Implement an OpenAI-compatible API proxy and user interface
Adds an API server with proxy routes for OpenAI and Anthropic clients, along with a frontend portal to interact with the API. Replit-Commit-Author: Agent Replit-Commit-Session-Id: dcd090ca-935a-42ae-8db4-503e1bbe165a Replit-Commit-Checkpoint-Type: full_checkpoint Replit-Commit-Event-Id: 61b72c06-3c4d-47af-a97e-9d527892d9c5 Replit-Commit-Screenshot-Url: https://storage.googleapis.com/screenshot-production-us-central1/40608c50-a053-47e8-aa03-1bdcc1a2698b/dcd090ca-935a-42ae-8db4-503e1bbe165a/ZBSB1hW Replit-Helium-Checkpoint-Created: true
This commit is contained in:
@@ -2,6 +2,7 @@ import express, { type Express } from "express";
|
||||
import cors from "cors";
|
||||
import pinoHttp from "pino-http";
|
||||
import router from "./routes";
|
||||
import proxyRouter from "./routes/proxy";
|
||||
import { logger } from "./lib/logger";
|
||||
|
||||
const app: Express = express();
|
||||
@@ -26,9 +27,10 @@ app.use(
|
||||
}),
|
||||
);
|
||||
app.use(cors());
|
||||
app.use(express.json());
|
||||
app.use(express.urlencoded({ extended: true }));
|
||||
app.use(express.json({ limit: "50mb" }));
|
||||
app.use(express.urlencoded({ extended: true, limit: "50mb" }));
|
||||
|
||||
app.use("/api", router);
|
||||
app.use("/v1", proxyRouter);
|
||||
|
||||
export default app;
|
||||
|
||||
@@ -0,0 +1,443 @@
|
||||
import { Router, type IRouter, type Request, type Response } from "express";
|
||||
import OpenAI from "openai";
|
||||
import Anthropic from "@anthropic-ai/sdk";
|
||||
|
||||
const router: IRouter = Router();
|
||||
|
||||
const openai = new OpenAI({
|
||||
baseURL: process.env["AI_INTEGRATIONS_OPENAI_BASE_URL"],
|
||||
apiKey: process.env["AI_INTEGRATIONS_OPENAI_API_KEY"],
|
||||
});
|
||||
|
||||
const anthropic = new Anthropic({
|
||||
baseURL: process.env["AI_INTEGRATIONS_ANTHROPIC_BASE_URL"],
|
||||
apiKey: process.env["AI_INTEGRATIONS_ANTHROPIC_API_KEY"],
|
||||
});
|
||||
|
||||
const MODELS = [
|
||||
{ id: "gpt-5.2", provider: "openai" },
|
||||
{ id: "gpt-5-mini", provider: "openai" },
|
||||
{ id: "gpt-5-nano", provider: "openai" },
|
||||
{ id: "o4-mini", provider: "openai" },
|
||||
{ id: "o3", provider: "openai" },
|
||||
{ id: "claude-opus-4-6", provider: "anthropic" },
|
||||
{ id: "claude-sonnet-4-6", provider: "anthropic" },
|
||||
{ id: "claude-haiku-4-5", provider: "anthropic" },
|
||||
] as const;
|
||||
|
||||
function authorize(req: Request, res: Response): boolean {
|
||||
const expected = process.env["PROXY_API_KEY"];
|
||||
if (!expected) {
|
||||
res.status(500).json({ error: { message: "PROXY_API_KEY not configured", type: "server_error" } });
|
||||
return false;
|
||||
}
|
||||
const header = req.header("authorization") ?? "";
|
||||
const match = header.match(/^Bearer\s+(.+)$/i);
|
||||
if (!match || match[1] !== expected) {
|
||||
res.status(401).json({ error: { message: "Invalid API key", type: "invalid_request_error" } });
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
router.get("/models", (req, res) => {
|
||||
if (!authorize(req, res)) return;
|
||||
const created = Math.floor(Date.now() / 1000);
|
||||
res.json({
|
||||
object: "list",
|
||||
data: MODELS.map((m) => ({
|
||||
id: m.id,
|
||||
object: "model",
|
||||
created,
|
||||
owned_by: m.provider,
|
||||
})),
|
||||
});
|
||||
});
|
||||
|
||||
interface ChatMessage {
|
||||
role: "system" | "user" | "assistant" | "developer" | "tool";
|
||||
content: unknown;
|
||||
name?: string;
|
||||
}
|
||||
|
||||
interface ChatCompletionRequest {
|
||||
model: string;
|
||||
messages: ChatMessage[];
|
||||
stream?: boolean;
