xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-17 22:27:26 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge 24617166ce2edf9ce5e78e98673a74d3f3880d38 into 254f6b986720c92ddf97fbb1a6a6465da8e87e29

Browse Source
This commit is contained in:
ゆり 2025-12-25 00:07:00 +00:00 committed by GitHub
commit 08e40e44c8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 18 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
docs/features/tool_calling.md
View File

@ -396,6 +396,24 @@ Supported models:
Flags: `--tool-call-parser gigachat3`
### GPT-OSS Models (`openai`)
OpenAI's GPT-OSS models use a special token-based format (Harmony format) for tool calling that is parsed using the `openai` tool parser.
Supported models:
* `openai/gpt-oss-120b`
* `openai/gpt-oss-20b`
Known limitations:
* Only `tool_choice='auto'` is currently supported
* Streaming tool calls have limited support
Flags: `--tool-call-parser openai --enable-auto-tool-choice`
For more detailed examples and usage patterns, see the [GPT-OSS cookbook](https://docs.vllm.ai/projects/recipes/en/latest/OpenAI/GPT-OSS.html).
### Models with Pythonic Tool Calls (`pythonic`)
A growing number of models output a python list to represent tool calls instead of using JSON. This has the advantage of inherently supporting parallel tool calls and removing ambiguity around the JSON schema required for tool calls. The `pythonic` tool parser can support such models.