Skip to content

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

Rust

stream, accumulate: map max_output_tokens cutoff to stop_reason max_tokens

Raine Virtada8e92fef094
diff --git a/src/translate/accumulate.ts b/src/translate/accumulate.ts
index f60a17b..cbf2c8b 100644
--- a/src/translate/accumulate.ts
+++ b/src/translate/accumulate.ts
@@ -45,6 +45,7 @@ export async function accumulateResponse(
>()
const itemToOutputIndex = new Map<string, number>()
let sawToolUse = false
+ let incomplete = false
let usage: any
const orderedIndices: number[] = []
@@ -113,8 +114,12 @@ export async function accumulateResponse(
if (b && b.kind === "tool" && !b.args && typeof item?.arguments === "string") b.args = item.arguments
continue
}
- if (t === "response.completed") {
+ if (t === "response.completed" || t === "response.incomplete") {
usage = p.response?.usage
+ const reason = p.response?.incomplete_details?.reason
+ if (t === "response.incomplete" || reason === "max_output_tokens" || p.response?.status === "incomplete") {
+ incomplete = true
+ }
continue
}
}
@@ -142,7 +147,7 @@ export async function accumulateResponse(
role: "assistant",
model: opts.model,
content,
- stop_reason: sawToolUse ? "tool_use" : "end_turn",
+ stop_reason: incomplete ? "max_tokens" : sawToolUse ? "tool_use" : "end_turn",
stop_sequence: null,
usage: {
input_tokens: usage?.input_tokens ?? 0,
diff --git a/src/translate/stream.ts b/src/translate/stream.ts
index 40c174b..4e40375 100644
--- a/src/translate/stream.ts
+++ b/src/translate/stream.ts
@@ -73,6 +73,7 @@ async function runStream(
const itemToOutputIndex = new Map<string, number>()
let sawToolUse = false
let finalUsage: CodexUsage | undefined
+ let incomplete = false
let rateLimitReached = false
let rateLimitResetSeconds: number | undefined
@@ -249,8 +250,12 @@ async function runStream(
continue
}
- if (type === "response.completed") {
+ if (type === "response.completed" || type === "response.incomplete") {
finalUsage = payload.response?.usage
+ const reason = payload.response?.incomplete_details?.reason
+ if (type === "response.incomplete" || reason === "max_output_tokens" || payload.response?.status === "incomplete") {
+ incomplete = true
+ }
continue
}
@@ -281,7 +286,7 @@ async function runStream(
emit("message_delta", {
type: "message_delta",
delta: {
- stop_reason: sawToolUse ? "tool_use" : "end_turn",
+ stop_reason: incomplete ? "max_tokens" : sawToolUse ? "tool_use" : "end_turn",
stop_sequence: null,
},
usage,