feat(provider): add GitHub Models provider (chat + embeddings)

examples/github-provider-example.yaml

@@ -0,0 +1,10 @@
+id: github-model-demo
+namespace: demo.ai
+tasks:
+  - id: github_chat
+    type: io.kestra.plugin.ai.chat.ChatCompletion
+    provider:
+      type: github
+      token: "{{ secrets.GITHUB_MODELS_TOKEN }}"
+      model: "gpt-4.1-mini"
+    prompt: "Explain Kestra in 2 lines."

src/main/java/io/kestra/plugin/ai/provider/github/GitHubModelsClient.java

@@ -0,0 +1,158 @@
+package io.kestra.plugin.ai.provider.github;
+
+import com.fasterxml.jackson.core.type.TypeReference;
+import com.fasterxml.jackson.databind.JsonNode;
+import com.fasterxml.jackson.databind.ObjectMapper;
+
+import java.net.URI;
+import java.net.http.HttpClient;
+import java.net.http.HttpRequest;
+import java.net.http.HttpResponse;
+import java.time.Duration;
+import java.util.*;
+
+/**
+ * Minimal GitHub Models client (chat + embeddings).
+ *
+ * Notes:
+ * - Uses the base URL from GitHubModelsConfig (defaults to https://api.github.com).
+ * - Supports chat with `messages` (preferred) or `input`.
+ * - Parses typical response shapes: choices[0].message.content, choices[0].output, data[].embedding.
+ */
+public class GitHubModelsClient {
+    private final GitHubModelsConfig config;
+    private final HttpClient http;
+    private final ObjectMapper mapper = new ObjectMapper();
+
+    public GitHubModelsClient(GitHubModelsConfig config) {
+        this.config = config;
+        int timeout = config.getTimeoutMs() != null ? config.getTimeoutMs() : 60_000;
+        this.http = HttpClient.newBuilder()
+            .connectTimeout(Duration.ofMillis(timeout))
+            .build();
+    }
+
+    private String base() {
+        return (config.getBaseUrl() != null && !config.getBaseUrl().isBlank())
+            ? config.getBaseUrl()
+            : "https://api.github.com";
+    }
+
+    /**
+     * Call GitHub chat completion endpoint.
+     *
+     * @param model    model id (e.g. "gpt-4.1-mini")
+     * @param messages list of message maps (role, content) OR null if using input
+     * @param input    fallback input string (used if messages == null)
+     * @param params   optional additional parameters (temperature, max_tokens, etc.)
+     * @return assistant textual response (best-effort parsing)
+     * @throws Exception on network / parsing errors
+     */
+    public String chat(String model, List<Map<String, String>> messages, String input, Map<String, Object> params) throws Exception {
+        String url = base() + "/inference/chat/completions";
+
+        Map<String, Object> payload = new LinkedHashMap<>();
+        payload.put("model", model != null ? model : config.getDefaultModel());
+        if (messages != null) {
+            payload.put("messages", messages);
+        } else {
+            payload.put("input", input != null ? input : "");
+        }
+
+        if (params != null) {
+            payload.putAll(params);
+        }
+
+        HttpRequest req = HttpRequest.newBuilder()
+            .uri(URI.create(url))
+            .timeout(Duration.ofMillis(config.getTimeoutMs() != null ? config.getTimeoutMs() : 60_000))
+            .header("Authorization", "Bearer " + config.getToken())
+            .header("Accept", "application/vnd.github+json")
+            .header("Content-Type", "application/json")
