hiding this comment.
Approval conditional to using only leaf sensitive fields, not setting the entire component as sensitive
sonleoracle
force-pushed
the
feat-gemini-support
branch
from
0159d0c to
a865490
Compare
|
|
Internal regression succeeded 🍏: Build ID #387 |
|
|
||
| See `Google Cloud authentication docs <https://cloud.google.com/docs/authentication/application-default-credentials>`_ | ||
| and `GeminiCLI <https://geminicli.com/docs/get-started/authentication/#b-vertex-ai---service-account-json-key>`_ docs for more details. | ||
| When ``credentials`` is omitted and ADC is used instead, the auth component may remain |
There was a problem hiding this comment.
This needs to be updated. The auth component is used in GeminiConfig.auth, and it is not sensitive, so it will not be omitted. GeminiVertexAIAuthConfig.credentials instead is marked as sensitive, so it will always be omitted. I think that conditional omission is not an option at the moment.
| | StreamableHTTPmTLSTransport | ca_file | | ||
| +----------------------------------+--------------------+ | ||
|
|
||
| For Gemini auth components, only the sensitive leaf field is externalized. |
There was a problem hiding this comment.
I don't think we need to specify this, it's the same behavior as all the others.
| llm = GeminiConfig( | ||
| name="gemini-aistudio-flash", | ||
| model_id="gemini-2.5-flash", | ||
| # Optional: if omitted, runtimes may load GEMINI_API_KEY from the environment. |
There was a problem hiding this comment.
Nit: this makes it look like auth is optional, while I guess you are referring to the GeminiAIStudioAuthConfig.api_key. Make it explicit in the comment, or change the indentation so that this comment ends up inside the GeminiAIStudioAuthConfig definition.
| llm = GeminiConfig( | ||
| name="gemini-aistudio-flash", | ||
| model_id="gemini-2.5-flash", | ||
| # Optional: if omitted, runtimes may load GEMINI_API_KEY from the environment. |
| Name of the model to use, for example ``gemini-2.5-flash`` or | ||
| ``gemini-2.0-flash-lite``. | ||
|
|
||
| .. option:: auth: GeminiAIStudioAuthConfig | GeminiVertexAIAuthConfig |
There was a problem hiding this comment.
Use the right parent class
|
|
||
| Default parameters for text generation with this model. | ||
|
|
||
| AI Studio authentication |
There was a problem hiding this comment.
Nit: Gemini AI Studio authentication or Google ...
| the Gemini service to use. | ||
|
|
||
| Meanwhile, the Vertex AI service can be authenticated with Google Cloud credentials. These credentials can be provided with a `service account JSON key <https://docs.cloud.google.com/iam/docs/keys-create-delete/>`_ | ||
| either inline or through a local file path. When omitted, runtimes may rely on Google Application Default Credentials (ADC), such as |
There was a problem hiding this comment.
Can you be more precise? I guess that if credentials is a string, then it is supposed to be a file path, and when it is a dictionary, then it's the JSON, but I am not fully sure from this description. Please make it a bit clearer.
| "model_id": "gemini-2.0-flash-lite", | ||
| "auth": { | ||
| "component_type": "GeminiVertexAIAuthConfig", | ||
| "id": "gemini-vertex-auth-id", |
| "vertex_ai/gemini-2.0-flash-lite", | ||
| ], | ||
| ) | ||
| def test_geminiconfig_preserves_prefixed_model_id(model_id: str) -> None: |
There was a problem hiding this comment.
| def test_geminiconfig_preserves_prefixed_model_id(model_id: str) -> None: | |
| def test_gemini_config_preserves_prefixed_model_id(model_id: str) -> None: |
4 participants
Add Agent Spec definitions for Gemini LLMs. We support Gemini models on Google AI Studio and Google Cloud Vertex AI as providers.