@gitlab/gitlab-ai-provider 3.3.1 → 3.4.1

package/CHANGELOG.md

@@ -2,6 +2,18 @@
 
 All notable changes to this project will be documented in this file. See [Conventional Commits](https://conventionalcommits.org) for commit guidelines.
 
+## <small>3.4.1 (2026-02-06)</small>
+
+- Merge branch 'vg/token_refresh_fix' into 'main' ([8c1f2c4](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/8c1f2c4))
+- fix: detect context overflow errors before token refresh ([ce272cd](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/ce272cd))
+
+## 3.4.0 (2026-02-02)
+
+- Merge branch 'feat/ai-gateway-headers' into 'main' ([f2d4c79](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/f2d4c79))
+- feat: add custom AI Gateway headers support ([2ef662d](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/2ef662d))
+- chore: added models testing script ([85dd9ba](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/85dd9ba))
+- chore: bump version to 3.3.1 to fix package-lock.json sync ([6e6e9e2](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/6e6e9e2))
+
 ## 3.3.0 (2026-01-23)
 
 - feat: add GPT-5.2 model support (duo-chat-gpt-5-2) ([fba8c4c](https://gitlab.com/gitlab-org/editor-extensions/gitlab-ai-provider/commit/fba8c4c))

package/README.md

@@ -272,6 +272,33 @@ const gitlab = createGitLab({
 });
 ```
 
+### AI Gateway Headers
+
+Custom headers can be sent to GitLab's AI Gateway (Anthropic/OpenAI proxy) for traffic identification and routing. By default, the provider sends `User-Agent: gitlab-ai-provider/{version}`.
+
+```typescript
+// Provider-level headers (apply to all agentic models)
+const gitlab = createGitLab({
+  apiKey: process.env.GITLAB_TOKEN,
+  aiGatewayHeaders: {
+    'X-Custom-Routing': 'premium-tier',
+  },
+});
+
+// Model-level headers (override provider-level)
+const model = gitlab.agenticChat('duo-chat-opus-4-5', {
+  aiGatewayHeaders: {
+    'X-Request-Priority': 'high',
+  },
+});
+```
+
+**Header Precedence (lowest to highest):**
+
+1. Default headers (`User-Agent: gitlab-ai-provider/{version}`)
+2. Provider-level `aiGatewayHeaders`
+3. Model-level `aiGatewayHeaders`
+
 ## 🏗️ Architecture
 
 ### Core Components
@@ -374,7 +401,8 @@ interface GitLabProviderSettings {
   apiKey?: string; // PAT or OAuth access token
   refreshToken?: string; // OAuth refresh token
   name?: string; // Provider name prefix
-  headers?: Record<string, string>; // Custom headers
+  headers?: Record<string, string>; // Custom headers for GitLab API
+  aiGatewayHeaders?: Record<string, string>; // Custom headers for AI Gateway proxy
   fetch?: typeof fetch; // Custom fetch implementation
   aiGatewayUrl?: string; // AI Gateway URL (default: 'https://cloud.gitlab.com')
 }
@@ -394,6 +422,7 @@ interface GitLabAgenticOptions {
   providerModel?: string; // Override the backend model (e.g., 'claude-sonnet-4-5-20250929' or 'gpt-5.1-2025-11-13')
   maxTokens?: number; // Default: 8192
   featureFlags?: Record<string, boolean>; // GitLab feature flags
+  aiGatewayHeaders?: Record<string, string>; // Custom headers for AI Gateway proxy (per-model)
 }
 ```
 

