[MAIN-2708] Gemini support (#238)

Tested on

- Robusta
- bedrock/anthropic.claude-3-5-sonnet-20240620-v1:0
- openai/gpt-4o
- vertex_ai/gemini-2.0-flash-exp
- gemini/gemini-pro

README.md

@@ -464,6 +464,45 @@ export OPENAI_API_KEY=123
 holmes ask "what pods are unhealthy in my cluster?" --model="openai/llama3.1"
 ```
 
+</details>
+<details>
+<summary>Gemini/Google AI Studio</summary>
+
+To use Gemini, set the `GEMINI_API_KEY` environment variable as follows:
+
+```bash
+export GEMINI_API_KEY="your-gemini-api-key"
+```
+
+Once the environment variable is set, you can run the following command to interact with Gemini:
+
+```bash
+holmes ask "what pods are unhealthy and why?" --model=gemini/<MODEL_NAME>
+```
+
+Be sure to replace `MODEL_NAME` with a model you have access to - e.g., `gemini-pro`,`gemini/gemini-1.5-flash`, etc.
+
+</details>
+<details>
+<summary>Vertex AI Gemini</summary>
+
+To use Vertex AI with Gemini models, set the following environment variables:
+
+```bash
+export VERTEXAI_PROJECT="your-project-id"
+export VERTEXAI_LOCATION="us-central1" 
+export GOOGLE_APPLICATION_CREDENTIALS="path/to/your/service_account_key.json"
+```
+
+Once the environment variables are set, you can run the following command to interact with Vertex AI Gemini models:
+
+```bash
+poetry run python holmes.py ask "what pods are unhealthy and why?" --model "vertex_ai/<MODEL_NAME>"
+```
+
+Be sure to replace `MODEL_NAME` with a model you have access to - e.g., `gemini-pro`,`gemini-2.0-flash-exp`, etc.
+Ensure you have the correct project, location, and credentials for accessing the desired Vertex AI model.
+
 </details>
 <details>
 <summary>Using other OpenAI-compatible models</summary>

helm/holmes/templates/holmes.yaml

@@ -55,6 +55,9 @@ spec:
             mountPath: /etc/robusta/config
           - name: custom-toolsets-configmap
             mountPath: /etc/holmes/config
+          {{- if .Values.additionalVolumeMounts -}}
+          {{ toYaml .Values.additionalVolumeMounts | nindent 10 }}
+          {{- end }}
         resources:
           requests:
             cpu: {{ .Values.resources.requests.cpu }}
@@ -74,7 +77,9 @@ spec:
           configMap:
             name: custom-toolsets-configmap
             optional: true
-
+        {{- if .Values.additionalVolumes -}}
+        {{ toYaml .Values.additionalVolumes | nindent 8 }}
+        {{- end }}
 ---
 apiVersion: v1
 kind: Service

helm/holmes/values.yaml

@@ -43,3 +43,6 @@ resources:
     memory: 1024Mi
   limits:
     memory: 1024Mi
+
+additionalVolumes: []
+additionalVolumeMounts: []

holmes/core/llm.py

@@ -121,6 +121,8 @@ def _strip_model_prefix(self) -> str:
             model_name = model_name[len('openai/'):]  # Strip the 'openai/' prefix
         elif model_name.startswith('bedrock/'):
             model_name = model_name[len('bedrock/'):]  # Strip the 'bedrock/' prefix
+        elif model_name.startswith('vertex_ai/'):
+            model_name = model_name[len('vertex_ai/'):]  # Strip the 'vertex_ai/' prefix
 
         return model_name
 

holmes/core/tools.py

@@ -103,6 +103,11 @@ def get_openai_format(self):
                 },
             },
         }
+
+        # gemini doesnt have parameters object if it is without params
+        if tool_properties is None:
+            result["function"].pop("parameters")
+
         return result
 
     @abstractmethod
@@ -391,7 +396,6 @@ def get_tool_by_name(self, name: str) -> Optional[YAMLTool]:
     def get_all_tools_openai_format(self):
         return [tool.get_openai_format() for tool in self.tools_by_name.values()]
 
-
 class ToolsetYamlFromConfig(Toolset):
     name: str
     enabled: bool = True

pyproject.toml

@@ -39,6 +39,7 @@ playwright = "1.48.0"
 bs4 = "^0.0.2"
 markdownify = "^0.13.1"
 starlette = "^0.40"
+google-api-python-client = "^2.156.0"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^8.3.3"