@leoustc/service-llm 0.1.0 → 0.2.1

package/LICENSE

@@ -0,0 +1,201 @@
+Apache License
+Version 2.0, January 2004
+http://www.apache.org/licenses/
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+1. Definitions.
+
+"License" shall mean the terms and conditions for use, reproduction,
+and distribution as defined by Sections 1 through 9 of this document.
+
+"Licensor" shall mean the copyright owner or entity authorized by
+the copyright owner that is granting the License.
+
+"Legal Entity" shall mean the union of the acting entity and all
+other entities that control, are controlled by, or are under common
+control with that entity. For the purposes of this definition,
+"control" means (i) the power, direct or indirect, to cause the
+direction or management of such entity, whether by contract or
+otherwise, or (ii) ownership of fifty percent (50%) or more of the
+outstanding shares, or (iii) beneficial ownership of such entity.
+
+"You" (or "Your") shall mean an individual or Legal Entity
+exercising permissions granted by this License.
+
+"Source" form shall mean the preferred form for making modifications,
+including but not limited to software source code, documentation
+source, and configuration files.
+
+"Object" form shall mean any form resulting from mechanical
+transformation or translation of a Source form, including but
+not limited to compiled object code, generated documentation,
+and conversions to other media types.
+
+"Work" shall mean the work of authorship, whether in Source or
+Object form, made available under the License, as indicated by a
+copyright notice that is included in or attached to the work
+(an example is provided in the Appendix below).
+
+"Derivative Works" shall mean any work, whether in Source or Object
+form, that is based on (or derived from) the Work and for which the
+editorial revisions, annotations, elaborations, or other modifications
+represent, as a whole, an original work of authorship. For the purposes
+of this License, Derivative Works shall not include works that remain
+separable from, or merely link (or bind by name) to the interfaces of,
+the Work and Derivative Works thereof.
+
+"Contribution" shall mean any work of authorship, including
+the original version of the Work and any modifications or additions
+to that Work or Derivative Works thereof, that is intentionally
+submitted to Licensor for inclusion in the Work by the copyright owner
+or by an individual or Legal Entity authorized to submit on behalf of
+the copyright owner. For the purposes of this definition, "submitted"
+means any form of electronic, verbal, or written communication sent
+to the Licensor or its representatives, including but not limited to
+communication on electronic mailing lists, source code control systems,
+and issue tracking systems that are managed by, or on behalf of, the
+Licensor for the purpose of discussing and improving the Work, but
+excluding communication that is conspicuously marked or otherwise
+designated in writing by the copyright owner as "Not a Contribution."
+
+"Contributor" shall mean Licensor and any individual or Legal Entity
+on behalf of whom a Contribution has been received by Licensor and
+subsequently incorporated within the Work.
+
+2. Grant of Copyright License. Subject to the terms and conditions of
+this License, each Contributor hereby grants to You a perpetual,
+worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+copyright license to reproduce, prepare Derivative Works of,
+publicly display, publicly perform, sublicense, and distribute the
+Work and such Derivative Works in Source or Object form.
+
+3. Grant of Patent License. Subject to the terms and conditions of
+this License, each Contributor hereby grants to You a perpetual,
+worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+(except as stated in this section) patent license to make, have made,
+use, offer to sell, sell, import, and otherwise transfer the Work,
+where such license applies only to those patent claims licensable
+by such Contributor that are necessarily infringed by their
+Contribution(s) alone or by combination of their Contribution(s)
+with the Work to which such Contribution(s) was submitted. If You
+institute patent litigation against any entity (including a
+cross-claim or counterclaim in a lawsuit) alleging that the Work
+or a Contribution incorporated within the Work constitutes direct
+or contributory patent infringement, then any patent licenses
+granted to You under this License for that Work shall terminate
+as of the date such litigation is filed.
+
+4. Redistribution. You may reproduce and distribute copies of the
+Work or Derivative Works thereof in any medium, with or without
+modifications, and in Source or Object form, provided that You
+meet the following conditions:
+
+(a) You must give any other recipients of the Work or
+    Derivative Works a copy of this License; and
+
+(b) You must cause any modified files to carry prominent notices
+    stating that You changed the files; and
+
+(c) You must retain, in the Source form of any Derivative Works
+    that You distribute, all copyright, patent, trademark, and
+    attribution notices from the Source form of the Work,
+    excluding those notices that do not pertain to any part of
+    the Derivative Works; and
+
+(d) If the Work includes a "NOTICE" text file as part of its
+    distribution, then any Derivative Works that You distribute must
+    include a readable copy of the attribution notices contained
+    within such NOTICE file, excluding those notices that do not
+    pertain to any part of the Derivative Works, in at least one
+    of the following places: within a NOTICE text file distributed
+    as part of the Derivative Works; within the Source form or
+    documentation, if provided along with the Derivative Works; or,
+    within a display generated by the Derivative Works, if and
+    wherever such third-party notices normally appear. The contents
+    of the NOTICE file are for informational purposes only and
+    do not modify the License. You may add Your own attribution
+    notices within Derivative Works that You distribute, alongside
+    or as an addendum to the NOTICE text from the Work, provided
+    that such additional attribution notices cannot be construed
+    as modifying the License.
+
+You may add Your own copyright statement to Your modifications and
+may provide additional or different license terms and conditions
+for use, reproduction, or distribution of Your modifications, or
+for any such Derivative Works as a whole, provided Your use,
+reproduction, and distribution of the Work otherwise complies with
+the conditions stated in this License.
+
+5. Submission of Contributions. Unless You explicitly state otherwise,
+any Contribution intentionally submitted for inclusion in the Work
+by You to the Licensor shall be under the terms and conditions of
+this License, without any additional terms or conditions.
+Notwithstanding the above, nothing herein shall supersede or modify
+the terms of any separate license agreement you may have executed
+with Licensor regarding such Contributions.
+
+6. Trademarks. This License does not grant permission to use the trade
+names, trademarks, service marks, or product names of the Licensor,
+except as required for reasonable and customary use in describing the
+origin of the Work and reproducing the content of the NOTICE file.
+
+7. Disclaimer of Warranty. Unless required by applicable law or
+agreed to in writing, Licensor provides the Work (and each
+Contributor provides its Contributions) on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+implied, including, without limitation, any warranties or conditions
+of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+PARTICULAR PURPOSE. You are solely responsible for determining the
+appropriateness of using or redistributing the Work and assume any
+risks associated with Your exercise of permissions under this License.
+
+8. Limitation of Liability. In no event and under no legal theory,
+whether in tort (including negligence), contract, or otherwise,
+unless required by applicable law (such as deliberate and grossly
+negligent acts) or agreed to in writing, shall any Contributor be
+liable to You for damages, including any direct, indirect, special,
+incidental, or consequential damages of any character arising as a
+result of this License or out of the use or inability to use the
+Work (including but not limited to damages for loss of goodwill,
+work stoppage, computer failure or malfunction, or any and all
+other commercial damages or losses), even if such Contributor
+has been advised of the possibility of such damages.
+
+9. Accepting Warranty or Additional Liability. While redistributing
+the Work or Derivative Works thereof, You may choose to offer,
+and charge a fee for, acceptance of support, warranty, indemnity,
+or other liability obligations and/or rights consistent with this
+License. However, in accepting such obligations, You may act only
+on Your own behalf and on Your sole responsibility, not on behalf
+of any other Contributor, and only if You agree to indemnify,
+defend, and hold each Contributor harmless for any liability
+incurred by, or claims asserted against, such Contributor by reason
+of your accepting any such warranty or additional liability.
+
+END OF TERMS AND CONDITIONS
+
+APPENDIX: How to apply the Apache License to your work.
+
+To apply the Apache License to your work, attach the following
+boilerplate notice, with the fields enclosed by brackets "[]"
+replaced with your own identifying information. (Don't include
+the brackets!)  The text should be enclosed in the appropriate
+comment syntax for the file format. We also recommend that a
+file or class name and description of purpose be included on the
+same "printed page" as the copyright notice for easier
+identification within third-party archives.
+
+Copyright [yyyy] [name of copyright owner]
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

