co-mono/packages/coding-agent/examples/extensions/custom-provider-gitlab-duo/test.ts
Mario Zechner 7019fdb92b fix(gitlab-duo): use Responses API for all GPT models
Ensures consistent tool call ID format when switching between models.
2026-01-25 00:29:05 +01:00

380 lines
11 KiB
TypeScript

/**
* Test script for GitLab Duo - full streaming flow
* Run: npx tsx test.ts [model-id]
*
* Examples:
* npx tsx test.ts # Test default (duo-chat-sonnet-4-5)
* npx tsx test.ts duo-chat-gpt-5-codex # Test GPT-5 Codex (Responses API)
* npx tsx test.ts duo-chat-gpt-5-1 # Test GPT-5.1 (Chat Completions API)
*/
import {
type Api,
type AssistantMessageEventStream,
type Context,
createAssistantMessageEventStream,
type Model,
registerApiProvider,
type SimpleStreamOptions,
streamSimple,
} from "@mariozechner/pi-ai";
import { readFileSync } from "fs";
import { homedir } from "os";
import { join } from "path";
// =============================================================================
// Constants (copied from index.ts)
// =============================================================================
const GITLAB_COM_URL = "https://gitlab.com";
const AI_GATEWAY_URL = "https://cloud.gitlab.com";
const ANTHROPIC_PROXY_URL = `${AI_GATEWAY_URL}/ai/v1/proxy/anthropic/`;
const OPENAI_PROXY_URL = `${AI_GATEWAY_URL}/ai/v1/proxy/openai/v1`;
const DIRECT_ACCESS_TTL = 25 * 60 * 1000;
type OpenAIApi = "openai-completions" | "openai-responses";
const MODEL_MAPPINGS: Record<string, { api: "anthropic-messages" | OpenAIApi; backendModel: string; baseUrl: string }> =
{
"duo-chat-opus-4-5": {
api: "anthropic-messages",
backendModel: "claude-opus-4-5-20251101",
baseUrl: ANTHROPIC_PROXY_URL,
},
"duo-chat-sonnet-4-5": {
api: "anthropic-messages",
backendModel: "claude-sonnet-4-5-20250929",
baseUrl: ANTHROPIC_PROXY_URL,
},
"duo-chat-haiku-4-5": {
api: "anthropic-messages",
backendModel: "claude-haiku-4-5-20251001",
baseUrl: ANTHROPIC_PROXY_URL,
},
// All GPT models use Responses API for consistent tool call ID format across model switches
"duo-chat-gpt-5-1": { api: "openai-responses", backendModel: "gpt-5.1-2025-11-13", baseUrl: OPENAI_PROXY_URL },
"duo-chat-gpt-5-mini": {
api: "openai-responses",
backendModel: "gpt-5-mini-2025-08-07",
baseUrl: OPENAI_PROXY_URL,
},
"duo-chat-gpt-5-codex": { api: "openai-responses", backendModel: "gpt-5-codex", baseUrl: OPENAI_PROXY_URL },
};
// Model definitions for cost tracking
const MODEL_DEFS: Record<
string,
{
name: string;
cost: { input: number; output: number; cacheRead: number; cacheWrite: number };
contextWindow: number;
maxTokens: number;
}
> = {
"duo-chat-opus-4-5": {
name: "GitLab Duo Claude Opus 4.5",
cost: { input: 15, output: 75, cacheRead: 1.5, cacheWrite: 18.75 },
contextWindow: 200000,
maxTokens: 32000,
},
"duo-chat-sonnet-4-5": {
name: "GitLab Duo Claude Sonnet 4.5",
cost: { input: 3, output: 15, cacheRead: 0.3, cacheWrite: 3.75 },
contextWindow: 200000,
maxTokens: 16384,
},
"duo-chat-haiku-4-5": {
name: "GitLab Duo Claude Haiku 4.5",
cost: { input: 1, output: 5, cacheRead: 0.1, cacheWrite: 1.25 },
contextWindow: 200000,
maxTokens: 8192,
},
"duo-chat-gpt-5-1": {
name: "GitLab Duo GPT-5.1",
cost: { input: 2.5, output: 10, cacheRead: 0, cacheWrite: 0 },
contextWindow: 128000,
maxTokens: 16384,
},
"duo-chat-gpt-5-mini": {
name: "GitLab Duo GPT-5 Mini",
cost: { input: 0.15, output: 0.6, cacheRead: 0, cacheWrite: 0 },
contextWindow: 128000,
maxTokens: 16384,
},
"duo-chat-gpt-5-codex": {
name: "GitLab Duo GPT-5 Codex",
cost: { input: 2.5, output: 10, cacheRead: 0, cacheWrite: 0 },
contextWindow: 128000,
maxTokens: 16384,
},
};
// =============================================================================