+            .POST(HttpRequest.BodyPublishers.ofString(mapper.writeValueAsString(payload)))
+            .build();
+
+        HttpResponse<String> resp = http.send(req, HttpResponse.BodyHandlers.ofString());
+        int status = resp.statusCode();
+        String body = resp.body();
+
+        if (status < 200 || status >= 300) {
+            throw new RuntimeException("GitHub Models API error: " + status + " -> " + body);
+        }
+
+        JsonNode root = mapper.readTree(body);
+
+        // 1) Try choices[0].message.content
+        JsonNode choices = root.path("choices");
+        if (choices.isArray() && choices.size() > 0) {
+            JsonNode first = choices.get(0);
+            JsonNode message = first.path("message");
+            if (message.isObject() && message.has("content")) {
+                return message.get("content").asText();
+            }
+
+            // 2) fallback to output field on choice
+            if (first.has("output")) {
+                return first.get("output").asText();
+            }
+        }
+
+        // 3) fallback to top-level output
+        if (root.has("output")) {
+            return root.get("output").asText();
+        }
+
+        // 4) fallback: return whole body
+        return root.toString();
+    }
+
+    /**
+     * Call GitHub embeddings endpoint and return the first embedding vector as List<Double>.
+     *
+     * @param model model id for embeddings (e.g. "text-embedding-3-large")
+     * @param input text to embed
+     * @return embedding vector as List<Double>
+     * @throws Exception on network / parsing errors
+     */
+    public List<Double> embeddings(String model, String input) throws Exception {
+        String url = base() + "/inference/embeddings";
+
+        Map<String, Object> payload = Map.of(
+            "model", model != null ? model : config.getDefaultModel(),
+            "input", input
+        );
+
+        HttpRequest req = HttpRequest.newBuilder()
+            .uri(URI.create(url))
+            .timeout(Duration.ofMillis(config.getTimeoutMs() != null ? config.getTimeoutMs() : 60_000))
+            .header("Authorization", "Bearer " + config.getToken())
+            .header("Accept", "application/vnd.github+json")
+            .header("Content-Type", "application/json")
+            .POST(HttpRequest.BodyPublishers.ofString(mapper.writeValueAsString(payload)))
+            .build();
+
+        HttpResponse<String> resp = http.send(req, HttpResponse.BodyHandlers.ofString());
+        int status = resp.statusCode();
+        String body = resp.body();
+
+        if (status < 200 || status >= 300) {
+            throw new RuntimeException("GitHub Models API embeddings error: " + status + " -> " + body);
+        }
+
+        JsonNode root = mapper.readTree(body);
+        JsonNode data = root.path("data");
+        if (data.isArray() && data.size() > 0) {
+            JsonNode first = data.get(0);
+            JsonNode emb = first.path("embedding");
+            if (emb.isArray()) {
+                List<Double> vec = new ArrayList<>();
+                for (JsonNode n : emb) {
+                    vec.add(n.asDouble());
+                }
+                return vec;
+            }
+        }
+
+        throw new RuntimeException("No embedding found in response: " + body);
+    }
+}