package/README.md

@@ -1,77 +1,195 @@
-# @leoustc/service-llm
+# service-llm
 
-Expose an OpenAI Codex subscription through OpenAI-compatible HTTP APIs and MCP.
+Use your ChatGPT Codex subscription through OpenAI-compatible HTTP APIs or MCP.
 
-The service never executes model tool calls. It can return tool-call output to a client, which remains responsible for deciding whether to execute it.
+`service-llm` supports Chat Completions, Responses, SSE streaming, function calls, and an MCP `ask` tool. It returns model tool calls but never executes them.
 
 ## Requirements
 
 - Node.js 22.19 or newer
-- A ChatGPT Plus or Pro account with Codex access
+- A ChatGPT account with Codex access
 
-## Install and log in
+## Install
 
 ```sh
 npm install --global @leoustc/service-llm
+```
+
+## Log in
+
+```sh
 service-llm login
 ```
 
-For a headless machine, use the device-code flow:
+For a headless or remote machine:
 
 ```sh
 service-llm login --device-code
 ```
 
-OAuth credentials are stored at `~/.service-llm/auth.json` with owner-only permissions.
+Credentials are stored securely at `~/.service-llm/auth.json`. The available Codex models are recorded at `~/.service-llm/models.json`.
+
+If you start the service before logging in, the first model request returns a device code and verification URL. Complete the browser login, then retry the request.
+
+## Choose a model
+
+List supported models and show the current default:
+
+```sh
+service-llm model
+```
+
+The default is marked with `*`. Set a different default with:
+
+```sh
+service-llm model set gpt-5.4-mini
+```
+
+An explicit valid `model` in an API request overrides this default for that request.
+
+## Start the server
+
+```sh
+service-llm serve
+```
+
+The server listens on `127.0.0.1:7060` by default.
+
+At startup it prints the version, listening URL, authentication state, default model, supported-model count, endpoint URLs, and debug status. API key values and OAuth credentials are never displayed.
+
+Choose a different address or port:
+
+```sh
+service-llm serve --host 0.0.0.0 --port 8080
+```
+
+Protect the API with one or more client keys:
 
-If an API request arrives before login, the service starts a device-code login and returns HTTP 200 with `status: "ok"`, a friendly instruction, `device_code`, and `verification_uri`. Complete authentication in the browser, then retry the request.
+```sh
+service-llm serve --api-key secret-one,secret-two
+```
+
+Clients may send a key through either header:
+
+```http
+Authorization: Bearer secret-one
+```
+
+```http
+X-API-Key: secret-one
+```
+
+Do not expose the server on a public interface without authentication and appropriate network controls.
+
+## Chat Completions
+
+```sh
+curl http://127.0.0.1:7060/v1/chat/completions \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "model": "gpt-5.4",
+    "messages": [
+      {"role": "user", "content": "Write a short story."}
+    ]
+  }'
+```
+
+Add `"stream": true` to receive OpenAI-compatible SSE chunks.
 