package/dist/index.d.mts

@@ -34,6 +34,11 @@ interface GitLabAnthropicConfig {
      * @default 'https://cloud.gitlab.com'
      */
     aiGatewayUrl?: string;
+    /**
+     * Custom headers for AI Gateway Anthropic proxy requests.
+     * Merged with headers from direct_access token response.
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 /**
  * GitLab Anthropic Language Model
@@ -60,6 +65,11 @@ declare class GitLabAnthropicLanguageModel implements LanguageModelV2 {
      * Check if an error is a token-related authentication error that can be retried
      */
     private isTokenError;
+    /**
+     * Check if an error is a context overflow error (prompt too long)
+     * These should NOT trigger token refresh and should be reported to the user.
+     */
+    private isContextOverflowError;
     /**
      * Convert AI SDK tools to Anthropic tool format
      */
@@ -154,6 +164,12 @@ interface GitLabAgenticOptions {
      * Feature flags to pass to the GitLab API
      */
     featureFlags?: Record<string, boolean>;
+    /**
+     * Custom headers for AI Gateway requests (per-model override).
+     * These headers are sent to the Anthropic/OpenAI proxy endpoints.
+     * Merged with provider-level aiGatewayHeaders (model-level takes precedence).
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 interface GitLabProviderSettings {
     /**
@@ -200,6 +216,12 @@ interface GitLabProviderSettings {
      * @default 'https://cloud.gitlab.com'
      */
     aiGatewayUrl?: string;
+    /**
+     * Custom headers to include in AI Gateway requests (Anthropic/OpenAI proxy).
+     * These headers are merged with the default headers from direct_access response.
+     * Default User-Agent: gitlab-ai-provider/{version}
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 declare function createGitLab(options?: GitLabProviderSettings): GitLabProvider;
 /**
@@ -214,6 +236,8 @@ declare function createGitLab(options?: GitLabProviderSettings): GitLabProvider;
  */
 declare const gitlab: GitLabProvider;
 
+declare const VERSION: string;
+
 interface GitLabOpenAIConfig {
     provider: string;
     instanceUrl: string;
@@ -228,6 +252,11 @@ interface GitLabOpenAIConfig {
     aiGatewayUrl?: string;
     /** Whether to use the Responses API instead of Chat Completions API */
     useResponsesApi?: boolean;
+    /**
+     * Custom headers for AI Gateway OpenAI proxy requests.
+     * Merged with headers from direct_access token response.
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 declare class GitLabOpenAILanguageModel implements LanguageModelV2 {
     readonly specificationVersion: "v2";
@@ -241,6 +270,11 @@ declare class GitLabOpenAILanguageModel implements LanguageModelV2 {
     get provider(): string;
     private getOpenAIClient;
     private isTokenError;
+    /**
+     * Check if an error is a context overflow error (prompt too long)
+     * These should NOT trigger token refresh and should be reported to the user.
+     */
+    private isContextOverflowError;
     private convertTools;
     private convertToolChoice;
     private convertPrompt;
@@ -318,6 +352,11 @@ declare class GitLabError extends Error {
     isRateLimitError(): boolean;
     isForbiddenError(): boolean;
     isServerError(): boolean;
+    /**
+     * Check if this error is a context overflow error (prompt too long).
+     * These errors occur when the conversation exceeds the model's token limit.
+     */
+    isContextOverflowError(): boolean;
 }
 
 declare const gitlabOAuthTokenResponseSchema: z.ZodObject<{
@@ -641,4 +680,4 @@ declare class GitLabDirectAccessClient {
     invalidateToken(): void;
 }
 