|
||||
temperature?: number;
|
||||
top_p?: number;
|
||||
max_tokens?: number;
|
||||
max_completion_tokens?: number;
|
||||
stop?: string | string[];
|
||||
frequency_penalty?: number;
|
||||
presence_penalty?: number;
|
||||
user?: string;
|
||||
[key: string]: unknown;
|
||||
}
|
||||
|
||||
function isOpenAIModel(model: string): boolean {
|
||||
return /^(gpt|o\d|chatgpt)/i.test(model);
|
||||
}
|
||||
|
||||
function isAnthropicModel(model: string): boolean {
|
||||
return /^claude/i.test(model);
|
||||
}
|
||||
|
||||
function newId(): string {
|
||||
return "chatcmpl-" + Math.random().toString(36).slice(2, 14);
|
||||
}
|
||||
|
||||
router.post("/chat/completions", async (req, res) => {
|
||||
if (!authorize(req, res)) return;
|
||||
|
||||
const body = req.body as ChatCompletionRequest;
|
||||
if (!body || typeof body !== "object" || !body.model || !Array.isArray(body.messages)) {
|
||||
res.status(400).json({
|
||||
error: { message: "model and messages are required", type: "invalid_request_error" },
|
||||
});
|
||||
return;
|
||||
}
|
||||
|
||||
const stream = body.stream === true;
|
||||
|
||||
if (isOpenAIModel(body.model)) {
|
||||
await handleOpenAI(req, res, body, stream);
|
||||
} else if (isAnthropicModel(body.model)) {
|
||||
await handleAnthropic(req, res, body, stream);
|
||||
} else {
|
||||
res.status(400).json({
|
||||
error: {
|
||||
message: `Unsupported model: ${body.model}`,
|
||||
type: "invalid_request_error",
|
||||
},
|
||||
});
|
||||
}
|
||||
});
|
||||
|
||||
function setupSse(res: Response): NodeJS.Timeout {
|
||||
res.setHeader("Content-Type", "text/event-stream");
|
||||
res.setHeader("Cache-Control", "no-cache, no-transform");
|
||||
res.setHeader("Connection", "keep-alive");
|
||||
res.setHeader("X-Accel-Buffering", "no");
|
||||
res.flushHeaders();
|
||||
const keepalive = setInterval(() => {
|
||||
try {
|
||||
res.write(": keepalive\n\n");
|
||||
res.flush?.();
|
||||
} catch {
|
||||
// ignore
|
||||
}
|
||||
}, 5000);
|
||||
return keepalive;
|
||||
}
|
||||
|
||||
async function handleOpenAI(
|
||||
req: Request,
|
||||
res: Response,
|
||||
body: ChatCompletionRequest,
|
||||
stream: boolean,
|
||||
): Promise<void> {
|
||||
const { stream: _ignored, ...rest } = body;
|
||||
|
||||
if (!stream) {
|
||||
try {
|
||||
const completion = await openai.chat.completions.create({
|
||||
...rest,
|
||||
stream: false,
|
||||
} as Parameters<typeof openai.chat.completions.create>[0]);
|
||||
res.json(completion);
|
||||
} catch (err) {
|
||||
handleProviderError(req, res, err, "openai");
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
let keepalive: NodeJS.Timeout | undefined;
|
||||
let closed = false;
|
||||
req.on("close", () => {
|
||||
closed = true;
|
||||
if (keepalive) clearInterval(keepalive);
|
||||
});
|
||||
|
||||
try {
|
||||
const upstream = await openai.chat.completions.create({
|
||||
...rest,
|
||||
stream: true,
|
||||
} as Parameters<typeof openai.chat.completions.create>[0]);
|
||||
|
||||
keepalive = setupSse(res);
|
||||
|
||||
for await (const chunk of upstream as AsyncIterable<unknown>) {
|
||||
if (closed) break;
|
||||
res.write(`data: ${JSON.stringify(chunk)}\n\n`);
|
||||
res.flush?.();
|
||||
}
|
||||
|
||||
if (!closed) {
|
||||
res.write("data: [DONE]\n\n");
|
||||
res.flush?.();
|
||||
}
|
||||
} catch (err) {
|
||||
if (!res.headersSent) {
|
||||
handleProviderError(req, res, err, "openai");
|
||||
return;
|
||||
}
|
||||
req.log?.error({ err }, "openai stream error");
|
||||
try {
|
||||
res.write(
|
||||
`data: ${JSON.stringify({ error: { message: errorMessage(err), type: "upstream_error" } })}\n\n`,
