Val Town Code SearchReturn to Val Town

1# OpenAI Proxy
2
3This OpenAI API proxy injects Val Town's API keys. For usage documentation, check out https://www.val.town/v/std/openai

1import { parseBearerString } from "https://esm.town/v/andreterron/parseBearerString";
2import { API_URL } from "https://esm.town/v/std/API_URL?v=5";
3import { OpenAIUsage } from "https://esm.town/v/std/OpenAIUsage";
4import { RateLimit } from "npm:@rlimit/http";
28  const authHeader = req.headers.get("Proxy-Authorization") || req.headers.get("Authorization");
29  const token = authHeader ? parseBearerString(authHeader) : undefined;
30  const meRes = await fetch(`${API_URL}/v1/me`, { headers: { Authorization: `Bearer ${token}` } });
31  if (!meRes.ok) {
32    return new Response("Unauthorized", { status: 401 });
42
43  // Proxy the request
44  const url = new URL("." + pathname, "https://api.openai.com");
45  url.search = search;
46
47  const headers = new Headers(req.headers);
48  headers.set("Host", url.hostname);
49  headers.set("Authorization", `Bearer ${Deno.env.get("OPENAI_API_KEY")}`);
50  headers.set("OpenAI-Organization", Deno.env.get("OPENAI_API_ORG"));
51
52  const modifiedBody = await limitFreeModel(req, user);

58   * Indicates that the input or output port represents base structured
59   * datatype containing multi-part content of a message, generated by an LLM.
60   * See [Content](https://ai.google.dev/api/rest/v1beta/Content) for details
61   * on the datatype.
62   */

86    if (!inputs.$key || inputs.$key != Deno.env.get("BB_SERVICE_KEY")) {
87      return {
88        $error: "Must provide an API key to access the service.",
89      };
90    }

131
132const boardToEndpoint = (board: string) => {
133  return board.replace(/\.json$/, ".api/run");
134};
135

4
5For now, requires you to be running a board server. This harness actually acts
6as a proxy to the board server [run API endpoint](https://breadboard-ai.github.io/breadboard/docs/reference/board-run-api-endpoint/),
7and puts a nice (well, somewhat nice) frontend on top of it.
8
11
12The script will look for the `BB_LIVE_KEY` in your Val Town environment, which
13must contain your board server API key.
14
15To use, create an HTTP val, then import the `proxy` function from this script and call it like this:

4
5For now, requires you to be running a board server. This harness actually acts
6as a proxy to the board server [run API endpoint](https://breadboard-ai.github.io/breadboard/docs/reference/board-run-api-endpoint/),
7and puts a nice (well, somewhat nice) frontend on top of it.
8
11
12The script will look for the `BB_LIVE_KEY` in your Val Town environment, which
13must contain your board server API key.
14
15To use, create an HTTP val, then import the `proxy` function from this script and call it like this:

68) {
69  return async (req: Request) => {
70    const { api } = await import("https://esm.town/v/pomdtr/api");
71    const { deleteCookie, getCookies, setCookie } = await import("jsr:@std/http/cookie");
72

112  "Access-Control-Allow-Methods": "GET,HEAD,PUT,PATCH,POST,DELETE",
113  "Access-Control-Allow-Headers":
114    "Content-Type, Access-Control-Allow-Headers, api-key",
115  "Access-Control-Max-Age": "2592000", // 30 days
116} as Record<string, string>;
142
143const boardToEndpoint = (board: string) => {
144  return board.replace(/\.json$/, ".api/run");
145};
146

234        <ul>
235          <li><strong>App</strong>: Represents an OpenAI-powered application with a name and URL.</li>
236          <li><strong>Config</strong>: Defines configuration options for OpenAI API integration and application settings.</li>
237          <li><strong>ConsoleLog</strong>: Captures console output from OpenAI model interactions and application processes.</li>
238          <li><strong>CronLog</strong>: Logs scheduled tasks related to OpenAI operations, such as model fine-tuning or dataset updates.</li>
239          <li><strong>HttpLog</strong>: Records HTTP requests made to and from the OpenAI API.</li>
240        </ul>
241      </div>
250            Use Case: Manage multiple AI-powered applications or services.
251            <br>
252            Example: An app named "SentimentAnalyzer" with a URL pointing to its API endpoint.
253          </dd>
254          
255          <dt>Config</dt>
256          <dd>
257            Use Case: Store OpenAI API keys, model preferences, and application settings.
258            <br>
259            Example: Configure the GPT model to use, set token limits, and specify custom domains for AI services.
264            Use Case: Debug AI model outputs and track application performance.
265            <br>
266            Example: Log completion tokens, response times, and any errors encountered during API calls.
267          </dd>
268          
276          <dt>HttpLog</dt>
277          <dd>
278            Use Case: Monitor and analyze API usage and performance.
279            <br>
280            Example: Track rate limits, response times, and payload sizes for OpenAI API calls.
281          </dd>
282        </dl>
288      <div class="collapsible-content">
289        <ul>
290          <li><strong>AI Service Management</strong>: Use the App and Config schemas to manage multiple AI services, each with its own settings and API keys.</li>
291          <li><strong>Performance Monitoring</strong>: Utilize ConsoleLog and HttpLog to track the performance of AI models and API calls, helping optimize usage and costs.</li>
292          <li><strong>Automated AI Workflows</strong>: Implement CronLog to manage and monitor automated tasks like periodic model retraining or batch processing of data through AI models.</li>
293          <li><strong>Debugging and Troubleshooting</strong>: Leverage detailed logs from ConsoleLog and HttpLog to quickly identify and resolve issues in AI-powered applications.</li>
294          <li><strong>Usage Analytics</strong>: Analyze HttpLog data to gain insights into API usage patterns, popular features, and potential areas for optimization or scaling.</li>
295        </ul>
296        <p>By implementing this schema, developers can create robust, scalable applications that effectively integrate and manage OpenAI's powerful AI capabilities while maintaining comprehensive logging and configuration control.</p>