[Bug]: treat `ERROR: Upstream response failed` as retryable for openai/codex

[danpiths]

Bug Description

so openai has these errors come up when their server is overloaded.

half the times they give the server_is_overloaded error messages and the other half of the times it says that upstream response failed.

it has a message within it that says, An error occurred while processing your request. You can retry a request or contact us through our help center., in which case I think we should retry because this is a very similar thing to server_is_overloaded.

since the server comes back online after a point when this error is sent by them, stopping the whole request, when this is happening causes some issues with the agent delegation flows that happen in forgecode. it would be great for this to be retried so that the sub-agents keep working and don't fail.

ERROR: Upstream response failed
    Caused by: {"message":"An error occurred while processing your request. You can retry your request, or contact us through our help center at help.openai.com if the error persists. Please include the request ID ccb50bdf-aa84-4a5d-a95b-543382f18054 in your message.","code":"server_error"}

Steps to Reproduce

1. make a request to openai when there are a lot of server_is_overloaded errors going on
2. openai at times will repond back with An error occurred while processing your request. You can retry your request, or contact us through our help center at help.openai.com if the error persists
3. this causes you request to stop there and not be retried
   a. will cause problems with agent delegation flows where the subagent stops midway and cannot be resumed from the point it stopped at and thus the whole work needs to be done again.

Expected Behavior

it should be retried

Actual Behavior

it fails and stops

Forge Version

forge 2.13.19

Operating System & Version

macOS Tahoe 26.5.2

AI Provider

OpenAI

Model

gpt-5.6-sol on LOW

Installation Method

Other

Configuration

"$schema" = "https://forgecode.dev/schema.json"

max_search_lines = 1000
max_search_result_bytes = 10240
max_fetch_chars = 50000
max_stdout_prefix_lines = 100
max_stdout_suffix_lines = 100
max_stdout_line_chars = 500
max_line_chars = 2000
max_read_lines = 2000
max_file_read_batch_size = 50
max_file_size_bytes = 104857600
max_image_size_bytes = 262144
tool_timeout_secs = 300
auto_open_dump = false
max_conversations = 100
max_sem_search_results = 100
sem_search_top_k = 10
services_url = "https://api.forgecode.dev/"
max_extensions = 15
max_parallel_file_reads = 64
model_cache_ttl_secs = 604800
use_forge_committer = true
max_commit_count = 20
top_p = 0.8
top_k = 30
max_tokens = 20480
max_tool_failure_per_turn = 3
max_requests_per_turn = 100
restricted = true
tool_supported = true
currency_symbol = "$"
currency_conversion_rate = 1.0
verify_todos = true
use_text_patch_fallback = false
research_subagent = true
subagents = true
auto_install_vscode_extension = false
merge_system_messages = false

[retry]
initial_backoff_ms = 200
min_delay_ms = 1000
backoff_factor = 2
max_attempts = 8
status_codes = [
    429,
    500,
    502,
    503,
    504,
    408,
    522,
    524,
    520,
    529,
]
suppress_errors = false

[http]
connect_timeout_secs = 30
read_timeout_secs = 900
pool_idle_timeout_secs = 90
pool_max_idle_per_host = 5
max_redirects = 10
hickory = false
tls_backend = "default"
adaptive_window = true
keep_alive_interval_secs = 60
keep_alive_timeout_secs = 10
keep_alive_while_idle = true
accept_invalid_certs = false

[session]
provider_id = "codex"
model_id = "gpt-5.6-sol"

[commit]
provider_id = "codex"
model_id = "gpt-5.5"

[suggest]
provider_id = "codex"
model_id = "gpt-5.5"

[updates]
frequency = "daily"
auto_update = false

[compact]
retention_window = 6
eviction_window = 0.2
max_tokens = 2000
token_threshold = 100000
message_threshold = 200
on_turn_end = false

[reasoning]
effort = "low"
enabled = true

[danpiths]

@amitksingh1490 sorry to ping but if you could just add this as retryable, it would be of great help, since openai servers have been overloaded since the past 2 days and my subagents have to restart the whole work again and again eating a lot into my usage 🙇🏻

if you think me opening up a PR for this would help, lmk and i can get the PR to you as well!

thanks!

[amitksingh1490]

sure will fix this