|
||||
);
|
||||
res.flush?.();
|
||||
} catch {
|
||||
// ignore
|
||||
}
|
||||
} finally {
|
||||
if (keepalive) clearInterval(keepalive);
|
||||
if (!closed) {
|
||||
try {
|
||||
res.end();
|
||||
} catch {
|
||||
// ignore
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
interface AnthropicConvertedMessages {
|
||||
system: string | undefined;
|
||||
messages: Array<{ role: "user" | "assistant"; content: unknown }>;
|
||||
}
|
||||
|
||||
function convertMessagesForAnthropic(messages: ChatMessage[]): AnthropicConvertedMessages {
|
||||
const systemParts: string[] = [];
|
||||
const out: Array<{ role: "user" | "assistant"; content: unknown }> = [];
|
||||
for (const msg of messages) {
|
||||
if (msg.role === "system" || msg.role === "developer") {
|
||||
if (typeof msg.content === "string") {
|
||||
systemParts.push(msg.content);
|
||||
} else if (Array.isArray(msg.content)) {
|
||||
for (const part of msg.content as Array<{ type?: string; text?: string }>) {
|
||||
if (part?.type === "text" && typeof part.text === "string") {
|
||||
systemParts.push(part.text);
|
||||
}
|
||||
}
|
||||
}
|
||||
} else if (msg.role === "user" || msg.role === "assistant") {
|
||||
out.push({ role: msg.role, content: msg.content });
|
||||
}
|
||||
// ignore tool messages for this passthrough
|
||||
}
|
||||
return {
|
||||
system: systemParts.length > 0 ? systemParts.join("\n\n") : undefined,
|
||||
messages: out,
|
||||
};
|
||||
}
|
||||
|
||||
async function handleAnthropic(
|
||||
req: Request,
|
||||
res: Response,
|
||||
body: ChatCompletionRequest,
|
||||
stream: boolean,
|
||||
): Promise<void> {
|
||||
const { system, messages } = convertMessagesForAnthropic(body.messages);
|
||||
const maxTokens =
|
||||
body.max_completion_tokens ??
|
||||
body.max_tokens ??
|
||||
8192;
|
||||
|
||||
const baseParams: Record<string, unknown> = {
|
||||
model: body.model,
|
||||
max_tokens: maxTokens,
|
||||
messages,
|
||||
};
|
||||
if (system) baseParams["system"] = system;
|
||||
if (typeof body.temperature === "number") baseParams["temperature"] = body.temperature;
|
||||
if (typeof body.top_p === "number") baseParams["top_p"] = body.top_p;
|
||||
if (body.stop !== undefined) {
|
||||
baseParams["stop_sequences"] = Array.isArray(body.stop) ? body.stop : [body.stop];
|
||||
}
|
||||
|
||||
if (!stream) {
|
||||
try {
|
||||
const message = await anthropic.messages.create(
|
||||
baseParams as Parameters<typeof anthropic.messages.create>[0],
|
||||
);
|
||||
const text = (message.content ?? [])
|
||||
.map((b) => (b.type === "text" ? b.text : ""))
|
||||
.join("");
|
||||
const finishReason = mapAnthropicStopReason(message.stop_reason ?? null);
|
||||
res.json({
|
||||
id: message.id,
|
||||
object: "chat.completion",
|
||||
created: Math.floor(Date.now() / 1000),
|
||||
model: message.model,
|
||||
choices: [
|
||||
{
|
||||
index: 0,
|
||||
message: { role: "assistant", content: text },
|
||||
finish_reason: finishReason,
|
||||
},
|
||||
],
|
||||
usage: {
|
||||
prompt_tokens: message.usage?.input_tokens ?? 0,
|
||||
completion_tokens: message.usage?.output_tokens ?? 0,
|
||||
total_tokens:
|
||||
(message.usage?.input_tokens ?? 0) + (message.usage?.output_tokens ?? 0),
|
||||
},
|
||||
});
|
||||
} catch (err) {
|
||||
handleProviderError(req, res, err, "anthropic");
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
let keepalive: NodeJS.Timeout | undefined;
|
||||
let closed = false;
|
||||
req.on("close", () => {
|
||||
closed = true;
|
||||
if (keepalive) clearInterval(keepalive);
|
||||
});
|
||||
|
||||
const id = newId();
|
||||