-## Start the service
+## Responses API
 
 ```sh
-service-llm --port 7060 --api-key secret-one,secret-two
+curl http://127.0.0.1:7060/v1/responses \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "model": "gpt-5.4",
+    "input": "Write a short story."
+  }'
 ```
 
-The default bind address is `127.0.0.1`. Use `--host 0.0.0.0` to listen on all interfaces. If no API keys are supplied, endpoints are unauthenticated.
+Add `"stream": true` to receive typed Responses SSE events.
 
-Clients can authenticate with either `Authorization: Bearer <key>` or `X-API-Key: <key>`.
+## Function calls
 
-From a source checkout, build and run with request/response debug logging enabled:
+Supply OpenAI-compatible tools and `tool_choice`. The model may return a function call, but `service-llm` will not execute it.
 
 ```sh
-make run
-make run PORT=8080 HOST=0.0.0.0 API_KEYS=secret-one,secret-two
+curl http://127.0.0.1:7060/v1/responses \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "model": "gpt-5.4",
+    "input": "Write a short story and submit it with save_story.",
+    "tools": [{
+      "type": "function",
+      "name": "save_story",
+      "description": "Save a generated story",
+      "parameters": {
+        "type": "object",
+        "properties": {
+          "title": {"type": "string"},
+          "story": {"type": "string"}
+        },
+        "required": ["title", "story"]
+      }
+    }],
+    "tool_choice": "required"
+  }'
 ```
 
-Debug output includes request bodies, JSON responses, and individual SSE events. Authentication headers are redacted. Because prompts and model output are logged, do not enable debug logging around sensitive content.
+## MCP
 
-Install the current checkout to `~/.bin` without root permissions:
+Connect an MCP client to:
+
+```text
+http://127.0.0.1:7060/mcp
+```
+
+The server provides one tool:
+
+- `ask`: accepts `prompt` and optional `model`, `tools`, and `tool_choice` arguments.
+
+Example JSON-RPC request:
 
 ```sh
-make install
-service-llm --help
+curl http://127.0.0.1:7060/mcp \
+  -H 'Content-Type: application/json' \
+  -d '{
+    "jsonrpc": "2.0",
+    "id": 1,
+    "method": "tools/call",
+    "params": {
+      "name": "ask",
+      "arguments": {"prompt": "Write a short story."}
+    }
+  }'
 ```
 
-If necessary, add it to your shell path with `export PATH="$HOME/.bin:$PATH"`. Run `make uninstall` to remove the symlink.
+## Model selection
 
-## Endpoints
+Specify a model with the request's `model` field. If it is omitted or unavailable, the service uses your configured default. On first login, the lowest-cost available model is selected with low reasoning effort. Responses always report the model actually used.
 
-- `GET /health`
-- `POST /v1/chat/completions`, including `stream: true` SSE responses
-- `POST /v1/responses`, including typed SSE responses
-- `POST /mcp`, providing the MCP `ask` tool. `ask` accepts a prompt plus optional `model`, `tools`, and `tool_choice`, and returns model tool calls without executing them.
+## Debug logging
 
-The default model is `gpt-5.4`. A different Codex model can be selected with the request's `model` field.
+Enable request, response, and SSE logs with:
 
-## Development
+```sh
+DEBUG=service-llm service-llm serve
+```
+
+Authentication headers are redacted. Prompts and model output are not redacted, so avoid debug logging for sensitive requests.
+
+## Health check
 
 ```sh
-make build
-make test
-make pack
-make publish
-make clean
+curl http://127.0.0.1:7060/health
+```
+
+Expected response:
+
+```json
+{"status":"ok"}
 ```
 
-Each test run also saves its complete output under `test/logs/` using a timestamped filename.
-The suite sends the live `Write a short story.` prompt through Chat Completions, Responses, and MCP `ask`. Override the HTTP request model with `make test TEST_MODEL=gpt-5.4-mini`.
+## License
+
+Apache License 2.0. See [LICENSE](LICENSE).

package/VERSION

@@ -0,0 +1 @@
+0.2.1

package/dist/auth.js

@@ -1,4 +1,5 @@
 import { chmod, mkdir, readFile, rename, writeFile } from "node:fs/promises";
+import { randomUUID } from "node:crypto";
 import { homedir } from "node:os";
 import { dirname, join } from "node:path";
 import {
@@ -6,16 +7,77 @@ import {
   loginOpenAICodexDeviceCode,
   refreshOpenAICodexToken,
 } from "@earendil-works/pi-ai/oauth";
+import { getModels } from "@earendil-works/pi-ai/compat";
 