-export { BUNDLED_CLIENT_ID, DEFAULT_AI_GATEWAY_URL, type DirectAccessToken, GITLAB_COM_URL, type GitLabAgenticOptions, type GitLabAnthropicConfig, GitLabAnthropicLanguageModel, GitLabDirectAccessClient, type GitLabDirectAccessConfig, GitLabError, type GitLabErrorOptions, GitLabOAuthManager, type GitLabOAuthTokenResponse, type GitLabOAuthTokens, type GitLabOpenAIConfig, GitLabOpenAILanguageModel, type GitLabProject, GitLabProjectCache, GitLabProjectDetector, type GitLabProjectDetectorConfig, type GitLabProvider, type GitLabProviderSettings, MODEL_ID_TO_ANTHROPIC_MODEL, MODEL_MAPPINGS, type ModelMapping, type ModelProvider, OAUTH_SCOPES, type OpenAIApiType, type OpenCodeAuth, type OpenCodeAuthApi, type OpenCodeAuthOAuth, TOKEN_EXPIRY_SKEW_MS, createGitLab, getAnthropicModelForModelId, getModelMapping, getOpenAIApiType, getOpenAIModelForModelId, getProviderForModelId, getValidModelsForProvider, gitlab, isResponsesApiModel };
+export { BUNDLED_CLIENT_ID, DEFAULT_AI_GATEWAY_URL, type DirectAccessToken, GITLAB_COM_URL, type GitLabAgenticOptions, type GitLabAnthropicConfig, GitLabAnthropicLanguageModel, GitLabDirectAccessClient, type GitLabDirectAccessConfig, GitLabError, type GitLabErrorOptions, GitLabOAuthManager, type GitLabOAuthTokenResponse, type GitLabOAuthTokens, type GitLabOpenAIConfig, GitLabOpenAILanguageModel, type GitLabProject, GitLabProjectCache, GitLabProjectDetector, type GitLabProjectDetectorConfig, type GitLabProvider, type GitLabProviderSettings, MODEL_ID_TO_ANTHROPIC_MODEL, MODEL_MAPPINGS, type ModelMapping, type ModelProvider, OAUTH_SCOPES, type OpenAIApiType, type OpenCodeAuth, type OpenCodeAuthApi, type OpenCodeAuthOAuth, TOKEN_EXPIRY_SKEW_MS, VERSION, createGitLab, getAnthropicModelForModelId, getModelMapping, getOpenAIApiType, getOpenAIModelForModelId, getProviderForModelId, getValidModelsForProvider, gitlab, isResponsesApiModel };

package/dist/index.d.ts

@@ -34,6 +34,11 @@ interface GitLabAnthropicConfig {
      * @default 'https://cloud.gitlab.com'
      */
     aiGatewayUrl?: string;
+    /**
+     * Custom headers for AI Gateway Anthropic proxy requests.
+     * Merged with headers from direct_access token response.
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 /**
  * GitLab Anthropic Language Model
@@ -60,6 +65,11 @@ declare class GitLabAnthropicLanguageModel implements LanguageModelV2 {
      * Check if an error is a token-related authentication error that can be retried
      */
     private isTokenError;
+    /**
+     * Check if an error is a context overflow error (prompt too long)
+     * These should NOT trigger token refresh and should be reported to the user.
+     */
+    private isContextOverflowError;
     /**
      * Convert AI SDK tools to Anthropic tool format
      */
@@ -154,6 +164,12 @@ interface GitLabAgenticOptions {
      * Feature flags to pass to the GitLab API
      */
     featureFlags?: Record<string, boolean>;
+    /**
+     * Custom headers for AI Gateway requests (per-model override).
+     * These headers are sent to the Anthropic/OpenAI proxy endpoints.
+     * Merged with provider-level aiGatewayHeaders (model-level takes precedence).
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 interface GitLabProviderSettings {
     /**
@@ -200,6 +216,12 @@ interface GitLabProviderSettings {
      * @default 'https://cloud.gitlab.com'
      */
     aiGatewayUrl?: string;
+    /**
+     * Custom headers to include in AI Gateway requests (Anthropic/OpenAI proxy).
+     * These headers are merged with the default headers from direct_access response.
+     * Default User-Agent: gitlab-ai-provider/{version}
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 declare function createGitLab(options?: GitLabProviderSettings): GitLabProvider;
 /**
@@ -214,6 +236,8 @@ declare function createGitLab(options?: GitLabProviderSettings): GitLabProvider;
  */
 declare const gitlab: GitLabProvider;
 
