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

Recover from quoted output-token caps (#1362)

Ali Khokhar554c7d978779
diff --git a/pyproject.toml b/pyproject.toml
index 5b63d43e..4c7e9b86 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
[project]
name = "free-claude-code"
-version = "4.17.3"
+version = "4.17.4"
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_chat/output_cap.py b/src/free_claude_code/providers/openai_chat/output_cap.py
index 08cea7e7..cfdd4b5d 100644
--- a/src/free_claude_code/providers/openai_chat/output_cap.py
+++ b/src/free_claude_code/providers/openai_chat/output_cap.py
@@ -24,14 +24,15 @@ _OUTPUT_TOKEN_FIELDS = ("max_completion_tokens", "max_tokens")
_OUTPUT_TOKEN_KEYWORDS = ("max_completion_tokens", "max_tokens")
# Comparator phrases that precede the allowed maximum in provider error text.
+_CAP_VALUE_PATTERN = r"[`'\"]?(\d+)[`'\"]?"
_CAP_PATTERNS: tuple[re.Pattern[str], ...] = (
- re.compile(r"less than or equal to\s+(\d+)"),
- re.compile(r"smaller than or equal to\s+(\d+)"),
- re.compile(r"<=\s*(\d+)"),
- re.compile(r"at most\s+(\d+)"),
- re.compile(r"must not exceed\s+(\d+)"),
- re.compile(r"maximum(?:\s+value)?(?:\s+for\s+\S+)?\s+is\s+(\d+)"),
- re.compile(r"maximum(?:\s+allowed)?(?:\s+value)?\s+of\s+(\d+)"),
+ re.compile(rf"less than or equal to\s+{_CAP_VALUE_PATTERN}"),
+ re.compile(rf"smaller than or equal to\s+{_CAP_VALUE_PATTERN}"),
+ re.compile(rf"<=\s*{_CAP_VALUE_PATTERN}"),
+ re.compile(rf"at most\s+{_CAP_VALUE_PATTERN}"),
+ re.compile(rf"must not exceed\s+{_CAP_VALUE_PATTERN}"),
+ re.compile(rf"maximum(?:\s+value)?(?:\s+for\s+\S+)?\s+is\s+{_CAP_VALUE_PATTERN}"),
+ re.compile(rf"maximum(?:\s+allowed)?(?:\s+value)?\s+of\s+{_CAP_VALUE_PATTERN}"),
)
diff --git a/tests/providers/test_openai_chat_output_cap.py b/tests/providers/test_openai_chat_output_cap.py
index b9420dc9..e03d8235 100644
--- a/tests/providers/test_openai_chat_output_cap.py
+++ b/tests/providers/test_openai_chat_output_cap.py
@@ -36,10 +36,11 @@ class _BadRequest(Exception):
def test_parse_cap_from_groq_message():
error = _BadRequest(
- "max_completion_tokens must be less than or equal to 40960, the maximum "
- "value for max_completion_tokens is less than the context_window for this model"
+ "`max_completion_tokens` must be less than or equal to `16384`, the maximum "
+ "value for `max_completion_tokens` is less than the `context_window` for this "
+ "model"
)
- assert parse_output_token_cap(error) == 40960
+ assert parse_output_token_cap(error) == 16384
@pytest.mark.parametrize(
@@ -136,7 +137,18 @@ async def test_create_stream_clamps_and_learns_on_cap_rejection(groq_provider):
assert body["max_completion_tokens"] == 64000
model = body["model"]
- error = _BadRequest("max_completion_tokens must be less than or equal to 40960")
+ error = _BadRequest(
+ "invalid request",
+ body={
+ "message": (
+ "`max_completion_tokens` must be less than or equal to `16384`, "
+ "the maximum value for `max_completion_tokens` is less than the "
+ "`context_window` for this model"
+ ),
+ "type": "invalid_request_error",
+ "param": "max_completion_tokens",
+ },
+ )
create = AsyncMock(side_effect=[error, object()])
with patch.object(groq_provider._client.chat.completions, "create", create):
@@ -147,9 +159,9 @@ async def test_create_stream_clamps_and_learns_on_cap_rejection(groq_provider):
await attempt.aclose()
assert create.call_count == 2
- assert create.call_args_list[1].kwargs["max_completion_tokens"] == 40960
- assert used_body["max_completion_tokens"] == 40960
- assert groq_provider._model_output_caps[model] == 40960
+ assert create.call_args_list[1].kwargs["max_completion_tokens"] == 16384
+ assert used_body["max_completion_tokens"] == 16384
+ assert groq_provider._model_output_caps[model] == 16384
@pytest.mark.asyncio
diff --git a/uv.lock b/uv.lock
index 2b0c7939..3f6ccdeb 100644
--- a/uv.lock
+++ b/uv.lock
@@ -620,7 +620,7 @@ wheels = [
[[package]]
name = "free-claude-code"
-version = "4.17.3"
+version = "4.17.4"
source = { editable = "." }
dependencies = [
{ name = "aiohttp" },