 export const authPath = join(homedir(), ".service-llm", "auth.json");
+export const modelsPath = join(homedir(), ".service-llm", "models.json");
 let pendingDeviceLogin;
+let refreshPromise;
+let writeQueue = Promise.resolve();
+
+function smallestModel(models) {
+  return [...models].sort((left, right) => {
+    const leftCost = (left.cost?.input ?? Infinity) + (left.cost?.output ?? Infinity);
+    const rightCost = (right.cost?.input ?? Infinity) + (right.cost?.output ?? Infinity);
+    return leftCost - rightCost || left.id.localeCompare(right.id);
+  })[0];
+}
+
+async function saveModelCatalog() {
+  const models = getModels("openai-codex");
+  const fallback = smallestModel(models);
+  if (!fallback) throw new Error("No OpenAI Codex models are available");
+  let previousDefault;
+  try {
+    previousDefault = JSON.parse(await readFile(modelsPath, "utf8")).default_model;
+  } catch (error) {
+    if (error?.code !== "ENOENT") console.error(`Unable to preserve model selection: ${error.message}`);
+  }
+  const availableIds = new Set(models.map((model) => model.id));
+  const catalog = {
+    updated_at: new Date().toISOString(),
+    default_model: availableIds.has(previousDefault) ? previousDefault : fallback.id,
+    default_reasoning_effort: "low",
+    models: models.map((model) => ({ id: model.id, name: model.name })),
+  };
+  await writeSecureJson(modelsPath, catalog);
+  return catalog;
+}
+
+export async function listModels() {
+  try {
+    return JSON.parse(await readFile(modelsPath, "utf8"));
+  } catch (error) {
+    if (error?.code !== "ENOENT") throw new Error(`Unable to read ${modelsPath}: ${error.message}`);
+    return saveModelCatalog();
+  }
+}
+
+export async function setDefaultModel(modelId) {
+  const catalog = await listModels();
+  if (!catalog.models.some((model) => model.id === modelId)) {
+    throw new Error(`Unsupported model: ${modelId}. Run 'service-llm model' to list supported models.`);
+  }
+  catalog.default_model = modelId;
+  catalog.updated_at = new Date().toISOString();
+  await writeSecureJson(modelsPath, catalog);
+  return catalog;
+}
+
+async function writeSecureJson(path, data) {
+  const operation = writeQueue.then(async () => {
+    await mkdir(dirname(path), { recursive: true, mode: 0o700 });
+    const temporaryPath = `${path}.${process.pid}.${randomUUID()}.tmp`;
+    await writeFile(temporaryPath, `${JSON.stringify(data, null, 2)}\n`, { mode: 0o600 });
+    await rename(temporaryPath, path);
+    await chmod(path, 0o600);
+  });
+  writeQueue = operation.catch(() => {});
+  return operation;
+}
 
 async function saveAuth(data) {
-  await mkdir(dirname(authPath), { recursive: true, mode: 0o700 });
-  const temporaryPath = `${authPath}.${process.pid}.tmp`;
-  await writeFile(temporaryPath, `${JSON.stringify(data, null, 2)}\n`, { mode: 0o600 });
-  await rename(temporaryPath, authPath);
-  await chmod(authPath, 0o600);
+  await writeSecureJson(authPath, data);
 }
 
 async function readAuth() {
@@ -46,7 +108,9 @@ export async function login({ deviceCode = false } = {}) {
   const data = await readAuth();
   data["openai-codex"] = { type: "oauth", ...credentials };
   await saveAuth(data);
+  const catalog = await saveModelCatalog();
   console.error(`Authentication saved to ${authPath}`);
+  console.error(`Loaded ${catalog.models.length} models; default is ${catalog.default_model} with low reasoning`);
 }
 
 export class DeviceLoginRequiredError extends Error {
@@ -86,6 +150,7 @@ async function beginDeviceLogin() {
       const data = await readAuth();
       data["openai-codex"] = { type: "oauth", ...credentials };
       await saveAuth(data);
+      await saveModelCatalog();
       console.error(`Authentication completed and saved to ${authPath}`);
     })
     .catch((error) => {
@@ -107,9 +172,21 @@ export async function getAccessToken() {
   }
 
   if (Date.now() >= credentials.expires - 60_000) {
-    credentials = { type: "oauth", ...(await refreshOpenAICodexToken(credentials.refresh)) };
-    data["openai-codex"] = credentials;
-    await saveAuth(data);
+    if (!refreshPromise) {
+      refreshPromise = (async () => {
+        const latest = await readAuth();
+        const latestCredentials = latest["openai-codex"];
+        if (!latestCredentials?.refresh) throw new Error("Stored OpenAI Codex refresh token is missing; log in again");
+        if (latestCredentials?.access && Date.now() < latestCredentials.expires - 60_000) return latestCredentials;
+        const refreshed = { type: "oauth", ...(await refreshOpenAICodexToken(latestCredentials.refresh)) };
+        latest["openai-codex"] = refreshed;
+        await saveAuth(latest);
+        return refreshed;
+      })().finally(() => {
+        refreshPromise = undefined;
+      });
+    }
+    credentials = await refreshPromise;
   }
   return credentials.access;
 }

package/dist/backend.js

@@ -1,20 +1,75 @@
-import { getModel, stream } from "@earendil-works/pi-ai/compat";
-import { getAccessToken } from "./auth.js";
+import { readFile } from "node:fs/promises";
+import { getModel, getModels, stream } from "@earendil-works/pi-ai/compat";
+import { getAccessToken, modelsPath } from "./auth.js";
 
-export const defaultModel = "gpt-5.4";
+const zeroUsage = {
+  input: 0,
+  output: 0,
+  cacheRead: 0,
+  cacheWrite: 0,
+  totalTokens: 0,
+  cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+};
+
+function smallestBuiltinModel() {
+  return [...getModels("openai-codex")].sort((left, right) => {
+    const leftCost = (left.cost?.input ?? Infinity) + (left.cost?.output ?? Infinity);
+    const rightCost = (right.cost?.input ?? Infinity) + (right.cost?.output ?? Infinity);
+    return leftCost - rightCost || left.id.localeCompare(right.id);
+  })[0];
+}
+
+export const defaultModel = smallestBuiltinModel()?.id ?? "gpt-5.4-mini";
+
+async function configuredModels() {
+  try {
+    const catalog = JSON.parse(await readFile(modelsPath, "utf8"));
+    return {
+      ids: new Set(catalog.models.map((model) => model.id)),
+      defaultModel: catalog.default_model,
+      reasoningEffort: catalog.default_reasoning_effort ?? "low",
+    };
+  } catch (error) {
+    if (error?.code !== "ENOENT") console.error(`Unable to read ${modelsPath}: ${error.message}`);
+    const models = getModels("openai-codex");
+    return { ids: new Set(models.map((model) => model.id)), defaultModel, reasoningEffort: "low" };
+  }
+}
+
+export async function resolveModel(requestedModel) {
+  const catalog = await configuredModels();
+  const selectedId = requestedModel && catalog.ids.has(requestedModel) ? requestedModel : catalog.defaultModel;
+  const model = getModel("openai-codex", selectedId);
+  if (!model) throw new Error("No usable OpenAI Codex model is available");
+  return { model, requestedModel, fallback: Boolean(requestedModel && requestedModel !== selectedId), reasoningEffort: catalog.reasoningEffort };
+}
 