+declare const VERSION: string;
+
 interface GitLabOpenAIConfig {
     provider: string;
     instanceUrl: string;
@@ -228,6 +252,11 @@ interface GitLabOpenAIConfig {
     aiGatewayUrl?: string;
     /** Whether to use the Responses API instead of Chat Completions API */
     useResponsesApi?: boolean;
+    /**
+     * Custom headers for AI Gateway OpenAI proxy requests.
+     * Merged with headers from direct_access token response.
+     */
+    aiGatewayHeaders?: Record<string, string>;
 }
 declare class GitLabOpenAILanguageModel implements LanguageModelV2 {
     readonly specificationVersion: "v2";
@@ -241,6 +270,11 @@ declare class GitLabOpenAILanguageModel implements LanguageModelV2 {
     get provider(): string;
     private getOpenAIClient;
     private isTokenError;
+    /**
+     * Check if an error is a context overflow error (prompt too long)
+     * These should NOT trigger token refresh and should be reported to the user.
+     */
+    private isContextOverflowError;
     private convertTools;
     private convertToolChoice;
     private convertPrompt;
@@ -318,6 +352,11 @@ declare class GitLabError extends Error {
     isRateLimitError(): boolean;
     isForbiddenError(): boolean;
     isServerError(): boolean;
+    /**
+     * Check if this error is a context overflow error (prompt too long).
+     * These errors occur when the conversation exceeds the model's token limit.
+     */
+    isContextOverflowError(): boolean;
 }
 
 declare const gitlabOAuthTokenResponseSchema: z.ZodObject<{
@@ -641,4 +680,4 @@ declare class GitLabDirectAccessClient {
     invalidateToken(): void;
 }
 
-export { BUNDLED_CLIENT_ID, DEFAULT_AI_GATEWAY_URL, type DirectAccessToken, GITLAB_COM_URL, type GitLabAgenticOptions, type GitLabAnthropicConfig, GitLabAnthropicLanguageModel, GitLabDirectAccessClient, type GitLabDirectAccessConfig, GitLabError, type GitLabErrorOptions, GitLabOAuthManager, type GitLabOAuthTokenResponse, type GitLabOAuthTokens, type GitLabOpenAIConfig, GitLabOpenAILanguageModel, type GitLabProject, GitLabProjectCache, GitLabProjectDetector, type GitLabProjectDetectorConfig, type GitLabProvider, type GitLabProviderSettings, MODEL_ID_TO_ANTHROPIC_MODEL, MODEL_MAPPINGS, type ModelMapping, type ModelProvider, OAUTH_SCOPES, type OpenAIApiType, type OpenCodeAuth, type OpenCodeAuthApi, type OpenCodeAuthOAuth, TOKEN_EXPIRY_SKEW_MS, createGitLab, getAnthropicModelForModelId, getModelMapping, getOpenAIApiType, getOpenAIModelForModelId, getProviderForModelId, getValidModelsForProvider, gitlab, isResponsesApiModel };
+export { BUNDLED_CLIENT_ID, DEFAULT_AI_GATEWAY_URL, type DirectAccessToken, GITLAB_COM_URL, type GitLabAgenticOptions, type GitLabAnthropicConfig, GitLabAnthropicLanguageModel, GitLabDirectAccessClient, type GitLabDirectAccessConfig, GitLabError, type GitLabErrorOptions, GitLabOAuthManager, type GitLabOAuthTokenResponse, type GitLabOAuthTokens, type GitLabOpenAIConfig, GitLabOpenAILanguageModel, type GitLabProject, GitLabProjectCache, GitLabProjectDetector, type GitLabProjectDetectorConfig, type GitLabProvider, type GitLabProviderSettings, MODEL_ID_TO_ANTHROPIC_MODEL, MODEL_MAPPINGS, type ModelMapping, type ModelProvider, OAUTH_SCOPES, type OpenAIApiType, type OpenCodeAuth, type OpenCodeAuthApi, type OpenCodeAuthOAuth, TOKEN_EXPIRY_SKEW_MS, VERSION, createGitLab, getAnthropicModelForModelId, getModelMapping, getOpenAIApiType, getOpenAIModelForModelId, getProviderForModelId, getValidModelsForProvider, gitlab, isResponsesApiModel };

