Skip to content

Use Claude Code, Codex, VSCode, Pi, and OpenCode (and 6 other harnesses) for free (1.3B+ free tokens) from your terminal, app, IDE, or phone, and now from the browser with native browser sessions (multi-harness + multi-model) like OpenClaw (voice supported + ToS friendly)

Python

Restore OpenAI subscription generation compatibility (#1326)

Ali Khokharfccf572746a1
diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
index 0ea25948..b73b1e6d 100644
--- a/ARCHITECTURE.md
+++ b/ARCHITECTURE.md
@@ -657,8 +657,15 @@ failure classification, model discovery, admission, and commit-boundary retry.
Neutral Anthropic-to-Responses input and Responses-to-Anthropic stream
conversion remain in
[core/openai_responses/](src/free_claude_code/core/openai_responses/), which
-never imports OAuth, account IDs, or provider endpoints. The Admin API exposes
-only safe connected-account state and never serializes token objects.
+never imports OAuth, account IDs, or provider endpoints. That neutral converter
+preserves the public Responses contract; the subscription provider owns the
+private Codex request projection and omits fields that backend does not accept,
+such as the public `max_output_tokens` cap and `metadata`. Successful private
+responses can omit `Content-Type`, so the provider accepts an absent media type
+for its always-streaming request and parses the body as SSE. An explicitly
+declared non-SSE media type retains the bounded diagnostic failure path. The
+Admin API exposes only safe connected-account state and never serializes token
+objects.
[providers/google_openai/](src/free_claude_code/providers/google_openai/) owns the
Google-specific protocol behavior shared by AI Studio and Vertex AI: literal
diff --git a/pyproject.toml b/pyproject.toml
index 2e62e3c7..5bc5dd4a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
[project]
name = "free-claude-code"
-version = "4.16.4"
+version = "4.16.5"
description = "Local proxy connecting coding agents to OpenAI-compatible AI providers"
readme = "README.md"
requires-python = ">=3.14.0"
diff --git a/src/free_claude_code/providers/openai_codex/provider.py b/src/free_claude_code/providers/openai_codex/provider.py
index 1c44b1ad..53623cdc 100644
--- a/src/free_claude_code/providers/openai_codex/provider.py
+++ b/src/free_claude_code/providers/openai_codex/provider.py
@@ -91,7 +91,7 @@ class OpenAICodexProvider(BaseProvider):
*,
reasoning: ReasoningPolicy = DEFAULT_REASONING_POLICY,
) -> None:
- """Validate lossless request conversion before any upstream I/O."""
+ """Validate and adapt the private Codex request before upstream I/O."""
self._build_body(request, reasoning=reasoning)
@@ -152,9 +152,14 @@ class OpenAICodexProvider(BaseProvider):
reasoning: ReasoningPolicy,
) -> dict[str, Any]:
try:
- return build_responses_provider_request(request, reasoning=reasoning)
+ body = build_responses_provider_request(request, reasoning=reasoning)
except ResponsesConversionError as exc:
raise InvalidRequestError(str(exc)) from exc
+ # The private Codex backend rejects these public Responses fields.
+ # Codex itself omits the output cap and uses separate internal metadata.
+ body.pop("max_output_tokens", None)
+ body.pop("metadata", None)
+ return body
async def _run_stream(
self,
@@ -232,8 +237,8 @@ class OpenAICodexProvider(BaseProvider):
truncated=body_truncated,
)
raise
- content_type = response.headers.get("content-type", "")
- if "text/event-stream" not in content_type.lower():
+ content_type = response.headers.get("content-type")
+ if content_type and "text/event-stream" not in content_type.lower():
body_bytes, body_truncated = await _read_bounded_body(response)
error = _TruncatedResponsesStream(
"OpenAI returned a non-streaming Responses payload."
diff --git a/tests/api/test_openai_codex_compatibility.py b/tests/api/test_openai_codex_compatibility.py
index e647f51f..30ed614b 100644
--- a/tests/api/test_openai_codex_compatibility.py
+++ b/tests/api/test_openai_codex_compatibility.py
@@ -62,8 +62,7 @@ async def test_messages_accepts_current_claude_controls_for_openai_provider() ->
upstream_requests.append(request)
return httpx.Response(
200,
- text=_complete_stream("hello"),
- headers={"content-type": "text/event-stream"},
+ content=_complete_stream("hello").encode(),
