Skip to content

Use Claude Code with your ChatGPT, Kimi, Cursor or Grok subscription via a local Anthropic-compatible proxy

Rust

readme: recommend disabling non-streaming fallback

Raine Virtaeb4b4c5b3b9d
diff --git a/README.md b/README.md
index 2bd2e31..c400ce9 100644
--- a/README.md
+++ b/README.md
@@ -107,6 +107,7 @@ ANTHROPIC_AUTH_TOKEN=unused \
ANTHROPIC_MODEL=gpt-5.4[1m] \
ANTHROPIC_SMALL_FAST_MODEL=gpt-5.4-mini[1m] \
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 \
+CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK=1 \
claude
# Kimi
@@ -115,9 +116,16 @@ ANTHROPIC_AUTH_TOKEN=unused \
ANTHROPIC_MODEL=kimi-for-coding[1m] \
ANTHROPIC_SMALL_FAST_MODEL=kimi-for-coding[1m] \
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 \
+CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK=1 \
claude
```
+`CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK=1` is recommended because the
+proxy always talks to upstream providers with streaming requests, even when it
+accumulates a non-streaming Anthropic response for Claude Code. Disabling Claude
+Code's streaming-to-non-streaming fallback avoids retrying a partially completed
+stream in a way that can duplicate tool calls.
+
Or set it persistently in `~/.claude/settings.json`:
```json
@@ -127,7 +135,8 @@ Or set it persistently in `~/.claude/settings.json`:
"ANTHROPIC_AUTH_TOKEN": "unused",
"ANTHROPIC_MODEL": "gpt-5.4[1m]",
"ANTHROPIC_SMALL_FAST_MODEL": "gpt-5.4-mini[1m]",
- "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1
+ "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1,
+ "CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK": 1
}
}
```
@@ -169,6 +178,7 @@ if [ -f "$HOME/.claude/claude-code-proxy-enabled" ]; then
export ANTHROPIC_MODEL="gpt-5.4[1m]"
export ANTHROPIC_SMALL_FAST_MODEL="gpt-5.4-mini[1m]"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"
+ export CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK="1"
fi
exec "$HOME/.local/bin/claude" "$@"