package/dist/index.js

@@ -43,6 +43,7 @@ __export(index_exports, {
   MODEL_MAPPINGS: () => MODEL_MAPPINGS,
   OAUTH_SCOPES: () => OAUTH_SCOPES,
   TOKEN_EXPIRY_SKEW_MS: () => TOKEN_EXPIRY_SKEW_MS,
+  VERSION: () => VERSION,
   createGitLab: () => createGitLab,
   getAnthropicModelForModelId: () => getAnthropicModelForModelId,
   getModelMapping: () => getModelMapping,
@@ -95,6 +96,17 @@ var GitLabError = class _GitLabError extends Error {
   isServerError() {
     return this.statusCode !== void 0 && this.statusCode >= 500;
   }
+  /**
+   * Check if this error is a context overflow error (prompt too long).
+   * These errors occur when the conversation exceeds the model's token limit.
+   */
+  isContextOverflowError() {
+    if (this.statusCode !== 400) {
+      return false;
+    }
+    const message = this.message?.toLowerCase() || "";
+    return message.includes("context overflow") || message.includes("prompt is too long") || message.includes("prompt too long") || message.includes("tokens") && message.includes("maximum");
+  }
 };
 
 // src/gitlab-direct-access.ts
@@ -238,11 +250,15 @@ var GitLabAnthropicLanguageModel = class {
   async getAnthropicClient(forceRefresh = false) {
     const tokenData = await this.directAccessClient.getDirectAccessToken(forceRefresh);
     const { "x-api-key": _removed, ...filteredHeaders } = tokenData.headers;
+    const mergedHeaders = {
+      ...filteredHeaders,
+      ...this.config.aiGatewayHeaders
+    };
     this.anthropicClient = new import_sdk.default({
       apiKey: null,
       authToken: tokenData.token,
       baseURL: this.directAccessClient.getAnthropicProxyUrl(),
-      defaultHeaders: filteredHeaders
+      defaultHeaders: mergedHeaders
     });
     return this.anthropicClient;
   }
@@ -261,6 +277,21 @@ var GitLabAnthropicLanguageModel = class {
     }
     return false;
   }
+  /**
+   * Check if an error is a context overflow error (prompt too long)
+   * These should NOT trigger token refresh and should be reported to the user.
+   */
+  isContextOverflowError(error) {
+    if (error instanceof import_sdk.default.APIError) {
+      if (error.status === 400) {
+        const message = error.message?.toLowerCase() || "";
+        if (message.includes("prompt is too long") || message.includes("prompt too long") || message.includes("tokens") && message.includes("maximum")) {
+          return true;
+        }
+      }
+    }
+    return false;
+  }
   /**
    * Convert AI SDK tools to Anthropic tool format
    */
@@ -438,6 +469,14 @@ var GitLabAnthropicLanguageModel = class {
         warnings: []
       };
     } catch (error) {
+      if (this.isContextOverflowError(error)) {
+        const apiError = error;
+        throw new GitLabError({
+          message: `Context overflow: ${apiError.message}. Please start a new session or use /compact to reduce context.`,
+          statusCode: 400,
+          cause: error
+        });
+      }
       if (!isRetry && this.isTokenError(error)) {
         this.directAccessClient.invalidateToken();
         return this.doGenerateWithRetry(options, true);
@@ -445,6 +484,7 @@ var GitLabAnthropicLanguageModel = class {
       if (error instanceof import_sdk.default.APIError) {
         throw new GitLabError({
           message: `Anthropic API error: ${error.message}`,
+          statusCode: error.status,
           cause: error
         });
       }
@@ -625,6 +665,19 @@ var GitLabAnthropicLanguageModel = class {
               });
             }
           }
+          if (self.isContextOverflowError(error)) {
+            const apiError = error;
+            controller.enqueue({
+              type: "error",
+              error: new GitLabError({
+                message: `Context overflow: ${apiError.message}. Please start a new session or use /compact to reduce context.`,
+                statusCode: 400,
+                cause: error
+              })
+            });
+            controller.close();
+            return;
+          }
           if (!isRetry && self.isTokenError(error)) {
             self.directAccessClient.invalidateToken();
             controller.enqueue({
@@ -642,6 +695,7 @@ var GitLabAnthropicLanguageModel = class {
               type: "error",
               error: new GitLabError({
                 message: `Anthropic API error: ${error.message}`,
+                statusCode: error.status,
                 cause: error
               })
             });
@@ -743,10 +797,14 @@ var GitLabOpenAILanguageModel = class {
   async getOpenAIClient(forceRefresh = false) {
     const tokenData = await this.directAccessClient.getDirectAccessToken(forceRefresh);
     const { "x-api-key": _removed, ...filteredHeaders } = tokenData.headers;
+    const mergedHeaders = {
+      ...filteredHeaders,
+      ...this.config.aiGatewayHeaders
+    };
     this.openaiClient = new import_openai.default({
       apiKey: tokenData.token,
       baseURL: this.directAccessClient.getOpenAIProxyUrl(),