const created = Math.floor(Date.now() / 1000);
|
||||
|
||||
try {
|
||||
const upstream = anthropic.messages.stream(
|
||||
baseParams as Parameters<typeof anthropic.messages.stream>[0],
|
||||
);
|
||||
|
||||
keepalive = setupSse(res);
|
||||
|
||||
res.write(
|
||||
`data: ${JSON.stringify({
|
||||
id,
|
||||
object: "chat.completion.chunk",
|
||||
created,
|
||||
model: body.model,
|
||||
choices: [{ index: 0, delta: { role: "assistant" }, finish_reason: null }],
|
||||
})}\n\n`,
|
||||
);
|
||||
res.flush?.();
|
||||
|
||||
let stopReason: string | null = null;
|
||||
|
||||
for await (const event of upstream) {
|
||||
if (closed) break;
|
||||
if (
|
||||
event.type === "content_block_delta" &&
|
||||
event.delta.type === "text_delta"
|
||||
) {
|
||||
const chunk = {
|
||||
id,
|
||||
object: "chat.completion.chunk",
|
||||
created,
|
||||
model: body.model,
|
||||
choices: [
|
||||
{
|
||||
index: 0,
|
||||
delta: { content: event.delta.text },
|
||||
finish_reason: null,
|
||||
},
|
||||
],
|
||||
};
|
||||
res.write(`data: ${JSON.stringify(chunk)}\n\n`);
|
||||
res.flush?.();
|
||||
} else if (event.type === "message_delta") {
|
||||
const delta = event as unknown as { delta?: { stop_reason?: string | null } };
|
||||
if (delta.delta?.stop_reason) {
|
||||
stopReason = delta.delta.stop_reason;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (!closed) {
|
||||
const finishChunk = {
|
||||
id,
|
||||
object: "chat.completion.chunk",
|
||||
created,
|
||||
model: body.model,
|
||||
choices: [
|
||||
{
|
||||
index: 0,
|
||||
delta: {},
|
||||
finish_reason: mapAnthropicStopReason(stopReason),
|
||||
},
|
||||
],
|
||||
};
|
||||
res.write(`data: ${JSON.stringify(finishChunk)}\n\n`);
|
||||
res.write("data: [DONE]\n\n");
|
||||
res.flush?.();
|
||||
}
|
||||
} catch (err) {
|
||||
if (!res.headersSent) {
|
||||
handleProviderError(req, res, err, "anthropic");
|
||||
return;
|
||||
}
|
||||
req.log?.error({ err }, "anthropic stream error");
|
||||
try {
|
||||
res.write(
|
||||
`data: ${JSON.stringify({ error: { message: errorMessage(err), type: "upstream_error" } })}\n\n`,
|
||||
);
|
||||
res.flush?.();
|
||||
} catch {
|
||||
// ignore
|
||||
}
|
||||
} finally {
|
||||
if (keepalive) clearInterval(keepalive);
|
||||
if (!closed) {
|
||||
try {
|
||||
res.end();
|
||||
} catch {
|
||||
// ignore
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
function mapAnthropicStopReason(reason: string | null): string {
|
||||
switch (reason) {
|
||||
case "end_turn":
|
||||
case "stop_sequence":
|
||||
return "stop";
|
||||
case "max_tokens":
|
||||
return "length";
|
||||
case "tool_use":
|
||||
return "tool_calls";
|
||||
default:
|
||||
return "stop";
|
||||
}
|
||||
}
|
||||
|
||||
function errorMessage(err: unknown): string {
|
||||
if (err instanceof Error) return err.message;
|
||||
if (typeof err === "string") return err;
|
||||
try {
|
||||
return JSON.stringify(err);
|
||||
} catch {
|
||||
return "unknown error";
|
||||
}
|
||||
}
|
||||
|
||||
function handleProviderError(
|
||||
req: Request,
|
||||
res: Response,
|
||||
err: unknown,
|
||||
provider: string,
|
||||
): void {
|
||||
req.log?.error({ err, provider }, "provider error");
|
||||
const status =
|
||||
typeof (err as { status?: number })?.status === "number"
|
||||
? (err as { status: number }).status
|
||||
: 500;
|
||||
if (!res.headersSent) {
|
||||
res.status(status).json({
|
||||
error: {
|
||||
message: errorMessage(err),
|
||||
type: "upstream_error",
|
||||
provider,
|
||||
},
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
export default router;
|
||||
Reference in New Issue
Block a user