src/main/java/io/kestra/plugin/ai/provider/github/GitHubProvider.java

@@ -0,0 +1,73 @@
+package io.kestra.plugin.ai.provider.github;
+
+import java.util.*;
+
+/**
+ * Adapter between Kestra plugin AI tasks and GitHubModelsClient.
+ *
+ * IMPORTANT:
+ * - This class is a neutral adapter. If your project requires implementation of a specific
+ *   provider interface (e.g., ChatProvider), you'll need to adapt method signatures accordingly.
+ * - The methods below are simple and designed to be easy to call from task code.
+ */
+public class GitHubProvider {
+    private final GitHubModelsConfig config;
+    private final GitHubModelsClient client;
+
+    public GitHubProvider(GitHubModelsConfig config) {
+        this.config = config;
+        this.client = new GitHubModelsClient(config);
+    }
+
+    /**
+     * Send a chat prompt and return the assistant reply as a String.
+     *
+     * @param model  model id (or null to use config.defaultModel)
+     * @param prompt user prompt text
+     * @return assistant reply text
+     * @throws Exception on errors
+     */
+    public String chat(String model, String prompt) throws Exception {
+        // Build a minimal messages list (user role)
+        Map<String, String> userMsg = Map.of(
+            "role", "user",
+            "content", prompt
+        );
+        List<Map<String, String>> messages = Collections.singletonList(userMsg);
+
+        // No extra params for now; can be extended to pass temperature, max_tokens, etc.
+        return client.chat(model != null ? model : config.getDefaultModel(), messages, null, null);
+    }
+
+    /**
+     * Send pre-constructed messages (for more advanced use).
+     *
+     * @param model    model id
+     * @param messages list of message maps with keys "role" and "content"
+     * @param params   extra parameters for the model call
+     * @return assistant text
+     * @throws Exception on errors
+     */
+    public String chatWithMessages(String model, List<Map<String, String>> messages, Map<String, Object> params) throws Exception {
+        return client.chat(model != null ? model : config.getDefaultModel(), messages, null, params);
+    }
+
+    /**
+     * Get embeddings for an input text.
+     *
+     * @param model model id for embeddings
+     * @param input text to embed
+     * @return embedding vector
+     * @throws Exception on errors
+     */
+    public List<Double> embeddings(String model, String input) throws Exception {
+        return client.embeddings(model != null ? model : config.getDefaultModel(), input);
+    }
+
+    /**
+     * Expose config for tests or consumers.
+     */
+    public GitHubModelsConfig getConfig() {
+        return this.config;
+    }
+}

src/test/java/io/kestra/plugin/ai/provider/github/GitHubProviderTest.java

@@ -0,0 +1,72 @@
+package io.kestra.plugin.ai.provider.github;
+
+import org.junit.jupiter.api.Test;
+
+import java.lang.reflect.Field;
+import java.util.List;
+import java.util.Map;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.mockito.Mockito.*;
+
+/**
+ * Unit tests for GitHubProvider that mock GitHubModelsClient so tests don't hit network/Docker.
+ */
+public class GitHubProviderTest {
+    @Test
+    public void chatParsesMessageContent() throws Exception {
+        // Arrange
+        GitHubModelsConfig cfg = GitHubModelsConfig.builder()
+            .token("dummy")
+            .defaultModel("gpt-4.1-mini")
+            .build();
+
+        // Create provider instance
+        GitHubProvider provider = new GitHubProvider(cfg);
+
+        // Mock client and response
+        GitHubModelsClient mockClient = mock(GitHubModelsClient.class);
+        Map<String, Object> choice = Map.of(
+            "message", Map.of("content", "hello from mock")
+        );
+        when(mockClient.chat(anyString(), anyList(), any(), anyMap()))
+            .thenReturn("hello from mock");
+
+        // Inject mock client into provider (reflection because client is private)
+        Field clientField = GitHubProvider.class.getDeclaredField("client");
+        clientField.setAccessible(true);
+        clientField.set(provider, mockClient);
+
+        // Act
+        String out = provider.chat("gpt-4.1-mini", "some prompt");
+
+        // Assert
+        assertEquals("hello from mock", out);
+        verify(mockClient, times(1)).chat(eq("gpt-4.1-mini"), anyList(), eq(null), eq(null));
+    }
+
+    @Test
+    public void embeddingsReturnsList() throws Exception {
+        // Arrange
+        GitHubModelsConfig cfg = GitHubModelsConfig.builder()
+            .token("dummy")
+            .defaultModel("embed-model")
+            .build();
+        GitHubProvider provider = new GitHubProvider(cfg);
+
+        GitHubModelsClient mockClient = mock(GitHubModelsClient.class);
+        when(mockClient.embeddings(anyString(), anyString()))
+            .thenReturn(List.of(0.1, 0.2));
+
+        Field clientField = GitHubProvider.class.getDeclaredField("client");
+        clientField.setAccessible(true);
+        clientField.set(provider, mockClient);
+
+        // Act
+        var resp = provider.embeddings("embed-model", "text to embed");
+
+        // Assert
+        assertEquals(2, resp.size());
+        verify(mockClient, times(1)).embeddings(eq("embed-model"), eq("text to embed"));
+    }
+}