 function textFromContent(content) {
   if (typeof content === "string") return content;
   if (!Array.isArray(content)) return "";
   return content
-    .filter((part) => part?.type === "text" || part?.type === "input_text")
+    .filter((part) => ["text", "input_text", "output_text"].includes(part?.type))
     .map((part) => part.text ?? "")
     .join("\n");
 }
 
-export function messagesFromChat(messages = []) {
+function assistantMessage(content, model) {
+  return {
+    role: "assistant",
+    content,
+    api: "openai-codex-responses",
+    provider: "openai-codex",
+    model,
+    usage: structuredClone(zeroUsage),
+    stopReason: content.some((part) => part.type === "toolCall") ? "toolUse" : "stop",
+    timestamp: Date.now(),
+  };
+}
+
+export function messagesFromChat(messages = [], model = defaultModel) {
   let systemPrompt;
   const result = [];
+  const toolNames = new Map();
   for (const message of messages) {
     const text = textFromContent(message.content);
     if (message.role === "system" || message.role === "developer") {
@@ -22,14 +77,27 @@ export function messagesFromChat(messages = []) {
     } else if (message.role === "user") {
       result.push({ role: "user", content: text, timestamp: Date.now() });
     } else if (message.role === "assistant") {
+      const content = text ? [{ type: "text", text }] : [];
+      for (const call of message.tool_calls ?? []) {
+        const fn = call.function;
+        if (!call.id || !fn?.name) continue;
+        let args;
+        try {
+          args = typeof fn.arguments === "string" ? JSON.parse(fn.arguments) : (fn.arguments ?? {});
+        } catch {
+          args = { raw: fn.arguments };
+        }
+        content.push({ type: "toolCall", id: call.id, name: fn.name, arguments: args });
+        toolNames.set(call.id, fn.name);
+      }
+      result.push(assistantMessage(content, model));
+    } else if (message.role === "tool") {
       result.push({
-        role: "assistant",
+        role: "toolResult",
+        toolCallId: message.tool_call_id,
+        toolName: message.name ?? toolNames.get(message.tool_call_id) ?? "tool",
         content: [{ type: "text", text }],
-        api: "openai-codex-responses",
-        provider: "openai-codex",
-        model: defaultModel,
-        usage: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, totalTokens: 0, cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 } },
-        stopReason: "stop",
+        isError: false,
         timestamp: Date.now(),
       });
     }
@@ -37,35 +105,63 @@ export function messagesFromChat(messages = []) {
   return { systemPrompt, messages: result };
 }
 
-export function messagesFromResponses(input) {
-  if (typeof input === "string") {
-    return { messages: [{ role: "user", content: input, timestamp: Date.now() }] };
+export function messagesFromResponses(input, model = defaultModel) {
+  if (typeof input === "string") return { messages: [{ role: "user", content: input, timestamp: Date.now() }] };
+  if (!Array.isArray(input)) return { messages: [] };
+
+  const result = [];
+  const toolNames = new Map();
+  let systemPrompt;
+  for (const item of input) {
+    if (item.type === "function_call") {
+      let args;
+      try {
+        args = typeof item.arguments === "string" ? JSON.parse(item.arguments) : (item.arguments ?? {});
+      } catch {
+        args = { raw: item.arguments };
+      }
+      const callId = item.call_id ?? item.id;
+      toolNames.set(callId, item.name);
+      result.push(assistantMessage([{ type: "toolCall", id: callId, name: item.name, arguments: args }], model));
+    } else if (item.type === "function_call_output") {
+      result.push({
+        role: "toolResult",
+        toolCallId: item.call_id,
+        toolName: toolNames.get(item.call_id) ?? "tool",
+        content: [{ type: "text", text: textFromContent(item.output) || String(item.output ?? "") }],
+        isError: false,
+        timestamp: Date.now(),
+      });
+    } else {
+      const converted = messagesFromChat([item], model);
+      if (converted.systemPrompt) systemPrompt = systemPrompt ? `${systemPrompt}\n\n${converted.systemPrompt}` : converted.systemPrompt;
+      result.push(...converted.messages);
+    }
   }
-  return messagesFromChat(Array.isArray(input) ? input : []);
+  return { systemPrompt, messages: result };
 }
 
 function toolsFromOpenAI(tools) {
   if (!Array.isArray(tools)) return undefined;
   return tools.flatMap((tool) => {
     const fn = tool.type === "function" ? (tool.function ?? tool) : null;
-    return fn?.name
-      ? [{ name: fn.name, description: fn.description ?? "", parameters: fn.parameters ?? { type: "object", properties: {} } }]
-      : [];
+    return fn?.name ? [{ name: fn.name, description: fn.description ?? "", parameters: fn.parameters ?? { type: "object", properties: {} } }] : [];
   });
 }
 