// Direct Access Token
// =============================================================================
interface DirectAccessToken {
token: string;
headers: Record<string, string>;
expiresAt: number;
}
let cachedDirectAccess: DirectAccessToken | null = null;
async function getDirectAccessToken(gitlabAccessToken: string): Promise<DirectAccessToken> {
const now = Date.now();
if (cachedDirectAccess && cachedDirectAccess.expiresAt > now) {
return cachedDirectAccess;
}
const url = `${GITLAB_COM_URL}/api/v4/ai/third_party_agents/direct_access`;
console.log("Fetching direct access token from:", url);
const response = await fetch(url, {
method: "POST",
headers: {
Authorization: `Bearer ${gitlabAccessToken}`,
"Content-Type": "application/json",
},
body: JSON.stringify({ feature_flags: { DuoAgentPlatformNext: true } }),
});
if (!response.ok) {
const errorText = await response.text();
throw new Error(`Failed to get direct access token: ${response.status} ${errorText}`);
}
const data = (await response.json()) as { token: string; headers: Record<string, string> };
console.log("Got direct access token");
cachedDirectAccess = {
token: data.token,
headers: data.headers,
expiresAt: now + DIRECT_ACCESS_TTL,
};
return cachedDirectAccess;
}
// =============================================================================
// Stream Function (copied from index.ts)
// =============================================================================
function streamGitLabDuo(
model: Model<Api>,
context: Context,
options?: SimpleStreamOptions,
): AssistantMessageEventStream {
const stream = createAssistantMessageEventStream();
(async () => {
try {
const gitlabAccessToken = options?.apiKey;
console.log("streamGitLabDuo called");
console.log(" model.id:", model.id);
console.log(" options.apiKey present:", !!gitlabAccessToken);
if (!gitlabAccessToken) {
throw new Error("No GitLab access token provided in options.apiKey");
}
const mapping = MODEL_MAPPINGS[model.id];
if (!mapping) throw new Error(`Unknown model: ${model.id}`);
// Get direct access token (cached)
const directAccess = await getDirectAccessToken(gitlabAccessToken);
// Create a proxy model that uses the backend API
const proxyModel: Model<typeof mapping.api> = {
...model,
id: mapping.backendModel,
api: mapping.api,
baseUrl: mapping.baseUrl,
};
// Merge GitLab headers with Authorization bearer token
const headers = {
...directAccess.headers,
Authorization: `Bearer ${directAccess.token}`,
};
console.log("Calling streamSimple with proxy model:");
console.log(" proxyModel.id:", proxyModel.id);
console.log(" proxyModel.api:", proxyModel.api);
console.log(" proxyModel.baseUrl:", proxyModel.baseUrl);
// Delegate to pi-ai's built-in streaming
const innerStream = streamSimple(proxyModel, context, {
...options,
apiKey: "gitlab-duo", // Dummy value to pass validation
headers,
});
// Forward all events
for await (const event of innerStream) {
// Patch the model info back to gitlab-duo
if ("partial" in event && event.partial) {
event.partial.api = model.api;
event.partial.provider = model.provider;
event.partial.model = model.id;
}
if ("message" in event && event.message) {
event.message.api = model.api;
event.message.provider = model.provider;
event.message.model = model.id;
}
if ("error" in event && event.error) {
event.error.api = model.api;
event.error.provider = model.provider;
event.error.model = model.id;
}
stream.push(event);
}
stream.end();
} catch (error) {
console.error("Stream error:", error);
stream.push({
type: "error",
reason: "error",
error: {
role: "assistant",
content: [],
api: model.api,
provider: model.provider,
model: model.id,
usage: {
input: 0,
output: 0,
cacheRead: 0,
