feat(translation): add gemini generateContent wire format support

.agents/skills/switchyard-codebase-exploration/SKILL.md

@@ -42,7 +42,7 @@ Before making a code change, be able to name:
 
 - the client ingress path: CLI, FastAPI endpoint, Python API, or route-table host
 - the chain slot touched: request-side component, `LLMBackend`, response-side component, `TranslationEngine`, or wiring only
-- the wire formats involved: OpenAI Chat, Anthropic Messages, OpenAI Responses, streaming/non-streaming
+- the wire formats involved: OpenAI Chat, Anthropic Messages, OpenAI Responses, Gemini generateContent, streaming/non-streaming
 - the profile/route-table/export path that exposes the behavior
 - the focused tests and the broad validation gate to run
 

AGENTS.md

@@ -162,11 +162,13 @@ switchyard/
 │   │   ├── base.py                 # ChatResponse, ChatResponseType compatibility re-export
 │   │   ├── openai_chat.py          # ResponseStream
 │   │   ├── openai_responses.py     # ResponsesApiStream
-│   │   └── anthropic.py            # AnthropicResponseStream
+│   │   ├── anthropic.py            # AnthropicResponseStream
+│   │   └── gemini.py               # GeminiResponseStream
 │   ├── backends/                   # LLMBackend implementations
 │   │   ├── openai_llm_backend.py           # OpenAiPassthroughBackend
 │   │   ├── openai_native_backend.py        # OpenAiNativeBackend
 │   │   ├── anthropic_native_llm_backend.py # AnthropicNativeBackend
+│   │   ├── gemini_native_llm_backend.py    # GeminiNativeBackend
 │   │   ├── latency_service_llm_backend.py  # LatencyServiceLLMBackend
 │   │   ├── llm_target.py                   # LlmTarget, BackendFormat
 │   │   ├── multi_llm_backend.py            # MultiLlmBackend helpers
@@ -187,6 +189,7 @@ switchyard/
 │   ├── endpoints/                  # FastAPI endpoint wrappers (require `nemo-switchyard[server]`)
 │   │   ├── openai_chat_endpoint.py         # OpenAIChatEndpoint
 │   │   ├── anthropic_messages_endpoint.py  # AnthropicMessagesEndpoint
+│   │   ├── gemini_endpoint.py              # GeminiEndpoint
 │   │   ├── responses_endpoint.py           # ResponsesEndpoint
 │   │   ├── stats_endpoint.py               # StatsEndpoint
 │   │   ├── sse_helpers.py
@@ -330,6 +333,7 @@ uvicorn.run(build_switchyard_app(switchyard), port=4000)
 | `OPENAI_API_KEY` | API key for OpenAI-compatible backends |
 | `OPENAI_BASE_URL` | Base URL for OpenAI-compatible API |
 | `ANTHROPIC_API_KEY` | API key for Anthropic Claude |
+| `GEMINI_API_KEY` | API key for Google Gemini generateContent backends |
 | `NVIDIA_API_KEY` | API key for NVIDIA NIM / Inference Hub |
 | `OPENROUTER_API_KEY` | OpenRouter key for examples and saved provider setup; pass via `--api-key` or `switchyard configure` |
 | `SWITCHYARD_INTAKE_CAPTURE_CONTENT` | Set truthy to include prompt/response text in intake; default off (metadata-only) |

crates/switchyard-components-v2/src/backend.rs

@@ -6,7 +6,9 @@
 use std::sync::Arc;
 
 use async_trait::async_trait;
-use switchyard_components::backends::{AnthropicNativeBackend, OpenAiNativeBackend};
+use switchyard_components::backends::{
+    AnthropicNativeBackend, GeminiNativeBackend, OpenAiNativeBackend,
+};
 use switchyard_core::{
     BackendFormat, ChatRequest, ChatResponse, LlmTarget, Result, SwitchyardError,
 };
@@ -49,6 +51,7 @@ pub(crate) fn native_target_backend(target: LlmTarget) -> Result<TargetBackend>
             Arc::new(OpenAiNativeBackend::new(target.clone())?)
         }
         BackendFormat::Anthropic => Arc::new(AnthropicNativeBackend::new(target.clone())?),
+        BackendFormat::Gemini => Arc::new(GeminiNativeBackend::new(target.clone())?),
         BackendFormat::Auto => {
             return Err(SwitchyardError::InvalidConfig(format!(
                 "target {} must have a resolved backend format",
@@ -72,3 +75,10 @@ impl ProfileBackend for AnthropicNativeBackend {
         self.call_without_context(request).await
     }
 }
+
+#[async_trait]
+impl ProfileBackend for GeminiNativeBackend {
+    async fn call(&self, request: &ChatRequest) -> Result<ChatResponse> {
+        self.call_without_context(request).await
+    }
+}

crates/switchyard-components/src/backends/anthropic.rs

@@ -273,12 +273,13 @@ impl AnthropicTransport for ReqwestAnthropicTransport {
 fn validate_target_format(target: &LlmTarget) -> Result<()> {
     match target.format {
         BackendFormat::Anthropic => Ok(()),
-        BackendFormat::Auto | BackendFormat::OpenAi | BackendFormat::Responses => {
-            Err(SwitchyardError::InvalidConfig(format!(
-                "AnthropicNativeBackend requires a target with resolved Anthropic format, got {:?} for {}",
-                target.format, target.id
-            )))
-        }
+        BackendFormat::Auto
+        | BackendFormat::OpenAi
+        | BackendFormat::Responses
+        | BackendFormat::Gemini => Err(SwitchyardError::InvalidConfig(format!(
+            "AnthropicNativeBackend requires a target with resolved Anthropic format, got {:?} for {}",
+            target.format, target.id
+        ))),
     }
 }
 

crates/switchyard-components/src/backends/common.rs

@@ -60,6 +60,7 @@ pub(crate) fn request_wire_format(request_type: ChatRequestType) -> WireFormat {
         ChatRequestType::OpenAiChat => WireFormat::OpenAiChat,
         ChatRequestType::OpenAiResponses => WireFormat::OpenAiResponses,
         ChatRequestType::Anthropic => WireFormat::AnthropicMessages,
+        ChatRequestType::Gemini => WireFormat::GeminiGenerateContent,
     }
 }