-export async function createCompletionStream({ context, model = defaultModel, tools, toolChoice, signal }) {
-  const selectedModel = getModel("openai-codex", model) ?? getModel("openai-codex", defaultModel);
-  if (!selectedModel) throw new Error(`Unknown Codex model: ${model}`);
+export async function createCompletionStream({ context, model, tools, toolChoice, signal }) {
+  const resolved = await resolveModel(model);
   const apiKey = await getAccessToken();
-  return stream(selectedModel, { ...context, tools: toolsFromOpenAI(tools) }, {
+  const events = stream(resolved.model, { ...context, tools: toolsFromOpenAI(tools) }, {
     apiKey,
     signal,
+    reasoningEffort: resolved.reasoningEffort,
     onPayload(payload) {
       if (!toolChoice || !payload || typeof payload !== "object") return payload;
       return { ...payload, tool_choice: toolChoice };
     },
   });
+  return { events, model: resolved.model.id, fallback: resolved.fallback, requestedModel: resolved.requestedModel };
 }
 
 export function contentFromMessage(message) {

package/dist/cli.js

@@ -1,16 +1,41 @@
 #!/usr/bin/env node
-import { login } from "./auth.js";
+import { listModels, login, setDefaultModel } from "./auth.js";
 import { startServer } from "./http.js";
 
 function usage() {
   console.error(`Usage:
   service-llm login [--device-code]
-  service-llm [--port 7060] [--host 127.0.0.1] [--api-key key1,key2]`);
+  service-llm model [list]
+  service-llm model set <model-id>
+  service-llm serve [--port 7060] [--host 127.0.0.1] [--api-key key1,key2]`);
+}
+
+async function modelCommand(args) {
+  if (!args[1] || args[1] === "list") {
+    const catalog = await listModels();
+    console.log(`Supported Codex models (${catalog.models.length}):`);
+    for (const model of catalog.models) {
+      console.log(`${model.id === catalog.default_model ? "*" : " "} ${model.id}${model.name ? ` — ${model.name}` : ""}`);
+    }
+    console.log(`\n* default (${catalog.default_reasoning_effort ?? "low"} reasoning)`);
+    return;
+  }
+  if (args[1] === "set") {
+    const modelId = args[2];
+    if (!modelId || args[3]) throw new Error("Usage: service-llm model set <model-id>");
+    await setDefaultModel(modelId);
+    console.log(`Default model set to ${modelId}`);
+    return;
+  }
+  throw new Error("Usage: service-llm model [list|set <model-id>]");
 }
 
 function valueAfter(args, names, fallback) {
   const index = args.findIndex((arg) => names.includes(arg));
-  return index < 0 ? fallback : args[index + 1];
+  if (index < 0) return fallback;
+  const value = args[index + 1];
+  if (!value || value.startsWith("-")) throw new Error(`${args[index]} requires a value`);
+  return value;
 }
 
 const args = process.argv.slice(2);
@@ -18,9 +43,15 @@ if (args.includes("--help") || args.includes("-h")) {
   usage();
 } else if (args[0] === "login") {
   await login({ deviceCode: args.includes("--device-code") });
-} else {
-  const port = Number(valueAfter(args, ["--port"], "7060"));
+} else if (args[0] === "model") {
+  await modelCommand(args);
+} else if (args[0] === "serve") {
+  const serveArgs = args.slice(1);
+  const port = Number(valueAfter(serveArgs, ["--port"], "7060"));
   if (!Number.isInteger(port) || port < 1 || port > 65535) throw new Error("--port must be between 1 and 65535");
-  const apiKeyValue = valueAfter(args, ["--api-key", "-api_key"], process.env.SERVICE_LLM_API_KEYS ?? "");
-  startServer({ port, host: valueAfter(args, ["--host"], "127.0.0.1"), apiKeys: apiKeyValue.split(",").map((key) => key.trim()) });
+  const apiKeyValue = valueAfter(serveArgs, ["--api-key", "-api_key"], process.env.SERVICE_LLM_API_KEYS ?? "");
+  startServer({ port, host: valueAfter(serveArgs, ["--host"], "127.0.0.1"), apiKeys: apiKeyValue.split(",").map((key) => key.trim()) });
+} else {
+  usage();
+  process.exitCode = args.length ? 1 : 0;
 }

package/dist/http.js

@@ -1,6 +1,7 @@
 import { createServer } from "node:http";
 import { randomUUID } from "node:crypto";
-import { DeviceLoginRequiredError } from "./auth.js";
+import { DeviceLoginRequiredError, listModels } from "./auth.js";
+import { version } from "./version.js";
 import {
   contentFromMessage,
   createCompletionStream,
@@ -47,9 +48,9 @@ function isAuthorized(request, apiKeys) {
   return typeof token === "string" && apiKeys.has(token);
 }
 
-function sse(response, event) {
+function sse(response, event, named = false) {
   debug("response SSE", event);
-  response.write(`data: ${JSON.stringify(event)}\n\n`);
+  response.write(`${named && event.type ? `event: ${event.type}\n` : ""}data: ${JSON.stringify(event)}\n\n`);
 }
 
 async function collect(streamResult) {
@@ -64,18 +65,23 @@ async function collect(streamResult) {
 