-      defaultHeaders: filteredHeaders
+      defaultHeaders: mergedHeaders
     });
     return this.openaiClient;
   }
@@ -762,6 +820,21 @@ var GitLabOpenAILanguageModel = class {
     }
     return false;
   }
+  /**
+   * Check if an error is a context overflow error (prompt too long)
+   * These should NOT trigger token refresh and should be reported to the user.
+   */
+  isContextOverflowError(error) {
+    if (error instanceof import_openai.default.APIError) {
+      if (error.status === 400) {
+        const message = error.message?.toLowerCase() || "";
+        if (message.includes("prompt is too long") || message.includes("prompt too long") || message.includes("tokens") && message.includes("maximum")) {
+          return true;
+        }
+      }
+    }
+    return false;
+  }
   convertTools(tools) {
     if (!tools || tools.length === 0) {
       return void 0;
@@ -1042,6 +1115,14 @@ var GitLabOpenAILanguageModel = class {
         warnings: []
       };
     } catch (error) {
+      if (this.isContextOverflowError(error)) {
+        const apiError = error;
+        throw new GitLabError({
+          message: `Context overflow: ${apiError.message}. Please start a new session or use /compact to reduce context.`,
+          statusCode: 400,
+          cause: error
+        });
+      }
       if (!isRetry && this.isTokenError(error)) {
         this.directAccessClient.invalidateToken();
         return this.doGenerateWithChatApi(options, true);
@@ -1049,6 +1130,7 @@ var GitLabOpenAILanguageModel = class {
       if (error instanceof import_openai.default.APIError) {
         throw new GitLabError({
           message: `OpenAI API error: ${error.message}`,
+          statusCode: error.status,
           cause: error
         });
       }
@@ -1104,6 +1186,14 @@ var GitLabOpenAILanguageModel = class {
         warnings: []
       };
     } catch (error) {
+      if (this.isContextOverflowError(error)) {
+        const apiError = error;
+        throw new GitLabError({
+          message: `Context overflow: ${apiError.message}. Please start a new session or use /compact to reduce context.`,
+          statusCode: 400,
+          cause: error
+        });
+      }
       if (!isRetry && this.isTokenError(error)) {
         this.directAccessClient.invalidateToken();
         return this.doGenerateWithResponsesApi(options, true);
@@ -1111,6 +1201,7 @@ var GitLabOpenAILanguageModel = class {
       if (error instanceof import_openai.default.APIError) {
         throw new GitLabError({
           message: `OpenAI API error: ${error.message}`,
+          statusCode: error.status,
           cause: error
         });
       }
@@ -1229,6 +1320,19 @@ var GitLabOpenAILanguageModel = class {
           controller.enqueue({ type: "finish", finishReason, usage });
           controller.close();
         } catch (error) {
+          if (self.isContextOverflowError(error)) {
+            const apiError = error;
+            controller.enqueue({
+              type: "error",
+              error: new GitLabError({
+                message: `Context overflow: ${apiError.message}. Please start a new session or use /compact to reduce context.`,
+                statusCode: 400,
+                cause: error
+              })
+            });
+            controller.close();
+            return;
+          }
           if (!isRetry && self.isTokenError(error)) {
             self.directAccessClient.invalidateToken();
             controller.enqueue({
@@ -1243,6 +1347,7 @@ var GitLabOpenAILanguageModel = class {
               type: "error",
               error: new GitLabError({
                 message: `OpenAI API error: ${error.message}`,
+                statusCode: error.status,
                 cause: error
               })
             });
@@ -1369,6 +1474,19 @@ var GitLabOpenAILanguageModel = class {
           controller.enqueue({ type: "finish", finishReason, usage });
           controller.close();
         } catch (error) {
+          if (self.isContextOverflowError(error)) {
+            const apiError = error;
+            controller.enqueue({
+              type: "error",
+              error: new GitLabError({
+                message: `Context overflow: ${apiError.message}. Please start a new session or use /compact to reduce context.`,
+                statusCode: 400,
+                cause: error
+              })
+            });
+            controller.close();
+            return;
+          }
           if (!isRetry && self.isTokenError(error)) {
             self.directAccessClient.invalidateToken();
             controller.enqueue({
@@ -1383,6 +1501,7 @@ var GitLabOpenAILanguageModel = class {
               type: "error",
               error: new GitLabError({
                 message: `OpenAI API error: ${error.message}`,
+                statusCode: error.status,
                 cause: error
               })
             });
@@ -1559,11 +1678,13 @@ var GitLabOAuthManager = class {
   }
 };
 
