Skip to content

Add support for /v1/messages API and refactor SessionServer #1893

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
braisedpork1964 wants to merge 9 commits into
base: main
Choose a base branch
from
Open

Add support for /v1/messages API and refactor SessionServer #1893

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
f53a984
refactor SessionServer
braisedpork1964 Jun 5, 2026
b892628
drop anthropic server-side tools
braisedpork1964 Jun 8, 2026
a8d9618
format
braisedpork1964 Jun 9, 2026
aba7530
disable role demotion
braisedpork1964 Jun 10, 2026
a0723f5
drop the orphaned placeholder from trie on exception
braisedpork1964 Jun 11, 2026
30a17cd
remove the orphaned placeholder on null output_token_ids
braisedpork1964 Jun 25, 2026
52446c0
ensure trace-store nodes never carry a null expert_key via single-pha…
braisedpork1964 Jul 9, 2026
a6fc67d
emit format-native error payload so dropped turns aren't recorded dow…
braisedpork1964 Jul 9, 2026
bad6d5f
fix typos
braisedpork1964 Jul 13, 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
3 changes: 3 additions & 0 deletions xtuner/v1/rl/rollout/chat_template.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,9 @@ def canonicalize_messages_for_chat_template(messages: list[dict]) -> list[dict]:
template-renderable shape before ``apply_chat_template``. It must not
mutate model responses returned to the agent/client, nor the generated
token ids, labels, or raw rollout artifacts used for training.

``tool_calls[].function.arguments`` strings are parsed to dicts so the
template's ``tojson`` filter sees a real object.
"""

messages = copy.deepcopy(messages)
Expand Down
Loading
Loading