 async function chatCompletions(request, response, body) {
   const id = `chatcmpl-${randomUUID().replaceAll("-", "")}`;
-  const model = body.model || defaultModel;
-  const streamResult = await createCompletionStream({
-    context: messagesFromChat(body.messages), model, tools: body.tools, toolChoice: body.tool_choice,
+  const completion = await createCompletionStream({
+    context: messagesFromChat(body.messages, body.model), model: body.model, tools: body.tools, toolChoice: body.tool_choice,
   });
+  const { events: streamResult, model } = completion;
   if (body.stream) {
     response.writeHead(200, { "content-type": "text/event-stream", "cache-control": "no-cache", connection: "keep-alive" });
     sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: { role: "assistant" }, finish_reason: null }] });
-    for await (const event of streamResult) {
-      if (event.type === "text_delta") sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: { content: event.delta }, finish_reason: null }] });
-      if (event.type === "toolcall_end") sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: { tool_calls: [{ index: event.contentIndex, id: event.toolCall.id, type: "function", function: { name: event.toolCall.name, arguments: JSON.stringify(event.toolCall.arguments) } }] }, finish_reason: null }] });
-      if (event.type === "done") sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: {}, finish_reason: event.reason === "toolUse" ? "tool_calls" : event.reason }] });
-      if (event.type === "error") throw new Error(event.error.errorMessage ?? "Model request failed");
+    let toolIndex = 0;
+    try {
+      for await (const event of streamResult) {
+        if (event.type === "text_delta") sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: { content: event.delta }, finish_reason: null }] });
+        if (event.type === "toolcall_end") sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: { tool_calls: [{ index: toolIndex++, id: event.toolCall.id, type: "function", function: { name: event.toolCall.name, arguments: JSON.stringify(event.toolCall.arguments) } }] }, finish_reason: null }] });
+        if (event.type === "done") sse(response, { id, object: "chat.completion.chunk", created: Math.floor(Date.now() / 1000), model, choices: [{ index: 0, delta: {}, finish_reason: event.reason === "toolUse" ? "tool_calls" : event.reason }] });
+        if (event.type === "error") throw new Error(event.error.errorMessage ?? "Model request failed");
+      }
+    } catch (error) {
+      sse(response, { error: { message: error.message ?? String(error), type: "service_error" } });
     }
     debug("response SSE", "[DONE]");
     response.end("data: [DONE]\n\n");
@@ -96,19 +102,47 @@ function responseObject(id, model, message, status = "completed") {
 
 async function responses(request, response, body) {
   const id = `resp_${randomUUID().replaceAll("-", "")}`;
-  const model = body.model || defaultModel;
-  const streamResult = await createCompletionStream({ context: messagesFromResponses(body.input), model, tools: body.tools, toolChoice: body.tool_choice });
+  const completion = await createCompletionStream({ context: messagesFromResponses(body.input, body.model), model: body.model, tools: body.tools, toolChoice: body.tool_choice });
+  const { events: streamResult, model } = completion;
   if (body.stream) {
     response.writeHead(200, { "content-type": "text/event-stream", "cache-control": "no-cache", connection: "keep-alive" });
     const created = { id, object: "response", created_at: Math.floor(Date.now() / 1000), status: "in_progress", model, output: [] };
-    sse(response, { type: "response.created", response: created });
+    sse(response, { type: "response.created", sequence_number: 0, response: created }, true);
     let sequenceNumber = 0;
     let outputIndex = 0;
-    for await (const event of streamResult) {
-      if (event.type === "text_delta") sse(response, { type: "response.output_text.delta", sequence_number: sequenceNumber++, output_index: 0, content_index: 0, delta: event.delta });
-      if (event.type === "toolcall_end") sse(response, { type: "response.output_item.done", sequence_number: sequenceNumber++, output_index: outputIndex++, item: { type: "function_call", id: event.toolCall.id, call_id: event.toolCall.id, name: event.toolCall.name, arguments: JSON.stringify(event.toolCall.arguments), status: "completed" } });
-      if (event.type === "done") sse(response, { type: "response.completed", sequence_number: sequenceNumber++, response: responseObject(id, model, event.message) });
-      if (event.type === "error") throw new Error(event.error.errorMessage ?? "Model request failed");
+    let textStarted = false;
+    let text = "";
+    let textItem;
+    try {
+      for await (const event of streamResult) {
+        if (event.type === "text_delta") {
+          if (!textStarted) {
+            textStarted = true;
+            textItem = { id: `msg_${randomUUID().replaceAll("-", "")}`, type: "message", status: "in_progress", role: "assistant", content: [] };
+            sse(response, { type: "response.output_item.added", sequence_number: ++sequenceNumber, output_index: outputIndex, item: textItem }, true);
+            sse(response, { type: "response.content_part.added", sequence_number: ++sequenceNumber, output_index: outputIndex, item_id: textItem.id, content_index: 0, part: { type: "output_text", text: "", annotations: [] } }, true);
+          }
+          text += event.delta;
+          sse(response, { type: "response.output_text.delta", sequence_number: ++sequenceNumber, output_index: outputIndex, item_id: textItem.id, content_index: 0, delta: event.delta }, true);
+        }
+        if (event.type === "toolcall_end") {
+          const item = { type: "function_call", id: event.toolCall.id, call_id: event.toolCall.id, name: event.toolCall.name, arguments: JSON.stringify(event.toolCall.arguments), status: "completed" };
+          sse(response, { type: "response.output_item.added", sequence_number: ++sequenceNumber, output_index, item: { ...item, status: "in_progress", arguments: "" } }, true);
+          sse(response, { type: "response.function_call_arguments.done", sequence_number: ++sequenceNumber, output_index, item_id: item.id, arguments: item.arguments }, true);
+          sse(response, { type: "response.output_item.done", sequence_number: ++sequenceNumber, output_index: outputIndex++, item }, true);
+        }
+        if (event.type === "done") {
+          if (textStarted) {
+            sse(response, { type: "response.output_text.done", sequence_number: ++sequenceNumber, output_index, item_id: textItem.id, content_index: 0, text }, true);
+            sse(response, { type: "response.content_part.done", sequence_number: ++sequenceNumber, output_index, item_id: textItem.id, content_index: 0, part: { type: "output_text", text, annotations: [] } }, true);
+            sse(response, { type: "response.output_item.done", sequence_number: ++sequenceNumber, output_index: outputIndex++, item: { ...textItem, status: "completed", content: [{ type: "output_text", text, annotations: [] }] } }, true);
+          }
+          sse(response, { type: "response.completed", sequence_number: ++sequenceNumber, response: responseObject(id, model, event.message) }, true);
+        }
+        if (event.type === "error") throw new Error(event.error.errorMessage ?? "Model request failed");
+      }
+    } catch (error) {
+      sse(response, { type: "error", sequence_number: ++sequenceNumber, error: { message: error.message ?? String(error), type: "service_error" } }, true);
     }
     debug("response SSE", "<stream closed>");
     response.end();
@@ -118,19 +152,19 @@ async function responses(request, response, body) {
 }
 