request=request,
)
@@ -101,6 +100,7 @@ async def test_messages_accepts_current_claude_controls_for_openai_provider() ->
"model": "openai/gpt-test",
"max_tokens": 1024,
"messages": [{"role": "user", "content": "hello"}],
+ "metadata": {"user_id": "example-user"},
"thinking": {"type": "adaptive", "display": "omitted"},
"context_management": {
"edits": [
@@ -126,5 +126,7 @@ async def test_messages_accepts_current_claude_controls_for_openai_provider() ->
payload = json.loads(upstream_requests[0].content)
assert payload["model"] == "gpt-test"
assert payload["reasoning"] == {"effort": "high", "summary": "auto"}
+ assert "max_output_tokens" not in payload
+ assert "metadata" not in payload
assert "context_management" not in payload
assert "output_config" not in payload
diff --git a/tests/providers/test_openai_codex_provider.py b/tests/providers/test_openai_codex_provider.py
index c4cd6618..30731434 100644
--- a/tests/providers/test_openai_codex_provider.py
+++ b/tests/providers/test_openai_codex_provider.py
@@ -12,6 +12,7 @@ from free_claude_code.core.anthropic.stream_contracts import (
parse_sse_text,
text_content,
)
+from free_claude_code.core.diagnostics import ERROR_DETAIL_DISPLAY_CAP_BYTES
from free_claude_code.core.failures import ExecutionFailure
from free_claude_code.core.reasoning import ReasoningEffort, ReasoningPolicy
from free_claude_code.providers.admission import ProviderAdmissionController
@@ -77,6 +78,7 @@ def _client_control_request() -> MessagesRequest:
"model": "gpt-test",
"max_tokens": 1024,
"messages": [{"role": "user", "content": "hello"}],
+ "metadata": {"user_id": "example-user"},
"thinking": {"type": "adaptive", "display": "omitted"},
"context_management": {
"edits": [
@@ -151,8 +153,7 @@ async def test_provider_uses_subscription_headers_and_visible_model_catalog() ->
assert request.url.path.endswith("/responses")
return httpx.Response(
200,
- text=_complete_stream("hello"),
- headers={"content-type": "text/event-stream"},
+ content=_complete_stream("hello").encode(),
request=request,
)
@@ -215,6 +216,7 @@ async def test_provider_accepts_claude_client_controls_before_upstream_io() -> N
_config(), auth=_FakeAuth(), admission=_admission(), client=client
)
request = _client_control_request()
+ original_request = request.model_dump()
reasoning = ReasoningPolicy.on(effort=ReasoningEffort.HIGH)
provider.preflight_stream(request, reasoning=reasoning)
@@ -235,8 +237,11 @@ async def test_provider_accepts_claude_client_controls_before_upstream_io() -> N
payload = json.loads(upstream.content)
assert payload["model"] == "gpt-test"
assert payload["reasoning"] == {"effort": "high", "summary": "auto"}
+ assert "max_output_tokens" not in payload
+ assert "metadata" not in payload
assert "context_management" not in payload
assert "output_config" not in payload
+ assert request.model_dump() == original_request
assert_anthropic_stream_contract(parse_sse_text(body))
await client.aclose()
@@ -487,13 +492,20 @@ async def test_pre_stream_retry_discards_held_message_start() -> None:
@pytest.mark.asyncio
async def test_non_streaming_success_is_retried_then_reports_bounded_body() -> None:
attempts = 0
+ tail_sentinel = "must-not-escape-the-diagnostic-cap"
+ upstream_body = (
+ "upstream contract changed\n"
+ + "x" * ERROR_DETAIL_DISPLAY_CAP_BYTES
+ + tail_sentinel
+ ).encode()
def handler(request: httpx.Request) -> httpx.Response:
nonlocal attempts
attempts += 1
return httpx.Response(
200,
- json={"error": {"message": "upstream contract changed"}},
+ content=upstream_body,
+ headers={"content-type": "application/json"},
request=request,
)
@@ -516,6 +528,10 @@ async def test_non_streaming_success_is_retried_then_reports_bounded_body() -> N
assert attempts > 1
assert "upstream contract changed" in exc_info.value.message
+ assert f"truncated after {ERROR_DETAIL_DISPLAY_CAP_BYTES} bytes" in (
+ exc_info.value.message
+ )
+ assert tail_sentinel not in exc_info.value.message
assert "Request ID: req_non_stream" in exc_info.value.message
await client.aclose()
diff --git a/uv.lock b/uv.lock
index abf1567f..ff0f7c54 100644
--- a/uv.lock
+++ b/uv.lock
@@ -620,7 +620,7 @@ wheels = [
[[package]]
name = "free-claude-code"
-version = "4.16.4"
+version = "4.16.5"
source = { editable = "." }
dependencies = [
{ name = "aiohttp" },