Skip to content

feat: let the codex backend run independent tool calls in parallel #38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
TheGreatAxios merged 2 commits into from
Nov 6, 2026
Merged

feat: let the codex backend run independent tool calls in parallel #38

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ff729b2
build(deps): require @corbits/openai-responses ^0.2.4
TheGreatAxios Nov 6, 2026
2a75576
feat: let the codex backend run independent tool calls in parallel
TheGreatAxios Nov 6, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,13 +5,13 @@ OpenAI Codex ("Login with ChatGPT") for `@intx/inference`: the OAuth client conf
## Why @corbits/codex-provider?

1. **Inference on a ChatGPT subscription.** Codex has no API-key path. This package supplies the PKCE login config, the code exchange and refresh, and the `chatgpt-account-id` the backend requires on every request.
2. **The exact request Codex accepts.** `store: false`, `parallel_tool_calls: false`, no `max_output_tokens`, the host prompt as a leading `developer` message, and the Codex CLI's `originator` header are fixed in the adapter, not left to each host.
2. **The exact request Codex accepts.** `store: false`, `parallel_tool_calls: true`, no `max_output_tokens`, the host prompt as a leading `developer` message, and the Codex CLI's `originator` header are fixed in the adapter, not left to each host.
3. **Backend bugs repaired, host identity required.** `withCodexContentTypeRepair` restores the `content-type` the backend drops on some SSE streams. The host names itself through `CodexQuirks`; there is no default product name.

## Install

```bash
bun add @corbits/codex-provider @corbits/oauth-core@^0.2.0 @corbits/openai-responses@^0.2.0 @intx/inference@^0.4.0 @intx/types@^0.4.0
bun add @corbits/codex-provider @corbits/oauth-core@^0.2.0 @corbits/openai-responses@^0.2.4 @intx/inference@^0.4.0 @intx/types@^0.4.0
```

Runs on Bun >= 1.2 or Node >= 24.
Expand Down
6 changes: 3 additions & 3 deletions bun.lock
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

4 changes: 2 additions & 2 deletions package.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,7 +53,7 @@
},
"devDependencies": {
"@corbits/oauth-core": "0.3.4",
"@corbits/openai-responses": "0.2.3",
"@corbits/openai-responses": "0.2.4",
"@intx/inference": "0.5.0",
"@intx/types": "0.5.0",
"@types/bun": "1.3.14",
Expand All @@ -63,7 +63,7 @@
},
"peerDependencies": {
"@corbits/oauth-core": "^0.3.4",
"@corbits/openai-responses": "^0.2.3",
"@corbits/openai-responses": "^0.2.4",
"@intx/inference": "^0.5.0",
"@intx/types": "^0.5.0"
},
Expand Down
8 changes: 4 additions & 4 deletions src/responses-adapter.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,8 +39,8 @@ function parseCodexBody(rawBody: string) {
// chatgpt-account-id header only rides when the option is set, session_id
// rides in both the header and the body, `instructions` is omitted (the
// host's own system prompt leads `input` as a wrapped developer message),
// and the three fields the backend requires explicit values for (`store`,
// `parallel_tool_calls`, `max_output_tokens`) are on the wire correctly.
// `store` and `max_output_tokens` are set as the backend requires, and
// `parallel_tool_calls` is on.
describe("Codex responses adapter — request shape", () => {
test("throws when the quirks bag is absent — there is no honest default product name", () => {
expect(() => createCodexResponsesAdapter(source, undefined)).toThrow();
Expand Down Expand Up @@ -104,13 +104,13 @@ describe("Codex responses adapter — request shape", () => {
expect(bridgeText).toContain("operate carefully");
});

test("always sends store:false and parallel_tool_calls:false, and omits max_output_tokens", () => {
test("always sends store:false and parallel_tool_calls:true, and omits max_output_tokens", () => {
const adapter = createCodexResponsesAdapter(source, quirks);
const request = adapter.buildRequest(turns, "gpt-5.5", { maxTokens: 4096 });
const body = parseCodexBody(request.body);

expect(body.store).toBe(false);
expect(body.parallel_tool_calls).toBe(false);
expect(body.parallel_tool_calls).toBe(true);
expect(body.max_output_tokens).toBeUndefined();
});

Expand Down
6 changes: 3 additions & 3 deletions src/responses-adapter.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,8 +15,8 @@ import {
import { wrapCodexBridgeMessage } from "./instructions.js";
import { parseCodexQuirks, type CodexQuirks } from "./quirks.js";

// `store` and `parallelToolCalls` need an explicit `false` on the wire, and
// `maxOutputTokens: false` because the backend 400s on `max_output_tokens`.
// `store` needs an explicit `false` on the wire, and `maxOutputTokens: false`
// omits `max_output_tokens`, which the backend 400s on.
function codexResponsesQuirks(quirks: CodexQuirks): ResponsesQuirks {
return {
path: CODEX_RESPONSES_PATH,
Expand All @@ -37,7 +37,7 @@ function codexResponsesQuirks(quirks: CodexQuirks): ResponsesQuirks {
systemPrompt: { role: "developer", shape: "parts" },
contentShape: "typed",
store: false,
parallelToolCalls: false,
parallelToolCalls: true,
maxOutputTokens: false,
reasoning: {
effortOption: CODEX_REASONING_EFFORT_OPTION,
Expand Down
Loading