Skip to content

fix(ai): GPT-5 fails when streaming — retry with max_completion_tokens on 400 #2774

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

Open
midozy wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix(ai): GPT-5 fails when streaming — retry with max_completion_tokens on 400 #2774

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
b788443
fix(ai): retry streaming OpenAI/Custom requests with max_completion_t…
midozy Oct 12, 2026
a83e02b
fix(ai): raise non-streaming chat timeout to 300s for reasoning models
midozy Oct 12, 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
21 changes: 18 additions & 3 deletions Strand/AI/Providers/Custom.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,10 +61,24 @@ struct CustomClient: AIProviderClient {
req.setValue("application/json", forHTTPHeaderField: "Content-Type")
req.httpBody = try JSONSerialization.data(withJSONObject: body)

try await performStreamingRequest(req, session: session) { payload in
if let delta = SseDeltas.openAiDelta(payload) {
onDelta(delta)
do {
try await performStreamingRequest(req, session: session) { payload in
if let delta = SseDeltas.openAiDelta(payload) {
onDelta(delta)
}
}
} catch let AICoachError.server(code, detail) where code == 400 {
// GPT-5 / reasoning models reject `max_tokens` + `temperature`. The 400 arrives before
// any delta, so fall back to `send` (which retries with `max_completion_tokens`) and
// emit the full reply as one delta.
let d = detail.lowercased()
guard d.contains("max_completion_tokens") || d.contains("max_tokens")
|| d.contains("temperature") || d.contains("unsupported") else {
throw AICoachError.server(code, detail)
}
let reply = try await send(key: key, model: model, systemPrompt: systemPrompt,
messages: messages, session: session)
onDelta(reply)
}
}

Expand Down Expand Up @@ -150,6 +164,7 @@ struct CustomClient: AIProviderClient {

var req = URLRequest(url: AIProvider.custom.endpoint)
req.httpMethod = "POST"
req.timeoutInterval = 300 // reasoning models (GPT-5) can sit silent >60s before the first byte
AIProvider.applyCustomAuthHeader(key, to: &req)
req.setValue("application/json", forHTTPHeaderField: "Content-Type")
req.httpBody = try JSONSerialization.data(withJSONObject: body)
Expand Down
21 changes: 18 additions & 3 deletions Strand/AI/Providers/OpenAI.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,10 +51,24 @@ struct OpenAIClient: AIProviderClient {
req.setValue("application/json", forHTTPHeaderField: "Content-Type")
req.httpBody = try JSONSerialization.data(withJSONObject: body)

try await performStreamingRequest(req, session: session) { payload in
if let delta = SseDeltas.openAiDelta(payload) {
onDelta(delta)
do {
try await performStreamingRequest(req, session: session) { payload in
if let delta = SseDeltas.openAiDelta(payload) {
onDelta(delta)
}
}
} catch let AICoachError.server(code, detail) where code == 400 {
// GPT-5 / reasoning models reject `max_tokens` + `temperature`. The 400 arrives before
// any delta, so fall back to `send` (which retries with `max_completion_tokens`) and
// emit the full reply as one delta.
let d = detail.lowercased()
guard d.contains("max_completion_tokens") || d.contains("max_tokens")
|| d.contains("temperature") || d.contains("unsupported") else {
throw AICoachError.server(code, detail)
}
let reply = try await send(key: key, model: model, systemPrompt: systemPrompt,
messages: messages, session: session)
onDelta(reply)
}
}

Expand Down Expand Up @@ -97,6 +111,7 @@ struct OpenAIClient: AIProviderClient {

var req = URLRequest(url: AIProvider.openAI.endpoint)
req.httpMethod = "POST"
req.timeoutInterval = 300 // reasoning models (GPT-5) can sit silent >60s before the first byte
req.setValue("Bearer \(key)", forHTTPHeaderField: "Authorization")
req.setValue("application/json", forHTTPHeaderField: "Content-Type")
req.httpBody = try JSONSerialization.data(withJSONObject: body)
Expand Down
Loading