cacheWrite: 0,
totalTokens: 0,
cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
},
stopReason: "error",
errorMessage: error instanceof Error ? error.message : String(error),
timestamp: Date.now(),
},
});
stream.end();
}
})();
return stream;
}
// =============================================================================
// Main Test
// =============================================================================
interface AuthData {
[provider: string]: {
type: "oauth" | "api_key";
refresh?: string;
access?: string;
expires?: number;
key?: string;
};
}
async function main() {
const modelId = process.argv[2] || "duo-chat-sonnet-4-5";
if (!MODEL_MAPPINGS[modelId]) {
console.error(`Unknown model: ${modelId}`);
console.error("Available models:", Object.keys(MODEL_MAPPINGS).join(", "));
process.exit(1);
}
// Read auth.json
const authPath = join(homedir(), ".pi", "agent", "auth.json");
console.log("Reading auth from:", authPath);
let authData: AuthData;
try {
authData = JSON.parse(readFileSync(authPath, "utf-8"));
} catch (e) {
console.error("Failed to read auth.json:", e);
process.exit(1);
}
const gitlabCred = authData["gitlab-duo"];
if (!gitlabCred || gitlabCred.type !== "oauth" || !gitlabCred.access) {
console.error("No gitlab-duo OAuth credentials found. Run /login gitlab-duo first.");
process.exit(1);
}
console.log("Found gitlab-duo OAuth credentials");
const gitlabAccessToken = gitlabCred.access;
// Register our custom API provider
console.log("\nRegistering gitlab-duo-api provider...");
registerApiProvider({
api: "gitlab-duo-api" as Api,
stream: (model, context, options) => streamGitLabDuo(model, context, options as SimpleStreamOptions),
streamSimple: streamGitLabDuo,
});
// Create a test model
const modelDef = MODEL_DEFS[modelId];
const mapping = MODEL_MAPPINGS[modelId];
// Enable reasoning for Anthropic models
const supportsReasoning = mapping.api === "anthropic-messages";
const testModel: Model<Api> = {
id: modelId,
name: modelDef.name,
api: "gitlab-duo-api" as Api,
provider: "gitlab-duo",
baseUrl: AI_GATEWAY_URL,
reasoning: supportsReasoning,
input: ["text"],
cost: modelDef.cost,
contextWindow: modelDef.contextWindow,
maxTokens: modelDef.maxTokens,
};
// Create test context
const context: Context = {
messages: [{ role: "user", content: "Say hello in exactly 3 words.", timestamp: Date.now() }],
};
// Check for --thinking flag
const useThinking = process.argv.includes("--thinking");
console.log("\nStarting stream test...");
console.log("Model:", testModel.id);
console.log("Backend:", MODEL_MAPPINGS[modelId].backendModel);
console.log("API:", MODEL_MAPPINGS[modelId].api);
console.log(
"Reasoning:",
supportsReasoning ? (useThinking ? "enabled" : "supported but not enabled") : "not supported",
);
console.log("Prompt:", context.messages[0].content);
console.log("");
// Call streamSimple
const stream = streamSimple(testModel, context, {
apiKey: gitlabAccessToken,
maxTokens: 100,
reasoning: useThinking && supportsReasoning ? "low" : undefined,
});
// Consume the stream
let inThinking = false;
for await (const event of stream) {
if (event.type === "thinking_start") {
inThinking = true;
console.log("[Thinking]");
} else if (event.type === "thinking_delta") {
process.stdout.write(event.delta);
} else if (event.type === "thinking_end") {
inThinking = false;
console.log("\n[/Thinking]\n");
} else if (event.type === "text_delta") {
process.stdout.write(event.delta);
} else if (event.type === "error") {
console.error("\nError:", event.error.errorMessage);
} else if (event.type === "done") {
console.log("\n\nDone! Stop reason:", event.reason);
console.log("Usage:", event.message.usage);
}
}
}
main().catch(console.error);