+// src/version.ts
+var VERSION = true ? "3.4.0" : "0.0.0-dev";
+
 // src/gitlab-provider.ts
 var fs = __toESM(require("fs"));
 var path = __toESM(require("path"));
 var os = __toESM(require("os"));
-var VERSION = "0.0.1";
 function getOpenCodeAuthPath() {
   const homeDir = os.homedir();
   const xdgDataHome = process.env.XDG_DATA_HOME;
@@ -1713,6 +1834,14 @@ function createGitLab(options = {}) {
       ...options.featureFlags,
       ...agenticOptions?.featureFlags
     };
+    const defaultAiGatewayHeaders = {
+      "User-Agent": `gitlab-ai-provider/${VERSION}`
+    };
+    const aiGatewayHeaders = {
+      ...defaultAiGatewayHeaders,
+      ...options.aiGatewayHeaders,
+      ...agenticOptions?.aiGatewayHeaders
+    };
     const baseConfig = {
       provider: `${providerName}.agentic`,
       instanceUrl,
@@ -1721,7 +1850,8 @@ function createGitLab(options = {}) {
       fetch: options.fetch,
       maxTokens: agenticOptions?.maxTokens,
       featureFlags,
-      aiGatewayUrl: options.aiGatewayUrl
+      aiGatewayUrl: options.aiGatewayUrl,
+      aiGatewayHeaders
     };
     if (mapping.provider === "openai") {
       return new GitLabOpenAILanguageModel(modelId, {
@@ -2042,6 +2172,7 @@ var GitLabProjectDetector = class {
   MODEL_MAPPINGS,
   OAUTH_SCOPES,
   TOKEN_EXPIRY_SKEW_MS,
+  VERSION,
   createGitLab,
   getAnthropicModelForModelId,
   getModelMapping,