 async function ask({ prompt, model = defaultModel, tools, toolChoice }) {
-  const streamResult = await createCompletionStream({
+  const completion = await createCompletionStream({
     context: messagesFromResponses(prompt),
     model,
     tools,
     toolChoice,
   });
-  const message = await collect(streamResult);
-  return { text: contentFromMessage(message), toolCalls: toolCallsFromMessage(message), model };
+  const message = await collect(completion.events);
+  return { text: contentFromMessage(message), toolCalls: toolCallsFromMessage(message), model: completion.model };
 }
 
 async function mcp(response, body) {
   const base = { jsonrpc: "2.0", id: body.id };
-  if (body.method === "initialize") return sendJson(response, 200, { ...base, result: { protocolVersion: body.params?.protocolVersion ?? "2025-03-26", capabilities: { tools: {} }, serverInfo: { name: "service-llm", version: "0.1.0" } } });
+  if (body.method === "initialize") return sendJson(response, 200, { ...base, result: { protocolVersion: body.params?.protocolVersion ?? "2025-03-26", capabilities: { tools: {} }, serverInfo: { name: "service-llm", version } } });
   if (body.method === "notifications/initialized") return response.writeHead(202).end();
   if (body.method === "ping") return sendJson(response, 200, { ...base, result: {} });
   if (body.method === "tools/list") return sendJson(response, 200, { ...base, result: { tools: [{ name: "ask", description: "Ask a Codex model a question and optionally allow it to return function tool calls. Tool calls are never executed.", inputSchema: { type: "object", properties: { prompt: { type: "string" }, model: { type: "string" }, tools: { type: "array", items: { type: "object" } }, tool_choice: {} }, required: ["prompt"], additionalProperties: false } }] } });
@@ -197,9 +231,27 @@ export function startServer({ port = 7060, host = "127.0.0.1", apiKeys = [] } =
       sendJson(response, error.status ?? 500, { error: { message: error.message ?? String(error), type: "service_error" } });
     }
   });
-  server.listen(port, host, () => {
-    console.error(`service-llm listening on http://${host}:${port}`);
-    if (debugEnabled) console.error("[service-llm] DEBUG logging enabled");
+  server.listen(port, host, async () => {
+    const baseUrl = `http://${host}:${port}`;
+    console.error(`service-llm v${version}`);
+    console.error(`Listening: ${baseUrl}`);
+    console.error(`Client authentication: ${allowedKeys.size ? `enabled (${allowedKeys.size} key${allowedKeys.size === 1 ? "" : "s"})` : "disabled"}`);
+    try {
+      const catalog = await listModels();
+      console.error(`Default model: ${catalog.default_model} (${catalog.default_reasoning_effort ?? "low"} reasoning)`);
+      console.error(`Supported models: ${catalog.models.length} (run 'service-llm model' to list)`);
+    } catch (error) {
+      console.error(`Model catalog: unavailable (${error.message ?? String(error)})`);
+    }
+    console.error("Endpoints:");
+    console.error(`  Health:           ${baseUrl}/health`);
+    console.error(`  Chat Completions: ${baseUrl}/v1/chat/completions`);
+    console.error(`  Responses:        ${baseUrl}/v1/responses`);
+    console.error(`  MCP:              ${baseUrl}/mcp`);
+    console.error(`Debug logging: ${debugEnabled ? "enabled" : "disabled"}`);
+    if (!allowedKeys.size && !["127.0.0.1", "localhost", "::1"].includes(host)) {
+      console.error("WARNING: The service is listening beyond localhost without client authentication.");
+    }
   });
   return server;
 }

package/dist/index.js

@@ -1,3 +1,4 @@
-export { authPath, getAccessToken, login } from "./auth.js";
+export { authPath, getAccessToken, listModels, login, modelsPath, setDefaultModel } from "./auth.js";
 export { defaultModel } from "./backend.js";
 export { startServer } from "./http.js";
+export { version } from "./version.js";

package/dist/version.js

@@ -0,0 +1,3 @@
+import { readFileSync } from "node:fs";
+
+export const version = readFileSync(new URL("../VERSION", import.meta.url), "utf8").trim();

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@leoustc/service-llm",
-  "version": "0.1.0",
-  "description": "Expose a Codex subscription through OpenAI-compatible and MCP endpoints.",
+  "version": "0.2.1",
+  "description": "Standalone CLI service exposing a Codex subscription through OpenAI-compatible and MCP endpoints.",
   "type": "module",
   "main": "./dist/index.js",
   "bin": {
@@ -11,7 +11,8 @@
     ".": "./dist/index.js"
   },
   "files": [
-    "dist"
+    "dist",
+    "VERSION"
   ],
   "scripts": {
     "build": "node ./scripts/build.mjs",
@@ -27,5 +28,5 @@
   "publishConfig": {
     "access": "public"
   },
-  "license": "MIT"
+  "license": "Apache-2.0"
 }